@            D IR EC TI O NS I N DE VE L O P M E N T
|A *
~~~~~~~~- j,;,gi,j,/iilea''G1..MNX 
* !Lmj 0 0 .s ..<I>
*              _                     X,,j:EUjAi l~~~~~~~~~~~~~~~~~~~~~~~~~~~44






DIRECTIONS IN DEVELOPMENT
Development Finance
Institutions
Measuring Their Subsidy






DIRECTIONS IN DEVELOPMENT
Development Finance
Institutions
Measuring Their Subsidy
Mark Schreiner
Jacob Yaron
THE WORLD BANK
WASHINGTON, D.C.



Copyright © 2001 The International Bank for Reconstruction
and Development/THE WORLD BANK
1818 H Street, N.W.
Washington, D.C. 20433, USA
All rights reserved
Manufactured in the United States of America
First printing October 2001
1 2 3 4 04 03 02 01
The findings, interpretations, and conclusions expressed in this book are entirely those
of the authors and should not be attributed in any manner to the World Bank, to its affil-
iated organizations, or to members of its Board of Executive Directors or the countries
they represent. The World Bank does not guarantee the accuracy of the data included
in this publication and accepts no responsibility for any consequence of their use.
The material in this publication is copyrighted. The World Bank encourages dis-
semination of its work and will normally grant permission to reproduce portions of
the work promptly.
Permission to photocopy items for internal or personal use, for the internal or per-
sonal use of specific clients, or for educational classroom use is granted by the World
Bank, provided that the appropriate fee is paid directly to the Copyright Clearance
Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA; telephone 978-750-8400,
fax 978-750-4470. Please contact the Copyright Clearance Center before photocopying
items.
For permission to reprint individual articles or chapters, please fax a request with
complete information to the Republication Department, Copyright Clearance Center,
fax 978-750-4470.
All other queries on rights and licenses should be addressed to the Office of the
Publisher, World Bank, at the address above or faxed to 202-522-2422.
Cover photo credit: Curt Carnemark, 1994; Morocco
Cover background: Tomas Sennett, undated; Brazil
Library of Congress Cataloging-in-Publication Data
Schreiner, Mark, 1969-
Development finance institutions: measuring their subsidy / Mark Schreiner,
Jacob Yaron.
p. , cm. -  (Directions in development)
Includes bibliographical references.
ISBN 0-8213-4984-8
1. Development credit corporations.  2. Development banks.  3. Subsidies.
4. Externalities (Economics)  1. Yaron, Jacob. II. Title.  111. Series.
HG3726 .S34 2001
332.2'8-dc2l                                            2001045424



Contents
Abstract ...............................................                                            viii
Acknowledgments   ................................................x
Glossary of Acronyms ................................................                                  xi
Glossary  of Notation  ................................................                               xii
Introduction  ................................................1
What Is the Subsidy Dependence Index? ................................................2
What Is Net Present Cost to Society? ................................................              6
Is the Subsidy Dependence Index Redundant? .......................................8
What Does the Rest of the Monograph Cover? ........................................9
1. Why Measure the Social Cost of Public Development
Finance Institutions? ...............................................                             10
What Is a Public Development Finance Institution? ............................. 10
Who Bears the Costs and Who Reaps the Benefits of
Public Development Finance Institutions? ......................................... 11
How Do Public Development Finance Institutions
Differ from  Other Public Projects? .............................................                12
Why Does Society Subsidize Public Development
Finance Institutions? .............................................                            12
How Can Society Measure the Benefits of Public
Development Finance Institutions? .............................................                  14
What Is the Opportunity Cost to Society of Public Funds
Used by Development Finance Institutions? ...................................... 16
Why Does the Measurement of Costs Boost Performance? ................. 21
v



vi                                              DEVELOPMENT FINANCE INSTITUTIONS
2. What Is a Measure of the Social Cost of a Public Development
Finance Institution in the Short Term?  ............................................... 22
How Does a Development Finance Institution Get Subsidies? .......... 22
What Forms of Subsidized Funds Does a Development
Finance Institution Get? .....................................................               23
Numerical Examples of the Subsidy Dependence Index .................... 32
Is Subsidy in the Subsidy Dependence Index Related to a
Subsidy-Adjusted Return on Equity? .................................................. 43
How Does the Subsidy Dependence Index Change as
Its Parts Change? ....................................................                       46
3. What Is a Measure of the Social Cost of a Public Development
Finance Institution  in the Long Term? ................................................. 53
How Does the Net Present Cost to Society Discount Flows? ............. 53
What Questions Does the Net Present Cost to Society Inform? ......... 55
What Are the Net Present Cost to Society and the Subsidy
Dependence Index in the Long Term for the
Example Development Finance Institution? ...................................... 58
4. What Are the Pitfalls When Calculating the Subsidy
Dependence Index or Net Present Cost to Society? ........................... 62
What Is the Social Opportunity Cost? .................................................... 62
What Can Be Done to Cope with Accounting Data? ........................... 62
What Are Other Pitfalls and Caveats? .................................................... 68
What Are the Key Caveats? .....................................................                 73
5. Recent Proposed Changes to the Subsidy Dependence Index ....... 74
The Subsidy Dependence Ratio of Khandker ........................................ 74
The Profitability Gap of Sacay ....................................................               79
The Average Subsidy Dependence Index of Hulme and Mosley ....... 81
Appendix: A Framework to Approximate the Opportunity
Costs of Private Entities ....................................................                      84
What Is the Price of Private Debt? ....................................................    84
What Is the Price of Private Equity? .................................................... 86
References ....................................................                                     90



CONTENTS                                                                                                 vii
Tables
2.1.   Types of Subsidized  Funds ...............................................                        23
2.2.   Balance Sheet ...............................................                                     34
2.3.   Income Statement ...............................................                                   35
2.4.   Calculation  of the Subsidy Dependence Index ................................ 36
2.5. Alternative Calculation of the Subsidy
Dependence Index ...............................................                               38
2.6.   ROE, SAROE, ROA, and  SAROA  ...............................................                        47
3.1.  Net Present Cost to Society ...............................................                        59
4.1.   Balance Sheet with  Loan  Losses ...............................................                    65
4.2.  Income Statement with  Loan  Losses ...............................................  66
4.3.  Summary with  Loan  Losses ...............................................                          67
A.1.  Private Opportunity  Costs ...............................................                          85
A.2. Subsidy Dependence Index with Private
Opportunity  Costs ...............................................                             88
Figures
1.1.   Five Possible Proxies of the Social Opportunity  Cost ..................... 18
2.1.   Profit Grants and  ROE ...............................................                             25
2.2.   Grameen: Inflation  and  Nominal and  Real Yields ........................... 31
2.3.   BancoSol: Inflation  and  Nominal and  Real Yields ........................... 31
2.4.   ROE  versus SAROE  for the Example DFI ......................................... 46
2.5.   ROA  versus SAROA  for the Example DFI ........................................ 46
2.6.  The SDI and  the Yield on Loans, i ...............................................                   49
2.7.   The SDI and  the Rate Paid  for Public Debt, c ................................... 49
2.8.  The SDI and  the Social Opportunity  Cost, m  ................................... 50
2.9.  The SDI and Administrative Expenses .............................................. 51
2.10. The SDI and  the Ratio of Deposits to Public Debt .......................... 52
Boxes
1.1.   Social Cost Is the Road Not Taken ................................................                  3
1.2.   Economic Value Added, Subsidy  for For-Profit Firms ......................4
1.3. Social Worthwhileness, Subsidy Independence, Private
Profitability, and  Self-Sustainability ................................................7
2.1.   How   Profit Grants Affect Profit and  Return  on Equity .................. 25
2.2.   Real Yields at Grameen  and  BancoSol ............................................... 31
2.3. The Subsidy Dependence Index and Average Equity at
Bank Rakyat Indonesia ...............................................                          42
2.4. The Subsidy Dependence Index and Subsidy-Adjusted
Return  on Equity  for an African  DFI .............................................. 48



Abstract
Measures of the social cost of development finance institutions (DFIs)
that receive public funds help to check whether DFIs are good uses of
public funds. Public funds are well-spent-and social welfare is
improved-if the social benefit of a DFI exceeds the social cost. The term
development finance institution encompasses not only government devel-
opment banks but also thousands of nongovernmental microfinance
organizations worldwide that use matching grants to attempt to promote
community development, decentralization of power, and local empower-
ment. This monograph describes the measurement of costs but not of
benefits, but even without precise knowledge of benefits, knowledge of
costs can help to spend funds well. It is less expensive to measure costs
than benefits, and "cost calculations can provide a useful 'reality check'
on proposed interventions. Whatever the true [unknown] size of external
benefits, the government must judge that at a minimum the external ben-
efit exceeds this cost for the intervention to be worth undertaking"
(Devarajan, Squire, and Suthiwart-Narueput 1997, p. 40).
This monograph presents two measures of social cost. The first is the
Subsidy Dependence Index (SDI). The SDI does not discount flows, so it
works best in short time frames or when the rate of time preference is low.
The SDI is the ratio of subsidy received to revenue from loans. Subsidy is
defined as the social cost of the public funds used to run a DFI. The mea-
sure of subsidy can also be used to adjust common measures of financial
performance such as Return on Equity (ROE) or Return on Assets (ROA).
The second measure is the Net Present Cost to Society (NPCS). Like
standard present-value measures, the NPC5 discounts cash flows and
works in any time frame.
The SDI and the NPCS are useful because common financial ratios-
such as ROE-may hide the true performance of DFIs because their mea-
sures of cost may not reflect social opportunity costs. The SDI and the
NPCS shift the paradigm from reported (accounting) costs-much of
which are routinely subsidized-to opportunity (economic) costs. The
two measures proposed here use standard tools of project analysis to
viii



ABSTRACT                                                      ix
answer questions from the point of view of society. The questions and
answers also matter to governments and donors who care about sustain-
ability. A sustainable DFI can meet its goals now and in the long term.
Sustainability improves social welfare if the consequent long-term
increase in the length, breadth, scope, and quality of outreach compen-
sate for the short-term increases in costs shifted to the target group. By
definition, a subsidy-independent DFI has no social cost in financial terms.
The SDI and the NPC5 are simple tools, and their results are only as
good as their data and assumptions. Like other yardsticks, they help to
establish benchmarks, chart trends, and compare a DFI with peers with
identical clients and services.
The measurement of the social cost of public DFIs matters because
funds earmarked for development are scarce. Subsidies for DFIs are not
bad unless they could improve social welfare more somewhere else. The
measurement of social cost as described in this monograph is a first step
toward wiser use of public funds.



Acknowledgments
We are thankful for comments from Stephanie Charitonenko, Carlos
Cuevas, Douglas Graham, Michael Lyne, Jonathan Morduch, Sergio
Navajas, Glenn Pederson, Richard Rosenberg, Michael Sherraden, and
participants in seminars at The Ohio State University. The opinions
expressed are not necessarily those of Washington University in St. Louis
nor of the World Bank.
x



Glossary of Acronyms
BancoSol     Banco Solidario, S.A.
BRI          Bank Rakyat Indonesia
CBA          Cost-benefit analysis
CEA          Cost-effectiveness analysis
CGAP         Consultative Group to Assist the Poorest
CNCA         Caisse Nationale de Credit Agricole
EVA          Economic Value Added
DFI          Development finance institution
GAAP         Generally accepted accounting principles
IADB         Inter-American Development Bank
IAS          International accounting standard
NGO          Nongovernmental organization
NPCS         Net Present Cost to Society
PG           Profitability Gap
ROA          Return on Assets
ROE          Return on Equity
SAROA        Subsidy-Adjusted Return on Assets
SAROE        Subsidy-Adjusted Return on Equity
SDI          Subsidy Dependence Index in a short time frame
SDIL         Subsidy Dependence Index in a long time frame
SDR          Subsidy Dependence Ratio
xi



Glossary of Notation
c            Rate paid for public debt by a DFI
A            Average public debt
A (i(m - c)  Discount on public debt
8            Social discount rate
d            Deposit interest rate
Dep          Average deposit liabilities
DG           Direct grants
DX           Discount on expenses
E            Average equity
EG           Equity grant
Yield on loans
I            Average investments
j            Yield on investments
IC           Yield received by DFI on required reserves
k            Reserve requirement
L            Leverage
LP           Average loan portfolio (net)
LP- i        Revenue from loans
m            Opportunity cost of public funds in nominal terms to
society or the opportunity cost of public debt to a
private entity
M            Opportunity cost of public equity to a private entity
n            Age of a DFI in years
P            Accounting profit
Q            Quantity of public funds
It           Inflation rate
PC           Paid-in capital
R            Nominal rate of interest
r            Real rate of interest
RG           Revenue grant
S            Subsidy
t            Index of years since the start of a time frame
T            Years in a time frame
TP           True profit
xii



Introduction
Why measure the social cost of public development finance institutions?
Governments and donors for decades have tried to improve social welfare
through public support for development finance institutions (DFIs).
Public DFIs, as with all projects that use public funds, are worthwhile in
principle only if their social benefits exceed their social costs. In practice,
it is so expensive to measure social benefits that a full-blown social cost-
benefit analysis (CBA) cannot be done each time a choice must be made to
spend public funds on a DFI. A less-expensive alternative to CBA is a sim-
ple measure of social cost. Social cost is defined as the opportunity cost of
public resources used by a DFI. It is the opportunity cost to society of the
public funds used by a DFI less what the DFI could pay for those funds
and still show a profit. A DFI with no social cost is subsidy-independent.
This monograph presents two measures of social cost. The first is the
Subsidy Dependence Index (SDI) proposed by Yaron (1992a, 1992b). The
second is the Net Present Cost to Society (NPCS) proposed by Schreiner
(1997) for the flows of public funds between society and a DFI. The SDI
works in short time frames such as a year. Like all other standard present-
value measures, the NPCS works in all time frames because it discounts
resource flows according to when they take place in time.
Common financial measures such as accounting profit or Return on
Equity (ROE) are based on prices paid as recorded in the accounts of the
DFI, and these prices may reflect market failures or nonmarket rules of
governments or donors. In contrast, the SDI and the NPCS are based on
social opportunity costs, the social return those funds would earn in their
best use outside of DFIs. This paradigm shift matters because the prices of
public funds entrusted to a DFI as loans or equity are almost always set
far below the social opportunity cost. Likewise, DFIs often record grants
in cash as revenues and/or fail to record grants in kind as expenses. This
practice lards profits but does not change business performance. Such
accounting window dressing can hide the truth about the use of public
funds by a DFI and its economic subsidy dependence. Generally accepted
accounting principles (GAAP) take the opportunity cost of equity as zero
1



2                            DEVELOPMENT FINANCE INSTITUTIONS
and ignore both low levels of inflation and the time value of money. Thus,
a DFI can boast of an accounting profit even as inflation shrinks its net
worth in real terms. A positive ROE does not always mean that a DFI can
compensate society for the opportunity cost of public funds.
The SDI and the NPC5 resolve the problems of accounting-based mea-
sures because they value funds at their opportunity costs. They help to
check whether a DFI uses public funds to increase or to decrease social
welfare. A DFI increases social welfare only if the net benefits from its use
of public funds exceed the net benefits from their use elsewhere.
Measurement of the social cost of public support for DFIs matters
because public funds are scarce. New measures-such as the SDI and the
NPCS-are needed because the old ones fail to measure social costs well
because they were designed for private firms, not for public DFIs. The
goal of the measurement of social cost is not to end subsidies for DFIs;
rather, the goal is to put a price tag on DFIs to make sure that subsidies
for DFIs are the best way to improve social welfare.
Appropriate measurements of performance matter because DFIs use a
big chunk of the development budget. For example, the World Bank had
loaned more than $30 billion (all dollar amounts are U.S. dollars unless
specified) for credit projects by 1989 (Von Pischke 1991). Even if donor
lending for DFIs dwindles-which seems unlikely given the myriad spe-
cialized-credit programs worldwide-the SDI and NPCS are still useful
as measures of the social cost of the explosion of public support for micro-
finance DFIs during the past decade. Microfinance DFIs outnumber tra-
ditional public-sector DFIs; a 1996 survey of more than 200 microfinance
DFIs found 13 million loans worth $7 billion and 45 million deposit
accounts worth $19 billion (Paxton 1996). Some microfinance advocates
hope to attract more than $21.6 billion to extend microfinance to 100 mil-
lion families in the next 10 years (RESULTS International 1996). It
behooves society to check whether the crusade for microfinance siphons
funds away from better ways to improve welfare (box 1.1; Mosley and
Hulme 1998; Buckley 1997; Rogaly 1996).
What Is the Subsidy Dependence Index?
Yaron (1992a, 1992b) proposed a two-part framework of outreach and
sustainability that has become the most common tool to measure the per-
formance of DFIs (e.g., Gonzalez-Vega and others 1997; Khandker 1996;
Chaves and Gonzalez-Vega 1996; Christen and others 1995; Benjamin
1994; Yaron 1994; Hossain 1988).
The SDI is a summary measure of sustainability. It is the ratio of sub-
sidy received by a DFI to revenue from loans to the target group and
indicates whether a DFI could compensate society for the opportunity



INTRODUCTION                                                        3
Box I.1. Social Cost Is the Road Not Taken
The social cost of public resources used in a DFI is the benefit lost because
those resources were not used in another project. For example, the social
cost as measured by the SDI of the Caisse Nationale de Credit Agricole
(CNCA), a rural development bank in Morocco, was about $85 million a
year. According to Devarajan, Squire, and Suthiwart-Narueput (1997, p. 40),
these subsidies "could conceivably be justified on the grounds that the
bank operated in an underserved rural credit market and reached poor
people. Although these benefits are hard to quantify, assessing the cost of
the subsidy is one way to ask whether this subsidy is a good use of scarce
public resources and to think about alternative uses. In this case, CNCA's
annual subsidy amounted to about 20 percent of the recurrent budget for
primary education and 160 percent of the recurrent budget for basic health
care. And this in a country where social indicators were quite unsatisfacto-
ry-primary enrollment is around 70 percent, and under-five mortality is
about 80 deaths per thousand live births." Even without a full-blown CBA,
the SDI can help to guide the best mix of public investments among, as in
this example, agricultural credit, education, and preventative health care.
cost of public funds used in a short time frame and still show a profit.
Such a DFI is called subsidy-independent.
The SDI looks at social cost. The second half of the framework-out-
reach-looks at social benefit. Outreach has six aspects (Schreiner 1999a):
worth to users, cost to users, breadth, length, depth, and scope of the out-
put of a DFI.
The SDI relates to subsidy in two ways. First, the SDI provides a frame-
work to measure subsidy as the social opportunity cost of the public
funds held by the DFI in a short time frame such as one year, minus the
price the DFI paid, minus (plus) accounting profit (loss). Thus, subsidy is
the implicit "rental cost" of public resources, minus the rented funds that
were lost and thus cannot be returned. Subsidy is positive for a subsidy-
dependent DFI and negative for a subsidy-independent DFI.
Second, the SDI is a ratio that uses the measurement of subsidy as its
numerator and revenue from loans as its denominator. The ratio can be
seen as the percentage change in the yield on loans that, all else constant,
would make the DFI subsidy-independent. It may also be seen as the
matching grant (subsidy in the numerator) awarded to the DFI by society
for each dollar of interest and fees paid by borrowers (revenue from loans
in the denominator).
The subsidy measure in the SDI can also be transformed into a
Subsidy-Adjusted ROE (SAROE). As will be shown later, the two mea-



4                               DEVELOPMENT FINANCE INSTITUTIONS
sures are equivalent in that the SDI is negative if and only if an SAROE
would exceed the social opportunity cost. The measure of subsidy in the
SDI has features like those of Economic Value Added (EVA), a popular
new measure of the financial performance of for-profit firms (box 1.2).
Box I.2. Economic Value Added, Subsidy for For-Profit Firms
The concept of the measurement of performance as opportunity cost less
profit is not unique to DFIs, having long been a staple in the analysis of not-
for-profit hospitals (Jennings 1993; Wheeler and Clement 1990; Silvers and
Kauer 1986; Pauly 1986; Conrad 1984, 1986). Nor is the concept unique to
not-for-profits. For-profit firms-'lost in ever darker muddles of account-
ing" (Tully 1993)-have turned to measures based on opportunity cost
because measures such as accounting profit and ROE do not tell owners
whether a firm increases private wealth, just as they do not tell policymak-
ers whether a DFI increases social welfare.
Subsidy in the SDI is analogous to the concept of Economic Value Added
(EVA), a new performance measure used by for-profit firms (The Economist
1997). EVA is after-tax profit minus the economic cost of funds used. If EVA
is positive, then the firm created financial value for its owners; likewise, if
subsidy is negative, then the DFI created financial value for society.
EVA is useful to stockholders because "stock prices track EVA far more
closely than they track such popular measures as earnings per share or
operating margins or ROE. That is because EVA shows what investors real-
ly care about-the net cash return on their capital-rather than some other
type of performance viewed through the often distorting lens of accounting
rules" (Tully 1993).
Wal-Mart, Coca-Cola, AT&T, and Proctor & Gamble use EVA because,
unlike standard accounting measures, EVA accounts for the total cost of
capital. One analyst said, "Capital looks free to a lot of managers. It doesn't
look free to investors who hand them the money" (Tully 1993). Just as EVA
reminds managers of for-profit firms of opportunity costs to investors, the
SDI reminds managers of DFIs of opportunity costs to society.
Like the SDI, a strength of EVA is its ease of use. Better measures-such
as Net Present Value for stockholders and the Net Present Cost to Society-
use discounting, but they are used less because they are more complex.
Measurement with the SDI or EVA boosts performance and ratchets
standards up a notch. One CFO said, "The effect is staggering. 'Good' is no
longer positive operating earnings. It's only when you beat the cost of cap-
ital" (Tully 1993).
Like the SDI, EVA "is powerful and widely applicable because in the end
it doesn't prescribe doing anything.... Instead, it is a way to see and under-
stand what is really happening" (Tully 1993).



INTRODUCTION                                                5
What Are the Strengths of the Subsidy Dependence Index?
The SDI has at least 12 strengths.
* The SDI quantifies subsidy and shows the extent of subsidy depen-
dence. Often governments and donors do not know just how much
their support for DFIs costs society because much of the subsidy is not
in terms of explicit cash flows from the public purse to the DFI.
Knowledge of subsidies is needed to compare support for DFIs with
other uses of public funds.
* The SDI compares subsidy with revenue from loans. This ratio can be
seen as a matching grant; that is, the amount of subsidy awarded to the
DFI by society for each dollar of interest paid by borrowers.
* The SDI is a measure of subsidy dependence through time. Whether or
not a DFI can declare complete subsidy independence, it can always
strive to improve.
* A negative SDI implies an SAROE higher than the social opportunity
cost. This patches the weaknesses in the common framework based on
unadjusted ROE.
* The SDI shifts the paradigm from accounting costs to opportunity
costs because accounting costs are often distorted by subsidies.
• The SDI highlights the possibility of covering costs with revenue from
loans.
* Although the SDI does not measure benefits, which is expensive, it
does measure costs, which is less expensive.
* The SDI is simple-if the financial data of the DFI conform with
GAAP-and well known.
* The use of the SDI can induce a disciplined approach to the judgment
of the social costs of public support for DFIs.
* Because the data needed for the SDI should be easy to extract, the use
of the SDI can highlight specific improvements that should be intro-
duced in the accounting systems.
* The SDI can help in the analysis of the sources and uses of subsidy
(Yaron 1992b, p. 24).
* The SDI-unlike the SAROE-worsens if DFIs keep profit constant but
shift resources away from loans to the target group and toward other
investments such as government bonds.
What Are the Limitations of the Subsidy Dependence Index?
The SDI has at least two limitations. Analysts should know them so that
they use the SDI only to answer the question to which it applies. The SDI
answers an important question-whether a DFI could compensate soci-



6                            DEVELOPMENT FINANCE INSTITUTIONS
ety for the opportunity cost of its funds and still show a profit-but it
does not answer all the important questions.
First, the SDI does not discount flows of funds. This is not a difficulty
in short time frames (such as one year) with low inflation. But not all time
frames are short, and inflation can be high. For example, suppose gov-
ernments or donors need to choose whether to start a new DFI from
scratch. To inform this choice, they might ask whether an existing DFI
would have been judged as subsidy-independent from its birth had its
eventual performance been known at its birth. Or they might want to
plan their support so that projected performance in a long time frame
meets a goal (Helms 1997). After all, newborn DFIs, just like all newborn
firms, lose money until time and growth spread start-up costs and hone
technology. Like private investors who judge firms by their Net Present
Value, governments and donors must judge DFIs not only in their first
year, not only in the most recent year, and not only in the next year, but
rather all through their whole lifetimes. Of course, pro forma data may
have a wide margin of error, but society should follow the lead of private
investors, who find that explicit present-value analysis is useful even if
based on data of doubtful quality.
Second, the SDI indicates subsidy independence but not self-sustain-
ability (box I.3). A subsidy-independent DFI could pay the social opportu-
nity cost of its funds and still show a profit, and a self-sustainable DFI can
meet its goals now and in the long term. Subsidy independence may not
guarantee self-sustainability. For example, private opportunity costs may
exceed social opportunity costs, so a subsidy-independent DFI might not
be able to pay market prices for private funds and still show a profit
should sources of public funds dry up. Also, for example, a subsidy-inde-
pendent DFI may fail to meet its goals in the long term if it drifts from its
development mission. Of course, the SDI-like all performance mea-
sures-indicates subsidy independence in the past and not in the future.
Past performance is not a guarantee of future results.
What Is Net Present Cost to Society?
The NPCs answers the question: What benefits did society lose because it
entrusted public funds to a DFI rather than to some other project? Like the
SDI, the NPC5 uses the social opportunity cost. Unlike the SDI, the NPCs
discounts flows. Discounting matters more and more as a time frame
lengthens.
The NPCs complements the SDI. To match the practice of the SDI, the
NPCs adds financial flows from society to the DFI and subtracts financial
flows from the DFI to society, so the NPCs is the negative of Net Present
Value, a basic yardstick in finance and economics.



INTRODUCTION                                                         7
Box 1.3. Social Worthwhileness, Subsidy Independence,
Private Profitability, and Self-Sustainability
The four concepts of social worthwhileness, subsidy independence, private
profitability, and self-sustainability are distinct (Schreiner 1997). A socially
worthwhile DFI has social benefits that exceed costs in present-value terms.
A subsidy-independent DFI could pay the social opportunity cost of public
funds and still show a profit. A privately profitable DFI could pay the private
opportunity cost of all funds and still show a profit. A self-sustainable DFI
could meet its goals now and in the long term.
Social worthwhileness matters because public support for DFIs aims to
improve social welfare. Subsidy independence matters because, if cus-
tomers benefit from a DFI and if there are no external social costs, then zero
social cost implies social worthwhileness. Private profitability matters
because if public funds are limited then DFIs will be few and small unless
private investors use their own funds to buy DFIs or to start new ones from
scratch. Finally, self-sustainability matters because society cares about
improved welfare both now and in the future.
Subsidy independence is necessary and sufficient for private profitabili-
ty only if the social opportunity cost equals or exceeds the private oppor-
tunity cost. Private profitability is needed, however, for self-sustainability.
Privately profitable DFIs may also improve social welfare more than sub-
sidy-dependent DFIs (Schreiner 1999a; Mosley and Hulme 1998; Chaves
and Gonzalez-Vega 1996; Rosenberg 1996; Yaron 1994). Privately profitable
DFIs may also attract private funds and thus produce more development
finance at less cost to the public purse (Rosenberg 1994).
While private profitability is needed for self-sustainability, it does not
guarantee it. Self-sustainability also requires a host of other nonfinancial
qualities such as organizational strength, efficient technology, a consistent
structure of incentives to give stakeholders reasons to act in the interests of
the mission of the DFI, and rules that build in flexibility to adjust through
time (Schmidt 1997; Schreiner 1995). An investor-whether public or pri-
vate-that contemplates the purchase of a DFI should check more than just
past financial performance because future success depends greatly on
intangible, nonfinancial assets.
Both Net Present Value and the NPCS answer the same questions and
are derived from standard benefit-cost theory (Gittinger 1982). The two
measures have the same magnitude and opposite signs. If customers get
more benefits than costs and if noncustomers do not bear any costs, then a
DFI started from scratch that is equivalent to a DFI with a negative NPCS
would be expected to be a good social investment. Likewise, a DFI with a
negative NPCS seen from now on is a good social investment from now on.



8                           DEVELOPMENT FINANCE INSTITUTIONS
Wise use of the NPCS recognizes three facts. First, the NPCs is not just
the sum of SDIs through a span of years. Second, a negative NPC5 as seen
from now on does not necessarily imply a negative NPCS as seen from
birth. Because past costs are sunk, public support for a DFI from now on
may make sense even though it would not have made sense had future
performance been known at the time of birth. Third, the NPCs ignores
benefits and costs to noncustomers. If these benefits exceed these costs,
then a DFI with a positive NPCS (or a positive SDI) could still be a good
use of public funds.
Is the Subsidy Dependence Index Redundant?
Unlike the SDI, the NPCS discounts flows. Thus, the NPCS measures
social cost better than the SDI, especially in long time frames or when the
social rate of time preference is high. All else constant, society would be
better off if it judged DFIs not only with the SDI but also with the NPCS,
especially in long time frames. Still, the SDI is far from redundant for
three reasons. First, the SDI is slightly easier to compute than the one-
year case of the NPCS. Second, the extra accuracy caused by discounting
in the one-year case of the NPCs may be dwarfed by the inaccuracies of
the basic data and by the coarse assumptions used by both the SDI and
the NPCS in the attempt to fit data based on GAAP accounting into a
economic framework. Third and most important, many people already
are familiar with ROE, so the SDI is useful in its guise as a Subsidy-
Adjusted ROE.
Because the NPCS discounts cash flows and the SDI does not, the two
measures do not answer the same question. Society might ask about
social cost (or the Subsidy-Adjusted ROE) of a DFI in a short time frame,
and the answer is contained in the SDI. Society might also ask about
social cost in long time frames, and the answer is the NPCs. Both ques-
tions and answers matter, and both the SDI and the NPCS are the right
tools for their own distinct purposes.
Neither the SDI nor the NPCs answers all questions about the perfor-
mance of a DFI. The two measures inform some questions, but they do
not inform all questions, nor do they fully inform any single question.
Like all financial ratios, the SDI and the NPCS do not tell directly why
performance is good or bad, nor do they tell directly how to improve.
Other quantitative indicators and qualitative analysis still have a role in
the full assessment of the performance of public DFIs. In particular, fur-
ther analysis and even full-blown benefit-cost analysis might be desirable
in some circumstances despite its high cost.



INTRODUCTION                                                9
What Does the Rest of the Monograph Cover?
Chapter 1 discusses why DFIs exist and why society would want to mea-
sure their performance.
Chapter 2 presents the formula of the SDI in terms of the basic
accounts of a set of financial statements for an example DFI. It also has
numerical examples and discusses what the SDI means and shows how
the measure of subsidy in the SDI can also be transformed into a Subsidy-
Adjusted ROE.
Chapter 3 derives the NPCS and gives numerical examples. It also
shows that the one-year case of the NPCS is not the same as the SDI.
Chapter 4 emphasizes that the SDI and the NPCs are only as good as
their data and assumptions. It notes pitfalls in their calculation, especial-
ly the need to adjust financial statements to reflect the repayment risk of
outstanding loans and to purge the effects of inflation.
Chapter 5 reviews three recent attempts to modify the SDI or to use
other standards to judge the performance of public DFIs. It makes explicit
the questions answered by these new proposals and argues that their use
does not lead to a better understanding of the social cost of public DFIs.



1
Why Measure the Social Cost
of Public Development
Finance Institutions?
The measurement of the social cost of public DFIs matters because pub-
lic funds budgeted for development are scarce. The poor can use loans
and deposits, but they can also use more and/or better food, water, air,
health, clothes, houses, schools, roads, fuels, skills, tools, laws, markets,
and/or safety. The SDI and the NPCS are two measures of social cost.
Neither the SDI nor the NPCS is equivalent to social CBA, but both mea-
sures are linked to self-sustainability and to social welfare.
Oversight of public DFIs is needed because the people who work for
governments and donors and who choose to support DFIs with public
funds do not bear most of the costs and benefits of that choice. Instead,
costs and benefits accrue to taxpayers (because they provide public
funds), customers of the DFI (from the use of services), and noncustomers
(from displacement by customers and from the loss of benefits from pro-
jects left unfunded). Oversight is also needed because DFIs tempt gov-
ernments and donors more than most development projects because they
involve self-help not with gifts but with loans (Mosley and Hulme 1998).
Measurement of the social cost of public DFIs aims to help to align pri-
vate incentives with the public good.
What Is a Public Development Finance Institution?
A public DFI is a financial intermediary that aims to improve social wel-
fare and that gets some resources from governments or donors. A public
DFI may be owned by the state and thus receive public resources as equi-
ty, but it may also have private owners (or no owners) and receive public
resources as gifts or loans. Public funds entrusted to a DFI are subsidized
because the unfettered market would charge more. If not, then the DFI
would refuse public funds and go straight to the market on its own. By
the same logic, a DFI subsidizes its clients. If services from a public DFI
were costlier than identical services from the market, then clients would
eschew the DFI.
10



WHY MEASURE THE SOCIAL COST OF PUBLIC DFIs?                 11
Who Bears the Costs and Who Reaps the Benefits
of Public Development Finance Institutions?
Society-all the people in the world or all the people in a country-bears
the cost of public DFIs. Subsidies to one person are taxes to another.
Furthermore, funds spent on a DFI are funds not spent to improve wel-
fare in some other way. The direct or primary benefits of a public DFI
accrue to its clients. Although indirect or secondary costs and benefits to
nonclients and to the employees of the DFI, governments, and donors
may be large, they are also very difficult to measure, so this monograph
ignores them.
If there are no externalities, then there is no need to measure social costs
and benefits when private people transfer their own funds to a DFI. It is
safe to assume that private people look out for their own good and thus
weigh benefits and costs as they see them. In contrast, there is a need to
measure social costs and benefits when governments and donors transfer
funds to DFIs. Public servants do not always look out for the public good
(Stiglitz 1998; Tollison 1984). Analysis is warranted because the group that
bears the costs is not the group that gets the benefits (Brent 1996).
Worse, the choice to subsidize DFIs may be in the hands of the very
same employees who stand to gain from the subsidies because more
funds for DFIs maintain their jobs, foster promotion, and expand their
influence. This group is small, organized, and vocal. Each member may
have a lot to gain from channeling more subsidies to DFIs.
Furthermore, the subsidies linked to loans from DFIs may attract rich
people. If loans are big so that subsidies are big, then rich people may
find it worthwhile to press for more public funds for DFIs. For exam-
ple, rich farmers and their lobbies sought and received big subsidies
from agricultural DFIs all over the world (Adams, Graham, and Von
Pischke 1984).
In contrast, the groups who bear the brunt of the costs of public DFIs
are taxpayers and nonclients left unhelped by projects left unfunded.
These groups are big, dispersed, and silent. The small cost to each mem-
ber of the group means that it is not worth their effort to press for cuts in
public support for DFIs.
Because changes in personal welfare caused by public support for
DFIs are not perfectly aligned with changes in social welfare, industry
lobbies and employees of governments, donors, and DFIs may be tempt-
ed to crusade for DFIs even if DFIs are not the best way to improve social
welfare. Alternatively, policymakers may lack the tools or the data to
check whether a DFI improves social welfare. The decision makers must
be watched because they may get benefits without bearing the costs. The



12                          DEVELOPMENT FINANCE INSTITUTIONS
measurement of costs can help to remind them of the worth of public
funds in DFIs versus in alternative uses.
How Do Public Development Finance Institutions
Differ from Other Public Projects?
Public DFIs resemble most other projects that get public funds. The peo-
ple who bear the costs are not the same people who get the benefits, and
there are small groups whose jobs and rents depend on more funds. DFIs
also hold an uncommonly tempting promise of development potential
because they work with financial capital, a factor often seen as a con-
straint on development. DFIs transfer control over assets, and the poor
are poor because they lack the assets that produce income (Sherraden
1991). DFIs are also politically correct. They do not give money away;
rather, they lend it at interest. Few dare oppose helping others help them-
selves.
DFIs differ from other public development projects in their unusual
susceptibility to abuse. The benefits of DFIs for clients are easy to see, but
costs are often obscure. No one can argue with the worth of a loan that
helps an orphan married at 12 and abandoned at 13 to buy land and to
send her child to school (RESULTS International 1996). In contrast, even
skilled financial analysts often overlook the opportunity costs of a public
DFI or the erosion of the real value of its equity. Measures of cost can help
decision makers to remember not only the faces of a few recipients but
also the faceless millions who do not get projects because funds go to
DFIs. The choice is not between a public DFI or nothing at all; rather, the
choice is between a public DFI or some other project to improve welfare.
DFIs are also susceptible to abuse because, on the surface, making
loans requires only money. Compared with other development projects,
DFIs are easy to start and to run because anyone with money-regardless
of technical expertise-can make loans (Ladman and Tinnermeier 1981).
DFIs may also attract-from a social point of view-too much donor
funds because they can absorb and disburse funds fast. It can be easier to
lend than to spend, especially if repayment is not a concern (Von Pischke
1991). Public support for DFIs may also offer politicians a convenient way
to hide transfers of wealth (Ladman and Tinnermeier 1981). To sum up,
the potential for the abuse of a DFI is high.
Why Does Society Subsidize Public
Development Finance Institutions?
Society subsidizes DFIs to improve social welfare (Yaron, Benjamin, and
Piprek 1997). The social benefit is the extra utility of clients with the DFI



WHY MEASURE THE SOCIAL COST OF PUBLIC DFIS?                   13
versus without it. The social cost is the benefit lost because the DFI was
funded instead of something else.
In principle, a market failure is required for public DFIs to improve
social welfare. A marketfailure is when competition fails to lead to a social-
ly efficient outcome (Besley 1994). This happens when a movement from
the status quo would improve social welfare, but no private entity can
capture enough of the gains to recoup its costs. The market fails because
the best private choice is not also the best social choice. In principle,
someone could be made better off and no one would be worse off.
In practice, market failures plague financial markets (Stiglitz 1993). But
market failure, though needed to justify public intervention, is not
enough. DFIs can be justified only if they mitigate a market failure so well
that the benefits due to the intervention exceed the costs due to the inter-
vention. Even in the absence of market failure, a public DFI might be the
best way to reach a social goal, for example, if no other tool addresses an
important social concern as efficiently (Yaron, Benjamin, and Piprek 1997).
DFIs venture where the market failed, and to find good borrowers
shunned by private lenders is a difficult task. Through adjustments to the
ways in which they judge and control risk, some DFIs have found prof-
itable ways to make loans and to get repaid without traditional collater-
al. Often, the most successful DFIs are those most concerned about the
measurement of their costs.
In the past, DFIs have often backfired, and they may even have hurt
those they meant to help (Yaron, Benjamin, and Piprek 1997; Hulme and
Mosley 1996; Krahnen and Schmidt 1994; Adams, Graham, and Von
Pischke 1984). "In practice, DFIs found it difficult to finance projects with
high economic but low financial rates of return and to remain financially
viable at the same time" (World Bank 1989, p. 106). Subsidies grew,
strained budgets, and failed to strengthen the DFIs so that they could sur-
vive without subsidies. For example, Mexico put more than $23 billion in
1992 dollars in agricultural DFIs from 1983 to 1992 before budget cuts
forced a decrease (World Bank 1994).
Of course, some DFIs are good and do mitigate market failures. But
some other DFIs may waste scarce funds or exacerbate market failures.
The theory is clear; if there is a market failure, then a DFI might have
scope to improve social welfare. In practice, however, market failure
alone is not enough to justify a DFI because DFIs themselves have costs
and can disrupt markets. Government failure may wreck attempts to fix
market failure, or a DFI might be inefficient. To choose well, society must
measure costs, and perhaps also benefits (Devarajan, Squire, and
Suthiwart-Narueput 1997).
A DFI is only one of a number of possibly complementary ways to
improve welfare through reduced poverty and increased incomes.



14                          DEVELOPMENT FINANCE INSTITUTIONS
According to Lipton and Ravaillon (1995, p. 2630), "Chronic poverty does
not appear to be due mainly to 'market failure' in credit or other markets,
but rather to low factor productivity and low endowments-per-person of
nonlabor factors."
How Can Society Measure the Benefits of
Public Development Finance Institutions?
The benefits of DFIs are the extra welfare of clients with a DFI versus
without. The comparison is not before-and-after but rather with-and-
without. A before-and-after comparison does not control for the changes
in welfare that would have happened regardless of access to the DFI
(Gittinger 1982). The problem is to know what would have happened
without the DFI, the standard counter-factual problem. This requires a
control group: people who cannot choose to use the DFI but who are just
like the people who can choose to use it in all ways. Getting a control
group usually requires random assignment of access to a DFI (or random
assignment of qualified applicants), but such social experiments consume
large amounts of funds, time, and expertise and are still subject to poten-
tially debilitating critiques (Heckman and Smith 1995). The only social
experiment with random assignment ever in development finance tests
the effects of access to Individual Development Accounts and is current-
ly taking place in Tulsa, Oklahoma (Schreiner 2000a; Sherraden and oth-
ers 2000).
Without a control group, there is no inexpensive way to measure the
impact of a DFI. For example, a dollar from a DFI is the same as a dollar
from any other source. This fungibility of money means that, without a
control group, the analyst cannot know if the loan caused an observed
outcome or if the outcome would have happened anyway (Adams 1988;
Adams and Von Pischke 1992; David and Meyer 1983; Von Pischke and
Adams 1980). Once the difficulties caused by fungibility became clear
and widely accepted, serious work to measure the benefits of DFIs went
dormant.
While fungibility does indeed wreck before-and-after comparisons, it
does not affect with-and-without comparisons between randomly
assigned treatment and control groups because random assignment con-
trols for all other factors that might affect outcomes. While good mea-
surements cost a lot and while wrong measurements could be worse than
no measurements, it does not follow that no one should try to make good
measurements.
In the absence of random assignment, the measurement of the benefits
of a DFI is similar to program evaluation with nonexperimental data.
Econometricians have grappled with this problem for at least 30 years



WHY MEASURE THE SOCIAL COST OF PUBLIC DFIs?                  15
(Moffitt 1991). Their research has concluded that the analyst must control
for the systematic differences-whether observed or unobserved-
between clients and nonclients. Rigorous attempts to do this for DFIs
include, among others, Montgomery, Johnson, and Faisal (2000); Amin,
Rai, and Topa (1999); Coleman (1999); McKnelly and Dunford (1998);
Morduch (1998); Mosley and Hulme (1998); Pitt and Khandker (1998);
Smith and Jain (1998); Carter and Olinto (1996); McKeman (1996); Sial
and Carter (1996); Lapar (1995); Bolnick and Nelson (1990); Feder and
others (1990); and Carter (1989).
Without random assignment, a good control group is hard to find.
Among the people who have access to a DFI, the people who choose to
use the DFI are not the same as the people who choose not to use the DFI;
the users are more likely to do well regardless of the DFI, perhaps
because they work more or accept higher risks. In contrast, the nonusers
likely would not do so well regardless of the DFI. Thus, simple compar-
isons of users to nonusers may overestimate the impact of the DFI.
Although the measurement of benefits is improving all the time, cred-
ible measures are still incomplete and require a long time, a lot of skill,
and a big budget. Thus, it is too expensive to measure the benefits of all
public DFIs (Yaron, Benjamin, and Piprek 1997). Also, while a study may
estimate the impact of a DFI on one outcome, it is much more difficult to
estimate the impact of a DFI on all outcomes of interest.
In contrast, it is less expensive to measure costs. In most cases, the
measurement of costs but not benefits is a better use of resources than the
measurement of both costs and benefits. This is the basic premise behind
cost-effectiveness analysis (CEA) (Garber and Phelps 1997; Weinstein and
Stason 1977). Whereas CBA compares costs with expensive-to-measure
benefits, CEA compares costs with inexpensive-to-measure outputs. Of
course, CEA is not as useful as full-blown CBA. For example, CEA cannot
rank projects that do not produce the same outputs for the same cus-
tomers. Also, CEA cannot tell whether benefits exceed costs. Examples of
CEA for development finance are Morduch (1999), Schreiner (1997),
Binswanger and Khandker (1995), and Gale (1991). CEA is the essence of
the measurement of the cost of public provision, suggested by Devarajan,
Squire, and Suthiwart-Narueput (1997) as one of two keys for better pro-
ject appraisal.
Of course, the fact that the measurement of costs is less expensive than
the measurement of benefits does not mean that only costs matter to the
exclusion of benefits. It just means that in discussions of the social worth
of DFIs, analysts will often be able to be more explicit about costs than
about benefits. Choices about what public projects to fund ultimately
must rely on informed judgments of both costs and benefits, even if
knowledge of costs is better than knowledge of benefits.



16                           DEVELOPMENT FINANCE INSTITUTIONS
What Is the Opportunity Cost to Society
of Public Funds Used by
Development Finance Institutions?
The social cost of a public DFI is the return its public funds could get in
their best other use. This return is called the opportunity cost, the efficiency
price, or the shadow price. A dollar used on one thing cannot be used on
something else.
The most important parameter in the measurement of the social cost of
public DFIs is the opportunity cost to society of public funds. The choice of
an appropriate opportunity cost will often drive the main results of the
analysis. This parameter is so expensive to measure that the analyst must
choose a proxy or simply make an assumption. While there are no foolproof
rules, this section provides some criteria and guidelines for the choice.
Are Social and Private Opportunity Costs the Same?
Social and private opportunity costs are not the same. For both society
and for private entities, the opportunity cost is the return that funds
could earn in a use of similar risk. But social and private opportunity
costs may diverge for two reasons. First, public and private investors do
not in general have the same opportunities, budgets, or constraints. For
example, public funds earmarked for development must be spent on
development projects. Private investors have no such constraint.
Second, private and social costs and benefits may diverge due to mar-
ket failures caused by externalities, public goods, transaction costs, prin-
cipal-agent problems, and/or information asymmetries. Private investors
count only their own costs and benefits. As a result, they may ignore
some projects with low private returns but high social returns. In con-
trast, public entities should count all costs and benefits to all people in
society. Of course, the existence of market failure does not necessarily jus-
tify public intervention (Besley 1994).
In general, the social opportunity cost is at least as high as the private
opportunity cost, because society accounts for benefits to all people while
private people only account for their own benefits. Furthermore, if a pri-
vate project has higher returns than a public one, then society can always
invest in it as if it were a private owner (ennings 1993).
Are Social or Private Opportunity Costs the Same as the Price
Paid by a Development Finance Institutionfor Public Funds?
Neither social nor private opportunity costs are necessarily equal to the
price paid by a DFI for public funds. The prices of public funds are set



WHY MEASURE THE SOCIAL COST OF PUBLIC DFIs?                 17
most often not by market feedback but by administrative fiat. For exam-
ple, one multi-lateral donor has made loans to DFIs with grace periods of
5 years, terms of 40 years, and interest rates of 1 percent. The price of this
public debt did not depend on its social opportunity cost, on the expect-
ed risk of the DFI, or on the rates and terms of a like loan in the market.
Because the price of public funds does not reflect opportunity costs, mea-
sures such as accounting profit and ROE do not reflect the performance
of a public DFI from a social nor private point of view.
Do Public Funds as Equity Have the Same
Opportunity Cost as Public Funds as Debt?
Public funds labeled as equity in a DFI do not necessarily have the same
opportunity cost as public funds labeled as debt because equity does not
have a fixed repayment obligation. In general, equity is riskier than debt,
so the opportunity cost of equity exceeds that of debt. The appendix
reviews a simple method developed by Benjamin (1994) to approximate
the private opportunity costs of debt and equity.
For society, debt and equity are the same for DFIs that are completely
state-owned or that have explicit or implicit state guarantees for all of
their debt. The examples here assume that the DFI is completely state-
owned, although most microfinance DFIs are not state-owned and do not
enjoy state guarantees on their debt.
If a DFI can take deposits, then it might replace public debt with
deposits instead of with private debt. In this case, the private opportuni-
ty cost of public debt would be the interest rate paid on deposits plus a
markup for the expected increases in the cost of administration and
reserve requirements. Still, loans from banks would often replace public
debt. In general, deposits will cost more and the DFI will be more likely
to use debt rather than deposits as the DFI is newer and smaller, has less
experience with deposits, is seen as risky by potential depositors, has
more debt compared with equity, has more competition, and has more
public funds to replace. Many DFIs-and certainly most microfinance
DFIs-cannot replace public funds with private deposits because they are
not licensed to take deposits.
What Are Proxiesfor the Social Opportunity Cost of Public Funds?
In practice, it is so expensive to measure the social opportunity cost of
public funds that less-expensive proxies are used. In some cases, govern-
ments or donors will have estimates of the return to some unfunded pro-
ject. But because project analysis itself is costly and is subject to dimin-
ishing returns, returns for all funded and unfunded projects will not be



18                               DEVELOPMENT FINANCE INSTITUTIONS
known. Furthermore, in practice, some projects are funded for reasons
other than their high net social benefits. If governments, donors, and their
employees are risk averse, then safe but low-return projects may get
funded before risky but high-return projects. With a budget constraint,
the best projects should be chosen until funds run out.
The goal of the choice of a social opportunity cost is to measure costs
well so as to help to spend public funds better. The choice has four cri-
teria. First, the number should be meaningful, that is, credibly close to
the true opportunity cost. Second, all public-sector analyses should use
the same opportunity cost because all public projects compete for pub-
lic funds and because comparisons across projects require the use of a
uniform opportunity cost. Third, higher rates are preferred to lower
rates, all else constant. This protects society from those who would use
low rates to give a false sense of rigor to support their pet projects.
Fourth, the rate chosen must be credible. Disagreements about the
social opportunity cost sidetrack project analyses more than any other
issue.
Figure 1.1 illustrates five possible proxies of the social opportunity
cost. The horizontal axis is the amount of funds Q. The vertical axis is the
nominal social opportunity cost m. As the amount Q increases, the mar-
Figure 1.1. Five Possible Proxies of the Social Opportunity Cost
Nominal social opp. cost, m
arginal return on public project
m
Private opp. cost
M|*
10 percent real = 0.1 + ic
Deposit interest rate
Inflation :               Ma   a cost of public fun s
M4                            Amount offunds,Q  Q
Amount of funds, Q



WHY MEASURE THE SOCIAL COST OF PUBLIC DFIS?                     19
ginal cost of public funds m increases, and the marginal return on pub-
lic projects decreases. Except for the signs of their slopes, the curves are
drawn arbitrarily and are assumed to include all factors that matter for
social costs and benefits, for example, the monopoly of the govemment
in the sale of riskless bonds due to its monopoly on the creation of legal
tender.
ZERO. Analyses that ignore opportunity costs implicitly assume a zero
nominal social opportunity cost. With positive inflation, this implies a
negative real social opportunity cost because real rates r are linked to
inflation xt and to nominal rates R through r - (R - rt)/(1 + it). Negative
opportunity costs are not credible because the net benefits of the mar-
ginal project are positive.
THE RATE OF INFLATION. A second possible proxy for the nominal social
opportunity cost is inflation 7t. Given positive inflation, then the real
social opportunity cost is zero. This is too low (Mishan 1988; Dasgupta
and Pearce 1978), in part because it implies that benefits in the present are
not preferred to benefits in the future. Some frameworks, however, do
use inflation as the nominal social opportunity cost (Rosenberg, Christen,
and Helms 1997; Holtmann and Mommartz 1996).
THE RATE OF INTEREST ON DEPOSITS. A third possible proxy is the interest
rate for treasury bills or, equivalently, the rate paid for time deposits by
state-owned DFIs plus a markup for the expected cost of administration
and reserve requirements, commonly assumed to be about two to three
percentage points but adjustable to the specific case (Yaron 1992b). Most
examples of the SDI use the deposit rate (e.g., Sacay, Randhawa, and
Agabin 1996; Khandker, Khalily, and Khan 1995; Yaron 1994). This
assumes that the social benefit of the marginal public project equals the
marginal cost of funds to the state.
If public funds were raised and spent to the point where marginal cost
equals marginal benefit (Q', m' in figure 1.1) and if all other markets were
perfectly competitive, frictionless, and without information or transac-
tion costs, then the deposit interest rate would equal both social and pri-
vate opportunity costs. In this equilibrium case, the social opportunity
cost is also a market rate, hence the symbol m.
In practice, all markets are not perfect. Instead, govemments and
donors have limited budgets and more potentially high-retum projects
than they can fund. In this disequilibrium, the cost of funds raised and
spent at Q is not the same as the return on the marginal project m. The
deposit rate is often less than the social opportunity cost. Thus, measures
of subsidy that use the deposit rate are lower bounds (Yaron 1992b).



20                             DEVELOPMENT FINANCE INSTITUTIONS
TEN PERCENT IN REAL TERMS. A fourth possible proxy for the social oppor-
tunity cost is 10 percent per year in real terms. This somewhat arbitrary
rate is used by most governments and by the World Bank as a uniform
rule of thumb (Belli 1996a; Katz and Welch 1993; Gittinger 1982). Like all
of the proxies described here, it may be adjusted for risk, although that is
not the best way to analyze risk (Norgaard and Howarth 1992;
Markandya and Pearce 1991).
If the real rate is r, then the nominal rate is r + 7i + r . i. Thus, the nom-
inal rate could be above or below the equilibrium rate m in figure 1.1.
Usually, funds will run out before projects with returns above 10 percent.
Although no one claims that 10 percent is particularly close to the true
real rate of return on the marginal public investment, the rate has been
favored in practice for at least three reasons. First, the true return on the
marginal public investment is unknown, and guesses as to its value
inevitably lead to endless debates. Second, compared with the known
rates already discussed, 10 percent is a higher lower bound on the true
marginal social return. Quirk and Terasawa (1991) find that the marginal
return to public investment is probably much higher than 10 percent, and
the estimates of Ballard, Shoven, and Whalley (1985) imply a minimum
social opportunity cost of 17 percent. Third, 10 percent is the number
most widely used. This not only defuses quibbles about its use, but it also
allows cost comparisons across projects. This view treats opportunity
costs less as marginal returns and more as tools to allocate scarce funds
from a budget (Belli 1996a).
Ten percent per year in real terms is a lower bound for the social
opportunity cost. According to Belli (1996a, p. 148), "A discount rate
lower than 10 percent might be difficult to justify." In particular, Gittinger
(1982, p. 315) says that "financial rates of interest, such as government
borrowing rates or the prime lending rate, are generally too low to justi-
fy their use in economic [from the point of view of society] analysis of
projects. Indeed, when inflation is high, these rates may even be negative
in real terms." The burden of proof for another opportunity cost rests on
the analyst (Gittinger 1982). The examples in this monograph use 10 per-
cent because it is the highest credible lower bound and because it helps
to make analyses comparable across projects and countries.
THE OPPORTUNrrY COST OF FuNDs TO PRIVATE ENTMES. A fifth possible
proxy for the social opportunity cost is the opportunity cost of private
entities. This is the risk-adjusted price to replace public funds with simi-
lar funds from private sources. For example, the private opportunity cost
of equity is the return required to attract and to retain private investors in
the long term. Likewise, the private opportunity cost of public debt is
what a DFI would pay for similar debt from private lenders. Of course,



WHY MEASURE THE SOCIAL COST OF PUBLIC DFIs?                 21
private opportunity costs vary through time and among DFIs due to dif-
ferences in risk, leverage, and the local cost of funds. The market price of
funds in figure 1.1 is drawn below the marginal return on public projects
because the state can always invest in private projects, should private
projects have a higher return than public projects (Conrad 1984, 1986;
Jennings 1993; Silvers and Kauer 1986).
Why Does the Measurement of Costs Boost Performance?
The measurement of costs sparks strong performance, casts light on bad
performance, and helps to reward good stewards in five ways (Schreiner
1997).
* Measurement forces DFIs and their sponsors to discuss their goals.
Foggy goals wither under attempts at measurement. Buzzwords lose
punch unless grounded in the nuts-and-bolts problems of measure-
ment (IADB 1994).
* Measurement changes goals. Those who measure costs worry about
costs and vice versa (Von Pischke 1996).
* Measurement highlights goals. A DFI that measures costs signals a
willingness to reduce costs. Success is more than only disbursement. If
donors measure only disbursements, then a DFI will learn to disburse
at any cost (Von Pischke 1998).
• Measurement helps to meet goals. Technical feedback helps managers
detect trends, set targets, benchmark progress, and compare to peers
(Richardson 1994; Koch 1992; Barltrop and McNaughton 1992).
* Measurement proves what is possible for DFIs. Governments and
donors want to demand better performance. But without measurement,
they are pestered by the fear that they ask for too much too fast. Unsure
donors expect less, and they get less (Schmidt and Zeitinger 1996).



2
What Is a Measure of the
Social Cost of a Public
Development Finance
Institution in the Short Term?
A measure of the social cost of a public DFI in the short term is the SDI
(Yaron 1992b). The SDI is the dollar value of subsidy divided by revenue
from interest and fees on loans. It answers the question: Howfar is the DFI
from being able to compensate society for the opportunity cost of its funds and
still show a profit? If a DFI could compensate for subsidy, then it is sub-
sidy-independent. Its SDI would be less than zero, and its Subsidy-
Adjusted ROE would exceed the social opportunity cost.
The SDI measures the cost of a public DFI and compares it with its
activity level. The SDI is a simple tool that shifts the paradigm from
reported (accounting) costs to opportunity (economic) costs. Accounting
profit and ROE often disguise the performance of public DFIs because
some expenses do not reflect social opportunity costs.
The SDI helps to measure progress toward "the phasing out of credit
subsidies, the assumption by the fiscal budget of funding responsibility
for any remaining subsidies, and the reduction and/or rationalization of
directed credit lines" as required by "World Bank Policies Guiding
Financial Sector Operations" (paragraph 17). The SDI can help link cur-
rent public support to progress toward future independence from public
support (Women's World Banking 1995).
How Does a Development Finance Institution
Get Subsidies?
A DFI gets subsidies from subsidized funds. Subsidizedfunds are public
funds. If a DFI accepts public funds, then it must be that they cost less
than private funds. Subsidy is the social opportunity cost minus the price
the DFI actually pays.
By definition, only public funds can be subsidized, and private funds,
regardless of their price, are not subsidized, unless a contribution is tax-
exempt or unless the market price is affected by an explicit or implicit state
22



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM               23
guarantee of the liabilities of a DFI. The distinction is particularly impor-
tant for microfinance DFIs that receive some private donations. Unlike
public donors, private donors spend their own money. The fact that the
owner of the funds agrees to entrust them to a DFI reveals that the benefits
of the transaction exceed the costs, at least from the point of view of the pri-
vate donor. For example, shares in a credit union held by members of their
own free will are not subsidized even if the credit union never pays divi-
dends or buys the shares back. The members choose to buy shares because
they judge that the benefits of membership are worth it. Likewise, com-
pensating balances that pay low rates are not subsidized. The lost earnings
are part of the price borrowers accept when they choose to borrow. Even
outright gifts are not subsidized, as long as they are private. The fact that
private funds are not subsidized does not necessarily mean that the DFI is
efficient, nor that it is privately profitable. It may still be useful to compute
private costs and/or benefits with common measures such as ROE or EVA
(box 1.2) that take the private point of view. But no public analysis is need-
ed. Still, the measurement of social cost for DFIs with some public funds
must be careful to exclude funds from private sources.
What Forms of Subsidized Funds
Does a Development Finance Institution Get?
Subsidized funds come in six forms (table 2.1). Three forms are equity
grants. Equity grants increase net worth directly but do not directly
change accounting profit reported in the year received. The other three
forms are profit grants. Profit grants increase accounting profit directly
because they inflate revenues and/or deflate expenses. This increases net
worth at year-end indirectly through retained earnings.
Compared with the case without the grant, all six forms increase net
worth one-for-one and have a social opportunity cost equal to m. As in
Yaron (1992b), this monograph ignores dividends and taxes on profits in
the interest of simplicity.
Table 2.1. Types of Subsidized Funds
Type of subsidizedfunds  Notation       Type of grant
Direct grant          DG                Equity grant (EG)
Paid-in capital       PC                Equity grant (EG)
Revenue grant         RG                Profit grant (PG)
Discount on public debt A . (m - c)     Profit grant (PG)
Discount on expenses   DX               Profit grant (PG)
True profit            TP               Equity grant (EG)
Source: Authors.



24                            DEVELOPMENT FINANCE INSTITUTIONS
Equity Grants
The first two forms of subsidized funds are equity grants EG. These cash
gifts increase net worth but do not change the accounting profit reported
in a period directly. Equity grants are the sum of direct grants DG and
paid-in capital PC:
equity grants = direct grants + paid-in capital  (2.1)
EG = DG + PC.
Direct grants DG are cash gifts. Direct grants increase net worth, but
they do not pass through the income statement, and so they do not inflate
accounting profit. Direct grants include both gifts in cash and gifts in
kind such as computers or trucks.
Paid-in capital PC comes from sales of shares to governments or
donors. Such a sale is like a direct grant because public funds pay for the
shares. Furthermore, most donors do not wield control like private own-
ers. This monograph assumes that all paid-in capital comes from public
sources.
Profit Grants
Profit grants are the third through fifth forms of subsidized funds (table
2.1). Like all equity grants, all forms of profit grants PG increase net
worth because they inflate accounting profit or reduce accounting loss
and thus increase retained earnings.
Profit grants are the sum of revenue grants RG, discounts on public
debt A. (m - c), and discounts on expenses DX:
profit grants = revenue grants + discount on public debt
+ discount on expenses                 (2.2)
PG = RG + A - (m - c) + DX.
Profit grants distort accounting profit P and thus ROE because they
depend not on business performance but on arbitrary choices by admin-
istrators and accountants. Donors can use profit grants to nudge account-
ing profit and ROE as high or as low as they like. In contrast to account-
ing profit and ROE, the SDI and the NPCS recognize that a dollar treated
as a profit grant has the same effect on business performance as a dollar
treated as an equity grant (box 2.1; figure 2.1).
Revenue grants RG are cash gifts. They are just like equity grants
except for the accounting choice to record them as revenue instead of as
direct injections to equity. Revenue grants increase net worth, but only



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                     25
Box 2.1. How Profit Grants Affect Profit and
Return on Equity
Governments and donors can give a DFI a dollar through equity grants or
profit grants. The choice affects nothing of substance because all grants
increase equity one-for-one and have the same opportunity cost. Unlike
equity grants, however, profit grants boost accounting profit and thus ROE.
Profit grants are equity injections, but they enter the accounts as if they
were operating revenue. Classifying a dollar as a profit grant instead of as
an equity grant increases accounting profit but does not change business
performance.
For example, suppose a donor injects $100 in a DFI at a smooth pace
through a year. The DFI starts with equity of $100. In the first case, the
donor gives all $100 as equity grants and none as profit grants. Equity
grants do not affect revenues or expenses, and the DFI posts an accounting
loss of $50. End equity is the sum of start equity, equity grants, and profit,
so average equity is (100 + 100 + 100 - 50)/2 = 125. ROE is -50/125 = -0.40.
In this first case, ROE correctly states that the DFI destroyed 40 cents for
each dollar of equity used (see figure 2.1).
Now suppose all $100 shifts from equity grants to profit grants.
Revenues increase and/or expenses decrease, so now profit is $50 even
though business performance is unchanged. Average equity is still 125, but
ROE is now 50/125 = 0.40. In this second case, ROE incorrectly states that
the DFI created 40 cents for each dollar of equity used (see figure 2.1).
Accounting profit and
ROE depend on the arbi-  Figure 2.1. Profit Grants and ROE
trary choice to record subsi-
dized  funds as equity            4'
grants or profit grants.
These measures may hide /
the true performance of a
DFI because they depend         ; 0
on the arbitrary form of
subsidized funds. Many
DFIs do not adhere to
GAAP and thus may record        -40
grants or reimbursements
of expenses not as equity           /o grants as profit grants
injections but as revenue.
The distinction matters for meaningful measures of financial performance.
Other common financial ratios have the same weaknesses. In contrast,
the SDI and the NPCS do not change as profit grants change. Christen
(1997) also proposes an elegant approach that adjusts the financial state-
ments themselves so that common ratios answer the questions they are
meant to answer for public DFIs.



26                           DEVELOPMENT FINANCE INSTITUTIONS
after they pass through the income statement and inflate accounting prof-
it. Because revenue grants are not the product of the business operations
of the DFI, they should not be in reported profit.
Discounts on public debt A - (m - c) and discounts on expenses DX are
the fourth and fifth forms of subsidized funds. They are noncash gifts,
expenses paid on behalf of the DFI by someone else. Discounts increase
the profit reported by the DFI because they decrease expenses.
The discount on public debt A. (m - c) is the opportunity cost of pub-
lic debt less what the DFI paid, where A is average public debt, c is the
rate the DFI paid for public debt, and m is the social opportunity cost of
public debt:
discount public debt = average public debt
- (opportunity cost public debt - rate paid) (2.3)
=A. (m  c).
Like all discounts, discounts on public debt are subsidized funds that
inflate profit and boost net worth because they cut expenses. Public debt
is like private debt linked to a grant of A - (m - c) (IADB 1994). Unlike the
discount on public debt, public debt itself does not increase net worth.
The average rate paid on public debt c is the expense for interest paid
on public debt, divided by the average public debt A:
c _ expense for interest for public debt  (2.4)
average public debt
The best way to estimate average public debt A is to track the dates
and the amounts of each inflow and outflow and then to find the average
daily balance. Usually, such detailed data are unavailable to external ana-
lysts. A common practice is to estimate A as half the sum of the public
debt at the start of the year Ao and at the end of the year A1, although
quarterly or monthly averages would be more accurate:
A = (Ao + A1)/2.                  (2.5)
Discounts on expenses DX are costs absorbed by governments or
donors that the DFI does not record as expenses. Classic examples are
technical help, free deposit insurance, coverage of organization costs or
feasibility studies, debt guarantees, fees for consultants, classes for loan
officers, and travel for employees. Although discounts on expenses often
leave no trace in the financial statements and are difficult to track, they
are common and may represent large resource transfers (Schreiner
2000b).



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM               27
True Profit
True profit TP, the sixth form of subsidized funds (table 2.1), is like an
equity grant. True profit is accounting profit P less profit grants (equa-
tion 2.2):
true profit = accounting profit - profit grants  (2.6)
TP=P-[RG+A. (m -c)+ DX].
All else constant, true profit is the change in retained earnings that
would obtain in the absence of profit grants. Positive true profits are a
social benefit because governments or donors could withdraw them from
the DFI for use in other development projects. By the same logic, negative
true profits (true losses) are social costs.
Although easily confused, the concept of true profit is distinct from the
concept of profit grants. Profit grants are cash gifts from government or
donors recorded as revenue. Profit grants inflate accounting profit. True
profit is accounting profit after the removal of profit grants and discounts
on expenses. True profit is what accounting profit would be in the
absence of distortions due to access to public funds.
What Is the Formula of the Subsidy Dependence Index?
Yaron (1992a) defines the SDI as subsidy S divided by revenue from loans
LP * i, where LP is the average loan portfolio and i is the yield on loans:
SDI        subsidy
revenue from lending                (2.7)
S
LP . i
The SDI is the percentage change in the yield on loans (or, equivalent-
ly, in revenue from loans) that, all else constant, would make subsidy zero.
For example, an SDI of 1.00 means that an increase in the yield of 100 per-
cent would wipe out subsidy and make the SDI equal zero. An SDI of zero
or less means the DFI could compensate society for its opportunity cost
and still show a profit. It also means that the Subsidy-Adjusted ROE
would exceed the social opportunity cost.
What Is the Denominator of the Subsidy Dependence Index?
The denominator of the SDI is revenue from loans. This is the product of
the average loan portfolio outstanding LP and the yield on loans i:



28                           DEVELOPMENT FINANCE INSTITUTIONS
revenue from loans = average loan portfolio * yield on loans    (2.8)
= LP  i.
The yield i is interest and fee revenue from loans, divided by the aver-
age loan portfolio:
interest and fees from loans          (2.9)
average loan portfolio
What Is the Numerator of the Subsidy Dependence Index?
Yaron (1992a) defines the numerator of the SDI as subsidy S:
S=m  E+A  (m-c)+K-P                   (2.10)
where S is subsidy, m is the social opportunity cost, E is average equity, A
is average public debt, c is the rate paid for public debt, K is revenue grants
and discounts on expenses, and P is accounting profit. Subsidy is the sum
of the opportunity cost of the funds lodged in the net worth of a DFI and
of the three types of profit grants, less the accounting profits the DFI could
use to compensate for opportunity costs while still showing a profit. This
monograph assumes that all net worth in average equity E comes from
public sources. K is "the sum of all other annual subsidies received by the
DFI (such as partial or complete coverage of the DFI's operational costs by
the state) . . . [and] all other miscellaneous subsidies that a DFI might
receive. These include subsidization of training costs, free use of govern-
ment facilities and vehicles, free computer facilities, full or partial exemp-
tion from the deposit reserve requirement, and full or partial guarantee by
the state of loan repayment by subborrowers in default" (Yaron 1992b, pp.
6, 12). In other words, K is revenue grants plus discounts on expenses:
K=RG+DX.                        (2.11)
Clarity about K matters because if K does not include revenue grants,
then the SDI will depend on the arbitrary form of subsidized funds.
Worse, the SDI would underestimate subsidy. Two recent attempts to
adjust the SDI (chapter 5) botch K.
Given K (equation 2.11), the level of subsidy S (equation 2.10) is:
S = m  E +A  (m-c) +RG + DX-P.            (2.12)
Given year-end financial statements and assuming that stocks grow
and flows occur at a constant pace through the year, average stocks are



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM              29
half the sum of the start and end stocks. The end stock of equity is the
start stock plus the change in the stock:
E = (Eo + Ej)/2
= (Eo + Eo + AE)/2                  (2.13)
= Eo + (1/2) AE.
The change in equity AE is the sum of flows of the six forms of subsi-
dized funds:
AE = equity grants + profit grants       (2.14)
=DG+PC+RG+A -(m-c)+DX+ TP.
We rewrite the formula for subsidy (equation 2.10) with the formula
for true profits (equation 2.6), K (equation 2.11), average equity E (equa-
tion 2.13), and the change in equity AE (equation 2.14):
S =m E + A- (m - c) + K - P
=m  [Eo + (1/2) - (DG + PC + RG + A * (m - c) + DX + TP)]  (2.15)
+ RG + A - (m - c) + DX - [TP + RG + A- (m - c) + DX]
=m-Eo + (m/2) - [DG + PC + RG + A - (m - c) + DX + TP] - TP.
This breaks the SDI into three terms. The first term, m EO, is the oppor-
tunity cost of the subsidized funds that the DFI used through the whole
year. The second term, (m/2).  [DG + PC + RG + A . (m - c) + DX + TP], is
the opportunity cost of the fresh subsidized funds that the DFI got in the
course of the year. On average-or in the absence of knowledge of the
times when flows of public funds were received-the DFI had the use of
half the change in the stock of these new funds. The third term, TP, is the
true profit that the DFI would record without subsidies, what the DFI
could use to compensate society if it had no profit grants. Subsidy S is
then equal to the unpaid social opportunity cost less profit generated
from operations.
The formula also shows that the SDI does not depend on the form of
subsidized funds. Total subsidized funds from past years (in EO) have a
marginal and average cost of m. Total fresh subsidized funds have an
average cost of (m/2).
Why Does the Subsidy Dependence Index Compare Subsidy
to Revenuefrom Loans?
The key of the SDI is the measurement of subsidy. The comparison of sub-
sidy to revenue from loans is important, but secondary. Many things affect



30                            DEVELOPMENT FINANCE INSTITUTIONS
subsidy, and subsidy could be compared with any item from the financial
statements. Yaron (1994) focuses not only on revenue from loans but also
on loan recuperation, deposit mobilization, and administrative costs. The
choice to focus first on revenue from loans makes sense for four reasons.
First, DFIs often set interest rates by decree or have rates set for them
by governments or donors. Within a range, the DFI can often change
them with a stroke of a pen. In theory, interest rate hikes can dampen
demand and prompt loan losses (Morduch 2000; Stiglitz and Weiss 1981).
In practice, few DFIs have doused demand or spawned a rash of default
with higher interest rates because demand outstrips supply (Rosenberg
1996) (box 2.2; figures 2.2, 2.3). An efficient DFI does not gouge when it
charges enough to cover its costs in the long term.
lqbal (1986) found that interest rates mattered much less to small farm-
ers than to big farmers. According to Singh, Squire, and Strauss (1986,
p. 175):
It follows that the elimination or reduction of subsidies to programs
providing agricultural credit may serve the dual purpose of increas-
ing efficiency in the capital market and simultaneously improving
equity, since the reduction in borrowing by "large" farmers will
exceed that by "small" ones.
Second, if public funds will be cut in the long term, then the chances
of survival increase as a DFI can cover more of the cost of private funds
with revenues from loans.
Third, revenue from loans is the biggest item in the income statement
and usually exceeds all other operational sources of income combined.
Most DFIs cannot reduce expenses or increase nonloan income enough to
compensate for subsidies, so higher prices for loans may be the best
option, at least in the short term.
Fourth, the comparison of subsidy with revenue from loans places
subsidy in the context of the size of the DFI. Measures of effective pro-
tection or of domestic resource cost do the same thing (Tweeten 1992;
Gittinger 1982). The comparison also allows the SDI to be seen as the
matching grant society provides the DFI (subsidy in the numerator of the
SDI) for each dollar of revenue from interest and fees earned from loans
to clients in the target group (the denominator of the SDI).
Could Subsidy Be Compared to Anything Else?
Subsidy can be compared with anything in units of dollars per unit of
time. Good candidates are average equity or average assets. Such com-



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                           31
Box 2.2. Real Yields at Grameen and BancoSol
Along with the unit-desa system of Bank Rakyat Indonesia, the most famous
DFIs in the world are the Grameen Bank of Bangladesh (Hashemi 1997;
Khandker 1996) and BancoSol of Bolivia (Gonzalez-Vega and others 1997;
Mosley 1996). The real yield on loans at Grameen and BancoSol varied
widely, but the loan portfolios grew consistently and default stayed low.
Thus, there may be room to increase interest rates and/or fees in the pur-
suit of subsidy independence.
From 1984 to 1994, Grameen earned a nominal yield that varied from 12
to 19 percent (figure 2.2). Inflation varied from 1 to 22 percent, and the real
yield varied from -1 to 14 percent. More than 99 percent of taka disbursed
were recovered (Schreiner 1999b), and the loan portfolio grew from $9 mil-
lion to $275 million. The portfolio grew and default was low despite big
changes in the real yield.
From 1987 to 1996, the nominal yield at BancoSol varied from 36 to 63
percent (figure 2.3). Inflation varied from 8 to 23 percent, and the real yield
varied from 11 to 49 percent. More than 99 percent of dollars disbursed
were recovered, and the loan portfolio grew from 0 to $47 million (Schreiner
1997). Again, huge swings in the real yield went side-by-side with huge
portfolio growth and low default.
Figure 2.2. Grameen:                    Figure 2.3. BancoSol:
Inflation and Nominal                   Inflation and Nominal
and Real Yields                         and Real Yields
0.22 -0.7-
0.122Nominal yield                       0.6  Nominal yield
0.5
0.14
0.4-
0.10 - lation                                        
0.06           Real ~~~~yield      0.3 Ra il 
0.02                                     0.1 0.1
Inflation
-0.02 84 8   8   '9 99 92 9394             0 87 88 89 90 91 92 93 94 95 96
Year                                    Year
Sources: Schreiner (1997); Yaron, Benjamin, and Piprek (1997); Khandker,
Khalily, and Khan (1995); Hashemi (1997); and IMF (various years).



32                           DEVELOPMENT FINANCE INSTITUTIONS
parisons result in a Subsidy-Adjusted Return on Equity (SAROE) or a
Subsidy-Adjusted Return on Assets (SAROA). If subsidy is less than zero,
then the SAROE will exceed its hurdle rate, the social opportunity cost.
A DFI might respond to the loss of subsidy or might decrease its sub-
sidy dependence in many ways (Yaron 1992b). For example, it could slash
administrative costs, dun borrowers more, grow the loan portfolio, or
boost productivity.
All these strategies require care and analysis. Also, the DFI is a price-
taker in its investment portfolio, and so it cannot increase those revenues
much without an increase in risk. Comparisons with the average loan
portfolio LP require care because the DFI cannot increase LP with the
same ease as interest rates. Rapid growth in loan volume seems more
likely to provoke a rash of default than would a rapid price increase
because rapid growth in the short term can only be achieved by accept-
ing lower-quality clients.
Does the Subsidy Dependence Index Prescribe
Interest Rate Hikes?
The SDI does not prescribe interest rate hikes; rather, the SDI describes
social cost and how much yields would have to increase-all else con-
stant-to eliminate subsidy.
The SDI does not condemn subsidies nor public DFIs; after all, they
may be the best way to improve social welfare. But society should pursue
knowledge of the cost of DFIs to check whether they are good stewards
of public funds. Governments and donors should not buy DFIs sight
unseen nor measure their performance with inappropriate tools. Of
course, they also need some estimate of benefits to check the social worth
of DFIs.
Numerical Examples of the Subsidy Dependence Index
The following describes the financial results of an example DFI and then
walks through the calculation of the SDI. The DFI was born on January 1
of Year 01, and the averages for Year 01 use the figures (all zero) for the
end of Year 00.
Description of Financial Results for Year 01
In the balance sheet of the first year of the example DFI (table 2.2), most
assets (more than two-thirds) are loans (lines Ad and Ag), and invest-
ments and fixed assets are modest. Cash is 20 percent of all assets. Half of
all liabilities are public debt, and half are deposits and private debt (lines



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                 33
Ah, Ai, and Aj). While governments or donors own some shares (line Al),
most net worth comes from direct grants (line Am). The example DFI is
highly subsidized.
The first-year income statement (table 2.3) shows that the DFI paid 25
in interest for its liabilities (line Bg), spent 600 in operating costs (line Bj),
and did not provide for loan losses (line Bi). Revenues from loans and
investments were 420 + 5 = 425 (lines Ba, Bb, and Bc). Operating revenue
less operating costs and financial costs produced an operating margin of
425 - (25 + 600) = -200 (line Bk). This would have been even more nega-
tive in the absence of the discount on expenses of 100 (line Bn). As it was,
this and a revenue grant of 400 (line BI) let the example DFI boast an
accounting profit of 200 (line Bm).
If gifts from discounts on expenses and revenue grants of 100 + 400 =
500 were called equity grants rather than profit grants, then accounting
profits would be negative. Thus, measures that use accounting profit can
obscure the true performance of a public DFI. The accounting treatment
of a gift should not change measures of business performance.
Rates of interest are ratios of revenues and expenses from the income
statement to average stocks from the balance sheet. The yield on loans i
for the example DFI in Year 01 is (420)/[(0 + 2,100)/2] = 0.40 (line Cv of
table 2.4). This result uses the formula for the yield on loans (equation
2.9), the revenue from loans (line Ba in table 2.3), and the start and end
stocks of the net loan portfolio (line Ad in table 2.2).
The interest rate on public debt is (10)/[(0 + 400)/2] = 0.05. This uses the
formula for c (equation 2.4), the interest expense on public debt (line Bf in
table 2.3), and the start and end stocks of public debt (line Aj of table 2.2).
With an assumed opportunity cost to society of public debt m of 10
percent per year in real terms (line Ck of table 2.4), the DFI would pay
(0.10) - [(0 + 400)/2] = 20 for equivalent private debt. The discount on
public debt is the social opportunity cost less what was actually paid, 20 -
10 = 10 (line Cl).
The example DFI paid an interest rate on deposits of (5)/[(0 + 200)/2]
= 0.05 (line He of table A.1). The interest rate paid on private debt was
(10)/[(0 + 200)/2] = 0.10. The DFI also earned a yield on investments j of
(5)/[(0 + 200)/2] = 0.05.
What Is the Subsidy Dependence Index
of the Example Development Finance Institutionfor Year 01?
The SDI of the example DFI for Year 01 was 100 percent (line Cx of
table 2.4). Table 2.4 uses the received formula for subsidy (equation 2.10),
but the alternative formula (equation 2.15) gives the same result (line Dq
of table 2.5). An SDI of 100 percent means that, all else constant, an



Table 2.2. Balance Sheet
Line                                                                 12/31/01     12/31/02     12/31/03
Assets
Aa  Cash                             Data                             600            700         800
Ab  Loan portfolio (gross)           Data                            2,100         3,300       5,200
Ac  Reserve for loan losses          Data                               0              0           0
Ad  Loan portfolio (net), LP         Ab + Ac                         2,100         3,300       5,200
Ae  Investments, I                   Data                             200            400         600
Af  Fixed assets (net)               Data                             100            200        200
Ag  Total assets                     Aa + Ad +Ae + Af                3,000         4,600       6,800
Liabilities
Ah  Deposit liabilities              Data                             200            400         600
Ai  Private debt                     Data                             200            300        400
Aj  Public debt, A                   Data                             400            800       1,200
Ak  Total liabilities                Ah + Ai + Aj                     800          1,500       2,200
Equity
Al  Paid-in capital, PC              Data                             300            645         910
Am  Direct grants, DG                Data                            1,700         2,000       2,300
An  Retained earnings                Ant - 1 + Bm                     200            455       1,390
Ao  Total equity                     Al + Am + An                    2,200         3,100       4,600
Ap  Total equity and liabilities     Ak +Ao                          3,000         4,600       6,800
Note: Monetary figures in constant units.
Source: Example of authors.



Table 2.3. Income Statement
Line                                                                   12/31/01     12/31/02     12/31/03
Ba  Revenue from loans, LP . i        Data                              420          1,080        1,700
Bb  Revenue from investments, I j    Data                                 5             15          25
Bc  Total revenue operations          Ba + Bb                           425          1,095        1,725
Bd  Exp. int. deposit liabilities     Data                                5             15          25
Be  Exp. int. private debt            Data                               10             25          35
Bf  Interest expense on public debt, A  cData                            10             30          50
Bg  Total int. exp.                   Bd + Be + Bf                       25             70         110
Bh  Financial margin                  Bc - Bg                           400          1,025        1,615
Bi  Exp. prov. reserve for loan losses   Data                             0              0           0
Bj  Exp. admin.                       Data                              600          1,170        1,080
Bk  Operating margin                  Bh - (Bi + Bj)                   (200)          (145)        535
Bl  Rev. grants, RG                   Data                              400            400         400
Bm  Accounting profit, P              Bk + Bl                           200            255         935
Memo item:
Bn  Discounts on expenses, DX         Data                              100            100         100
Note: Monetary figures in constant units.
Source: Example of authors.



Table 2.4. Calculation of the Subsidy Dependence Index
Line                                                              12/31/01     12/31/02    12/31/03
Ca  Start equity                       Alt -1 + Am, -l + An           0         2,200       3,100
Cb  End equity                         Al + Am + An                2,200        3,100       4,600
Cc  Average equity, E                  (Ca + Cb)/2                 1,100        2,650       3,850
Cd  Opportunity cost of society, m     Data                        0.10          0.10        0.10
Ce  Subsidy on equity, E - m           Cc Cd                        110           265        385
Cf  Start public debt                  Ajt 1                          0           400        800
Cg  End public debt                    Aj                           400           800       1,200
Ch  Average public debt, A             (Cf + Cg)/2                  200           600       1,000
Ci  Exp. int. public debt, A c         Bf                            10            30         50
Cj  Rate paid for public debt, c       Ci/Ch                        0.05         0.05        0.05
Ck  Opportunity cost public debt, m    Data                        0.10          0.10        0.10
Cl  Discount on public debt, A. (m - c)  Ch (Ck - Cj)                10            30         50
Cm  Revenue grants, RG                 Bl                           400           400        400
Cn  Discounts on expenses, DX          Bn                           100           100        100
Co  Revenue grants and discounts
on expenses, K                   Cm + Cn                      500          500         500



Cp  Accounting profit, P                 Bm                           200            255         935
Cq  Subsidy, S                           Ce + Cl + Co-Cp              420            540           0
Cr  Start loan portfolio (net)           Adt_1                          0          2,100       3,300
Cs  End loan portfolio (net)             Ad                          2,100         3,300       5,200
Ct  Average loan portfolio (net), LP     (Cr + Cs)/2                 1,050         2,700       4,250
Cu  Revenue from loans, LP. i            Ba                           420          1,080       1,700
Cv  Yield on lending, i                  Cu/Ct                        0.40          0.40        0.40
Cw  Revenue from lending, LP. i          Ct Cv                        420          1,080       1,700
Cx  Subsidy Dependence Index, S/(LP i)   Cq/Cw                        1.00          0.50        0.00
Cy  Yield on lending, i                  Cv                           0.40          0.40        0.40
Cz  Change in yield                      Cy . Cx                      0.40          0.20        0.00
Caa Subsidy-free yield                   Cy + Cz                      0.80          0.60        0.40
Note: Monetary figures in constant units. Average equity includes profit.
Source: Example of authors.



Table 2.5. Alternative Calculation of the Subsidy Dependence Index
Line                                                                  12/31/01     12/31/02     12/31/03
Da  Opportunity cost of society, m       Cd                            0.10          0.10        0.10
Db  Start equity, Eo                     Ca                              0          2,200       3,100
Dc  EO m                                Da Db                            0            220         310
Dd  End direct grants                    Am                           1,700         2,000       2,300
De  Start direct grants                 Am, -                            0          1,700       2,000
Df  Change direct grants, DG            Dd - De                       1,700           300         300
Dg  End paid-in capital                  Al                            300            645         910
Dh  Start paid-in capital                Alt-1                           0            300         645
Di  Change paid-in capital, PC           Dg - Dh                       300            345         265
Dj  Discount public debt, A. (m -c)     Cl                              10             30          50
Dk  Rev. grants, RG                      Bl                            400            400         400
Dl  Discounts on expenses, DX            Bn                            100            100         100
Dm  Accounting profit, P                 Bm                            200            255         935
Dn  True profit, TP                     Dm - (Dj + Dk + Dl)           (310)         (275)         385
Do  Subsidy, S                           Dc + (Da/2) - (Df + Di + Dj
+ Dk-Dl + Dn) -Dn           420           540           0
Dp  Rev. from loans, LP * i              Ba                            420          1,080       1,700
Dq  Subsidy Dependence Index, S/(LP i)  Do/Dp                          1.00          0.50        0.00
Note: Monetary figures in constant units. Average equity includes profit.
Source: Example of authors.



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                 39
increase of 100 percent in the yield on loans would allow the DFI to show
a profit and still compensate for the social opportunity cost of its funds.
The subsidy on equity is 1,100 0.10 = 110 (line Ce of table 2.4). This is
the product of an average equity E of [(0 + 0 + 0) + (300 + 1,700 + 200)]/2
= 1,100 (line Cc) and of a social opportunity cost of equity m of 10 percent
(line Cd).
The discount on public debt (line Cl) is [(0 + 400)/2] * (0.10 - 0.05) = 10.
This is the product of average public debt A (line Ch) and of the opportu-
nity cost to society of public debt m (line Ck) less the rate paid c (line Cj).
The amount K (equation 2.11) is 400 + 100 = 500 (line Co). This is the
sum of revenue grants RG (line Cm) and discounts on expenses DX (line
Cn). Accounting profit P is 200 (line Cp). Finally, revenue from loans LP -
i is [(0 + 2,100)/2] - 0.40 = 420 (line Cw). This is the product of the aver-
age loan portfolio LP (line Ct) and the yield on loans i (line Cv). Thus, the
SDI for Year 01 is (equation 2.7):
S
SDI01 = LPSt
m . E + A - (m - c) + K-P
LP . i
0.10 * 1,100 + 200. (0.10 - 0.05) + 500 - 200  (2.16)
1,050 - 0.40
= (110 + 10 + 500 - 200)/420
= 420/420 = 1.00.
What Does the Subsidy Dependence Indexfor Year 01 Mean?
All else constant, the SDI for Year 01 of 100 percent means the DFI could
compensate for the social opportunity cost of its funds and still show a
profit if revenue from loans increased by 100 percent. If the size of the
loan portfolio does not change, then this would mean doubling the yield.
In general,
Subsidy-free yield = actual yield- (1 + SDI)             (2.17)
= actual yield + implied change in yield.
The SDI is a relative measure; it measures the implied change in the
yield that would compensate for subsidies, relative to the actual yield.
The actual yield varies from year to year and from DFI to DFI. Also, the
nominal yield varies with inflation even if the real yield does not. Thus,
good analysis will consider, in both real and nominal terms, the absolute
level of subsidy S (a dollar amount), the ratio of subsidy to revenue from
loans (the SDI, a percentage), the actual yield i (a percentage), the implied



40                            DEVELOPMENT FINANCE INSTITUTIONS
change in the yield (a percentage), and the subsidy-free yield (a percent-
age). In this example, the actual yield is 40 percent (line Cy in table 2.4).
The subsidy-free yield is 0.40 + 0.40 * 1.00 = 0.80 (line Caa). The implied
change is 0.80 - 0.40 = 0.40 (line Cz). Because inflation is assumed to be
zero, the real yield equals the nominal yield.
Does the Subsidy Dependence Index De end
on How Average Equity Is DefinedT
Like ROE, the SDI depends on how average equity is defined.
Accountants do not agree on the best way to define average equity. All
agree that average equity should include start equity Eo and fresh injec-
tions to equity such as equity grants and paid-in capital in the year,
weighted for the time of injection. Average equity should also include
revenue grants RG, the discount on public debt A * (m - c), and the dis-
count on expenses DX because these profit grants are equity grants in
disguise. Opinion differs, however, whether average equity should also
include true profit in the current year. When average equity excludes
true profit in the current year, then the measure of subsidy in the SDI
more closely resembles a present-value measure. For this reason, it is
preferable to exclude true profit in the current year from measures of
average equity.
As in Yaron (1992a, 1992b), the formulae in this monograph include
true profit TP in average equity (e.g., equation 2.15). This recognizes that
owners could, in principle, withdraw true profit as it accrues. If they
choose not to, then it is as if they withdrew true profit but injected it back
as paid-in capital. If average equity includes true profit, then the measure
does not depend on this arbitrary choice made by owners. Also, the use
of average equity can be seen as a practical solution to imperfect data
without knowledge of the timing of injections to equity, accrual of profit,
and dividend pay-outs. The use of start equity would be more consistent
with economic paradigms based on net present cost, but this approach
would require more accurate data than are normally available to external
analysts. The use of average equity rather than start equity is a compro-
mise necessitated by the available data.
Unfortunately, this definition of average equity is inconsistent with the
way some other rates are measured. For example, suppose that a DFI has
a constant balance through a year of 100 of debt and 100 of paid-in capi-
tal. Suppose that in a year the DFI pays 10 in interest for the debt and that
it accrues a true profit of 10. The common way to measure the rate of
interest on the debt is as 10/ [(100 + 100)/2] = 0.10. At the same time, the
rate of Return on Equity with true profit included in average equity is
10/[(100 + 100 + 10)/21 = 0.095. These two measures are inconsistent.



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                 41
This can be reconciled in two ways. The first recognizes that the inter-
est on the debt was paid throughout the year and not just at the end, so
the DFI had to finance the interest with 10 of debt. Then the measure of
the interest rate on debt is 10/ [(100 + 100 + 10)/2] = 0.095, the same as the
common way to measure rates of Retum on Equity.
The second way to reconcile the measures is to assume that owners
measure their returns not on average funds used but rather on start funds
invested. This is similar to the practice of measuring the effective annual
interest rate on a loan as if its balance did not change in the course of a
year. In this case, the measure of the rate of Return on Equity is 10/[(100
+ 100)/2] = 0.10, the same as the common way to measure rates of inter-
est on debt.
The method used to measure average equity affects the SDI (Schreiner
1997; Yaron 1992b). The choice of method matters more as the absolute
value of true profit grows relative to start equity (box 2.3). For the exam-
ple DFI in Year 01, suppose that average equity does not include true
profit. Thus, average equity does not change when the yield on loans
doubles and when true profit increases by 420. The new SDI is:
SDI'01 = 0.10 . 1,100 + 200 . (0.10 - 0.05) + 500 - (200 + 420)
SDI'oi =    1,050 . [0.40 . (1 + 1.00)]
110 + 10 + 500 - 620                            (2.18)
840
= 0/840 = 0.
If the change in true profit does affect average equity, then the new SDI is:
SDI"01 - 0.10. (1,100 + 200) + 200 (0.10 - 0.05) + 500 - (200 + 420)
SDI"oi =      1,050  [0.40 (1 + 1.00)]
130 + 10 + 500 - 620                             (2.19)
840
= 20/840- 0.02.
What Are the Subsidy Dependence Indexes for
Year 02 and Year 03?
The example DFI cut its SDI in Years 02 and 03. It did not increase its
yield on loans in Year 02; rather, it increased the average loan portfolio
by 157 percent (line Ct of table 2.4) while administrative costs
increased only 95 percent (line Bj of table 2.3). In short, the DFI got
more efficient. Perhaps costs were high in the first year because the DFI
was born with all it would ever need. It bought office space, hired a full
complement of administrators, set up a computer system, and hired



42                              DEVELOPMENT FINANCE INSTITUTIONS
Box 2.3. The Subsidy Dependence Index and
Average Equity at Bank Rakyat Indonesia
The case of the unit-desa system of Bank Rakyat Indonesia illustrates the
sensitivity of the SDI and ROE to the inclusion of current-year profit in the
measure of average equity E (Charitonenko, Patten, and Yaron 1998).
The stock of equity in the unit-desa system at the start of 1995 Eo was
about 72 billion rupiah ($1 was worth about 2,200 rupiah). Accounting prof-
its P in 1995 were about 393 billion rupiah. Thus, ROE computed in terms
of start equity Eo was 393/72 -5.45, or about 545 percent. With profits in the
year included in the measure of average equity, ROE was 393/[(72 + 393)
/2] - 1.69, or about 169 percent. Thus, the inclusion of current-year profits
in equity had a large effect on the estimate of ROE.
To compute the SDI, note that the value of K-including discounts due to
exemption from reserve requirements-was -43 billion. The unit-desa system
used no subsidized debt, so the discount on public debt A * (m - c) was zero.
The social opportunity cost m used in Charitonenko, Patten, and Yaron (1998)
was 17.9 percent, and revenue from loans LP i was 861 billion rupiah.
With the inclusion of current-year profits in average equity E, the SDI
was:
l[(E0 + El)/2] m + A * (m - c) + K - P)/LP i
= ([(72+72+393)/2] - 0.179 + 0 + (-43)-393]/861
- -388/861
_-0.45.
That is, the unit-desa system could have reduced the yield on loans by 45
percent (from about 32 percent to about 17 percent), compensated for the
social opportunity cost of its public funds, and still shown a profit. This
extremely high level of subsidy independence is one reason why
Charitonenko, Patten, and Yaron (1997, p. 5) conclude that "no other suc-
cessful, sustainable micro or rural finance institution shows the extent of
outreach or the degree of financial self-sustainability that the BRI Unit Desa
system has achieved."
What if the measurement of average equity E excludes current profits?
The SDI is then:
([(72 + 72)/2] * 0.179 + 0 + (-43) - 3931/861 - -423/861
- -0.49.
Subsidy independence increases, although the change is only about four
percentage points. Given the extremely high leverage of the unit-desa sys-
tem (equity in 1995 was only 1.4 percent of assets) and its extremely high
profitability (ROA in 1995 was 6.1 percent), the SDI does not seem very sen-
sitive to the inclusion or exclusion of current-year profits in the measure-
ment of average equity.



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM               43
several loan officers. It took time for the loan officers to get up to speed
with a full portfolio. In the meantime, costs per unit of output were
high because the costs of the infrastructure were spread over a small
portfolio.
Discounts on expenses DX did not change (line Bn of table 2.3), and
discounts on public debt tripled (line Cl of table 2.4). Average equity
increased by 140 percent (line Cc of table 2.4). The SDI was cut in half:
SDI02 = m  E + A. (mr-c) + K-P
SDI02    LP . 
0.10 * 2,650 + 600 * (0.10 - 0.05) + 500 - 255  (2.20)
2,700 0.40
= (265 + 30 + 500 - 255)/1,080
= 540/1,080 = 0.50.
Subsidy S rose from 420 to 540, but the SDI fell (lines Cq and Cx of
table 2.4). The subsidy-free yield was 0.40 * (1 + 0.50) = 0.60 (line Caa), and
the implied change was 0.20 (line Cz).
In the third year, the average portfolio grew by more than 50 percent
(line Ct of table 2.4). Administrative expenses fell by about 8 percent (line
Bj of table 2.3). Discounts on expenses DX did not change. Discounts on
public debt grew by 20 (line Cl of table 2.4). The example DFI still got
fresh flows of all six forms of subsidized funds, but increased profits
drove the SDI to zero:
SDIo3 = m . E + A. (m - c) + K - P
SD03       LP . i
0.10 * 3,850 + 1,000 - (0.10 - 0.05) + 500 - 935  (2.21)
4,250 - 0.40
(385 + 50 + 500 - 935)/1,700
0/1,700 = 0.00.
Of course, similar examples could be presented in which the measure
of equity E does not include profit in the current period.
Is Subsidy in the Subsidy Dependence Index Related
to a Subsidy-Adjusted Return on Equity?
The measure of subsidy in the SDI is closely related to a Subsidy-
Adjusted ROE. This is useful because ROE is the most common measure
of the financial performance of a private firm. Most users of financial
information know and understand ROE. ROE compares accounting prof-
it (after tax) with average equity:



44                          DEVELOPMENT FINANCE INSTITUTIONS
ROE = accounting profit (after tax)    (2.22)
average equity
ROA resembles ROE except it compares accounting profit with aver-
age assets:
ROA = accounting profit (after tax)    (2.23)
average assets
ROA is a useful tool for comparisons between peers in the same
macroeconomic environments because it removes the effects of financial
leverage. Barltrop and McNaughton (1992) and Mould (1987) explain the
use of ROE, ROA, and other common financial ratios in the analysis of
DFIs.
ROE and ROA use accounting profits, and accounting profits depend
on whether a gift is called an equity grant or a profit grant. A Subsidy-
Adjusted ROE (SAROE) (or a Subsidy-Adjusted ROA [SAROA]) would
replace accounting profit with true profit. An SAROE compares true prof-
its with average equity:
SAROE = true profit                   (2.24)
average equity
Likewise, an SAROA compares true profit with average assets. The
SAROE and the SAROA are useful to compare public DFIs with peers
(Christen 1997). Peer comparisons are the standard way to benchmark
the performance of banks (Barltrop and McNaughton 1992; Koch
1992).
The SDI and the SAROE are closely related. Yaron (1992b, p. 5) hints at
this when he says that subsidy is less than zero when "the Return on
Equity, net of any subsidy received, equals or exceeds the opportunity
cost of funds." The SDI is negative if and only if the SAROE exceeds the
social opportunity cost.
The proof that a negative SDI implies an SAROE higher than the hur-
dle rate-that is, the opportunity cost of funds-uses the alternative for-
mula for subsidy (equation 2.15), the formula for the change in equity
(equation 2.14), and the formula for average equity (equation 2.13):
S = m  Eo + (m/2) - [DG + PC + RG + A  (m - c) + DX + TPI - TP
= m * Eo + m .(1/2)- AE - TP                       (2.25)
= m  [Eo + (1/2) . AE] - TP
= m  E- TP.



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM              45
This simple formula shows that subsidy S is the opportunity cost of
the equity used in a year less what the DFI could have paid for that equi-
ty and still shown a true profit. A negative SDI implies an SAROE above
the social hurdle rate:
S < O
m . E - TPO 0
m  ES TP                        (2.26)
m S TP/E
opportunity cost of capital S Subsidy-Adjusted ROE.
A strength of the SDI is that it answers the same question as an
SAROE. Figure 2.4 compares ROE and SAROE for the example DFI (table
2.6). In the three years, ROE goes from 0.18 to 0.10 to 0.24. ROE seems to
show that performance improved in the second year and worsened in the
third. In contrast, SAROE goes from -0.28 to -0.10 to 0.10, showing that
improvement was constant. This shows that ROE is not a good measure
of the financial performance of subsidized DFIs. By the third year (when
the SDI was zero), the DFI could have compensated society for its oppor-
tunity cost and still have shown a profit. ROA and SAROA follow the
same pattern as ROE and SAROE (figure 2.5).
If the measure of subsidy in the SDI and the SAROE gives the same
answer for one question, then why use the SDI? After all, the process of
adjusting the financial statements as required to compute the SAROE
helps to ensure that all standard, widely understood financial ratios are
meaningful. The SDI, however, is more than just the measure of subsidy
in its numerator, and thus the SDI has at least three features that the
SAROE does not. First, subsidy independence is zero with the SDI but m
with the SAROE. Given human psychology, naive users may celebrate a
positive SAROE even if it is still less than m. The chances that the SAROE
is positive and yet less than m increase in low-income countries where
inflation may be high and where real interest rates tend to be high due
to the underdevelopment of the financial sector. Second, the SDI is a
measure of the matching grant provided by society (the numerator) for
each dollar of interest paid by the clients of a DFI (the denominator). For
example, if the SDI of a DFI is 1.00, then the SDI contrasts the dollar pro-
vided by society with the dollar provided by the client in a way that the
SAROE does not. Thus, the SDI allows analysts to compare the matching
grant provided to the target group through a DFI with matching grants
(potential or actual) provided through other channels. Third, the SDI
worsens if a DFI abandons its mission and puts resources in investments
other than loans to the target group, all else constant, because revenue
from loans LP . i in the denominator decreases. The SAROE, however,



46                           DEVELOPMENT FINANCE INSTITUTIONS
Figure 2.4. ROE versus           Figure 2.5. ROA versus
SAROE for the Example DFI        SAROA for the Example DFI
0.3 -                            0.3
ROE
0.2    ROE               E       0.21               ROA
0.                                         ROAA  
0         OSAROE                 0
-0.1                             -0.1         SAROA
-0.2-                            -0.2 -SAROA
-0.3                              .3 -
1       2      3                 1       2       3
Year                             Year
stays the same, and may even improve, if the other investments are more
profitable than loans to the target group (box 2.4).
How Does the Subsidy Dependence Index
Change as Its Parts Change?
The SDI has many parts, among them the yield on loans i. Knowledge of
how changes in these parts drive changes in the SDI may help to map
concrete plans to reduce subsidy dependence.
Of course, increased subsidy independence may not always be possi-
ble or even preferred. For example, the DFI controls some parts of the SDI
but not all. Still, knowledge of how social cost might be reduced is always
useful to society.
The figures that follow show how the SDI for Year 01 of the example
DFI changes as one of its parts changes, all else constant. The figures
show the direction of change better than the level of change. The level
depends on the units and on the levels of the part being changed as well
as the units and levels of all other parts of the SDI.
How Does the Subsidy Dependence Index Change
as the Yield on Loans Changes?
As the yield on loans i increases, the SDI decreases at a decreasing rate
(figure 2.6). All else constant, an increase in i increases the denominator
of the SDI because it increases revenue from loans. This decreases the



Table 2.6. ROE, SAROE, ROA, and SAROA
Line                                                                 12/31/01      12/31/02    12/31/03
Ea  Accounting profit, P             Bm                               200            255         935
Eb  Revenue grants, RG               BI                               400            400         400
Ec  Discount on public debt, A  (m - c) Cl                              10            30          50
Ed  Discounts on expenses, DX        Bn                                100           100         100
Ee  True profit, TP                  Ea - (Eb + Ec + Ed)             (310)         (275)         385
Ef  Start equity                     Aot  1                             0          2,200       3,100
Eg  End equity                       Ao                              2,200         3,100       4,600
Eh  Average equity, E                (Ef + Eg)/2                     1,100         2,650       3,850
Ei  Start assets                     Agt_l1                             0          3,000       4,600
Ej  End assets                       Ag                              3,000         4,600       6,800
Ek  Average assets                   (Ei + Ej)/2                     1,500         3,800       5,700
El  ROA                              Ea/Ek                            0.13          0.07        0.16
Em  Subsidy-Adjusted ROA             Ee/Ek                           (0.21)        (0.07)       0.07
En  ROE                              Ea/Eh                            0.18          0.10        0.24
Eo  Subsidy-Adjusted ROE             Ee/Eh                           (0.28)        (0.10)       0.10
Note: Monetary figures in constant units. Average equity includes profit.
Source: Example of authors.



48                             DEVELOPMENT FINANCE INSTITUTIONS
Box 2.4. The Subsidy Dependence Index and
Subsidy-Adjusted Return on Equity for an African DFI
A large African DFI illustrates how the SDI improves when funds shift
away from other investments to loans to the target group, all else constant,
even though the SAROE stays unchanged.
In 1998, the DFI had an average public debt A of 9.91, an opportunity
cost m of 15.5 percent, and an actual rate paid c of 3.9 percent. Average equi-
ty E was 1.5, and K was zero. Of the 9.91 + 1.5 = 11.41 resources in use at the
DFI, 2.47 were in the average loan portfolio LP, and 8.94 were invested in
treasury bills. The yield on loans i was 23 percent, so LP. i was 0.57. Finally,
accounting profit P was -1.42.
Thus, the SDI was [1.5 * 0.155 + 9.91 * (0.155 - 0.039) + 0 - (-1.42)1/0.57
- 2.8/0.57 = 492 percent. The SAROE was -2.8/1.5 --187 percent, far from
the hurdle rate.
With profit held constant, a shift of 1 unit from treasury bills to the loan
portfolio would not affect the SAROE. The SDI, however, would improve
from 492 percent to 2.8/ (3.47 - 0.23) - 350 percent. The SDI is more sensitive
than the SAROE to the allocation of funds between loans to the target group
and other investments.
Source: Authors.
SDI. The increase in i also increases true profit and so decreases subsidy
in the numerator (as long as m < 200 percent). Thus, the effects of an
increase in i in both the numerator and the denominator serve to decrease
the SDI.
How Does the Subsidy Dependence Index Change
as the Rate Paid on Public Debt Changes?
If the measure of equity E in the SDI includes profits in the current
period, then an increase in the rate paid on public debt c decreases the
SDI (figure 2.7; equation 2.15). This happens because a higher rate
paid for debt decreases the equity injected by the discount on public
debt A * (m - c). In turn, this decreases the social cost of the public
funds in the net worth of the DFI and so decreases the SDI. If the mea-
sure of equity E does not include profit in the current period, then a
change in c does not affect the SDI, because the decrease in the dis-
count on public debt A * (m - c) is exactly balanced by an increase in
accounting profit P.



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                      49
Figure 2.6. The SDI and the Yield on Loans, i
SDI
3.00
2.50 -
2.00 -
1.50 -
1.00            Base case
0.50
0.00-
-0.50  -              l     l    l     l
0.2  0.3  0.4  0.5  0.6  0.7  0.8  0.9
Yield on lending, i
Figure 2.7. The SDI and the Rate Paid for Public Debt, c
SDI
1.002
1.001
1.000         Base case
0.999
0.998
0.997
0.996
0     0.05     0.1    0.15    0.2
Rate paid on public debt, c
How Does the Subsidy Dependence Index Change
as the Social Opportunity Cost Changes?
The SDI increases as the social opportunity cost m increases (figure 2.8).
This happens in two ways. First, the increase in m increases social cost
through the discount on public debt A - (m - c) (because the spread



50                            DEVELOPMENT FINANCE INSTITUTIONS
between m and c widens) and decreases the ability to compensate for
social cost through true profit (due to the increase in A - m). Second, the
increase in m increases the social cost of the public funds in the net worth
of the DFI. This is often the bulk of the social cost of a DFI.
Small changes in m can lead to big changes in the SDI (figure 2.8). This
is one reason why the choice of m matters so much. At the same time, the
fact that the SDI depends on m has no policy implication for a DFI. The
opportunity cost of the public funds in the DFI depends on the perfor-
mance not of the DFI but of the marginal public project. (Subsidy-and
the SDI-can be positive even if m is zero because subsidy is the social
opportunity cost of public funds used, less the ability to compensate for
that cost out of true profits. If true profits are negative, then subsidy is
positive even if the use of public funds has no opportunity cost.)
How Does the Subsidy Dependence Index Change
as Administrative Expenses Change?
Increases in administrative expenses decrease true profit and so
increase the SDI (figure 2.9). In the first year of the example DFI,
decreases in administrative expenses could make the subsidy zero. This
is not always the case-social cost could be positive even if administra-
tive costs are zero, for example, if social cost exceeds the level of admin-
istrative costs.
Figure 2.8. The SDI and the Social Opportunity Cost, m
SDI
1.25
1.20
1.15
1.10
1.05
1.00               Base case
0.95
0.90
0.851
0.05   0.075   0.1   0.125   0.15
Social opportunity cost, m



SOCIAL COST OF A PUBLIC DFI IN THE SHORT TERM                     51
Figure 2.9. The SDI and Administrative Expenses
SDI
1.50
Base case/
1.00
0.50
0.00
-0.50
0        300       600       900
Administrative expenses
All else constant, the slope of the graph of figure 2.9 does not depend
on the type of expense that changes. For example, the effect on the SDI of
a $1 change in provision for loan losses is the same as the effect of a $1
change in administrative expenses. Provision for loan losses is discussed
in chapter 4.
How Does the Subsidy Dependence Index Change as Liabilities
Shiftfrom Public Debt to Deposits?
The SDI decreases as deposits replace public debt (figure 2.10). If d is the
cost (both financial and administrative) of deposits, then a shift of one
dollar from public debt to deposits affects subsidy S in three ways. First,
the shift changes the discount on public debt injected in equity by
-(m/2) - (m - c). Because m > c, the discount is negative and so decreas-
es subsidy. Second, the shift changes the true profit injected in equity by
(m/2) . (m - d). True profit increases by m because that is the opportuni-
ty cost of a unit of public debt, but it decreases by d because that is the
unit expense on deposits. Third, the switch changes the true profit avail-
able to compensate for subsidies by (m - d). Because m > d, the change in
true profit is positive. Because in most cases (m/2) < 1, the net effect of
the increase in true profit on subsidy is negative. Because the effect on
subsidy of the decrease in the discount on public debt is also negative, a
shift of one dollar from public debt to deposits decreases subsidy and
thus the SDI.



52                               DEVELOPMENT FINANCE INSTITUTIONS
Figure 2.10. The SDI and the Ratio of Deposits to Public Debt
SDI
1.02
1.01\
\.01  Base case
1.00
0.99
0.98
0.97
0.00        0.33         0.67         1.00
Deposits/(public debt + deposits)



3
What Is a Measure
of the Social Cost of a Public
Development Finance Institution
in the Long Term?
The NPCS is a measure of the social cost of flows of resources between
society and a public DFI in any time frame. Because the SDI does not dis-
count, it does not properly measure social cost in long time frames. Just
like private investors, society should look at the present value of projects
in the long term. Even if society plans to own equity for only a short time,
the present value of the equity at the end of the time frame depends on
expected performance after that point.
The NPCS complements short-term measures of social cost. Just like
dams, DFIs should not be judged only by output in their tenth year but
rather by discounted costs and benefits in their whole lifetimes.
How Does the Net Present Cost to Society
Discount Flows?
What Is the Discount Rate?
The NPCs discounts flows by when they take place in time. The discount
rate is the price of gains and costs in the present in terms of gains and
costs in the future. The social discount rate 6, for a flow one year past the
start of the time frame is one divided by one plus the social opportunity
cost in the first year, m1 (Gittinger 1982).
This assumes that all flows take place at the end of the year. Often the
analyst has year-end financial statements and assumes that flows and
changes in stocks take place at a constant pace in the course of the year.
In this case, flows are discounted as if they took place halfway through
the year. For year t in time frames that last more than one year, the dis-
count rate for a flow or a change in a stock would be:
t-0.5    1  )1-0.5  t-l   1
53



54                            DEVELOPMENT FINANCE INSTITUTIONS
The subscript t is a time index. Likewise, the superscript t - 0.5 is not
an exponent but a label. The opportunity cost m may change through
time.
What Is the Formula of the Net Present Cost to Society?
Outflows from society to a DFI are social costs, and inflows back to soci-
ety from a DFI are social gains. As a cost measure, the NPC5 adds dis-
counted outflows and subtracts discounted inflows. Like all discounted
measures, the NPC5 ignores flows sunk before the start of the time frame.
As presented here, the NPCS assumes all equity injections come from
society, but this assumption can be relaxed (Schreiner 1997).
The stock of equity at the start of the time frame E. is not a sunk
flow. At time 0, society chooses to keep this net worth in a DFI rather
than withdraw it for use elsewhere. Thus, society counts Eo as an out-
flow:
NPCS start net worth = 5 * Eo = Eo.           (3.2)
After the start of the time frame, the DFI builds net worth from fresh
flows of funds FFt from grants, paid-in capital, and discounts. The dis-
counted cost of these outflows from society to the DFI is:
T t 0.5
NPC5 fresh flows of funds =  X  8    * [DGt + PCt + RGt
t=1 
+ At * (mt - c1) + DXtJ         (3.3)
T  t -0.5
t=1 t
True profit accrues through each year. Society could withdraw true
profit as it accrues, but, in the absence of precise knowledge of the tim-
ing of flows, it is assumed that society lets the DFI keep true profit.
Hence, true profit is like an inflow back-to-back with an outflow, and
the two flows cancel out of the NPCS. The treatment here ignores taxes
and dividends, but Schreiner (1997) adjusts the framework to handle
them.
At the end of the time frame, it is assumed that society gets an inflow
equal to the net worth then present in the DFI. Net worth at the end
includes all outflows from society to the DFI up to time T plus true prof-
it. The flow is discounted by Tr:



SOCIAL COST OF A PUBLIc DFI IN THE LONG TERM                  55
T
NPC5 end net worth = 6T ' {EO +  I  [DGt + PCt + RGt
+ At (mt - ct) + DXt + TPt]J         (3.4)
rT - [E + j  (FFt + TPt)]
t=l
The NPCs adds the discounted outflows (equations 3.2 and 3.3) and
subtracts the discounted inflows (equation 3.4):
NPCS = discounted outflows - discounted inflows
T                      T
= E0 +  I  t   * F1 -T [EO + t  (FFt + TPd)]       (3.5)
1=1                    t1
T                       T
T + X  (8t  0.5 -T .F    8 
=(1 -AT )* Eo +         T -AT) *FF- *    TPt.
The NPCS of the flows of funds between society and a DFI from time
0 to time T has three terms. The first term is the cost of funds put in at the
start. For society at time 0, start equity is worth Eo when entrusted to the
DFI at time 0 but only 6  - Eo when it comes back at time T. The cost is
the present value of funds when they are put in less their present value
when they come back.
The second term is the cost of fresh funds FFt injected after the start of
the time frame. Seen from time 0, these funds are worth 8t - 0.5 when
entrusted to the DFI but only 5r when the DFI gives them back. The cost
is the difference.
The third term is the cost (or the gain) of the true profit built up by the
DFI. Society gets this inflow at the end of the time frame, so the discount
factor is T . For most DFIs, the sum of true profit since birth is negative,
and this decreases the inflow back to society. This means that-in real,
nominal, and present-value terms-society gets back fewer dollars than
it put in. Of course, the NPC5-like the SDI-cannot account for social
benefits or costs not reflected in the financial statements of the DFI.
What Questions Does the Net Present Cost
to Society Inform?
The NPC5 informs two important social questions that involve long time
frames. In the first, the NPCS informs the question of whether it improves
social welfare to use public funds to start a new DFI from scratch, ignor-
ing all costs and benefits borne by members of the target group. In the



56                          DEVELOPMENT FINANCE INSTITUTIONS
second, the NPC5 informs the question of whether it improves social wel-
fare to maintain public support for a DFI from now on, again ignoring all
costs and benefits borne by members of the target group.
The NPC5 is negative if the worth of the inflows to society exceeds the
worth of the outflows. Thus, the concept of net present cost mirrors the
concept of present value. If a DFI imposes no costs on nonclients, then a
negative NPC5 indicates that a DFI would be a good social investment
because its return exceeds that of the marginal public project. This requires
true profits so large that, even when discounted from the end of the time
frame back to the start, they exceed the cost of the funds used by the DFI.
Is There a Long-Term Analog to the Subsidy Dependence Index?
A long-term SDI (SDIL) tells the percentage change in revenue from loans
that would make the NPC5 zero (Schreiner 1997). To derive this, first
write true profit TPt as revenue from loans LPt- it plus OROE1, all other
revenues less other expenses:
TPt = LPt* it + OROEt*               (3.6)
Now set the NPCS (equation 3.5) equal to zero and solve for the SDIL,
the percentage change in revenue from loans that would make the NPCs
zero:
O  (  ti )  o +tLl(6tO5 -    t )  UF-T * 1 LPt *it
*(1 + SDIL) + ORoEt]
T                    T               (
0DIL =                  T                           T
t  r    t =             = t     T t1       
With data since birth, the SDIL tells how far a public DFI has been from
subsidy independence since birth. With projected data from now on, the
SDIL predicts how far a DFI is expected to be from subsidy independence.
Is Subsidy in the Subsidy Dependence Index Just the One-Year
Case of Net Present Cost to Society?
If average equity includes true profit, then subsidy in the SDI (equation
2.15) is not just the one-year case of the NPCS (equation 3.5). If true prof-



SOCIAL COST OF A PUBLIC DFI IN THE LONG TERM                  57
it is positive (negative), then subsidy in the SDI is less (more) than in the
NPCS because the SDI does not discount the flow of true profit, implying
less (more) subsidy (Schreiner 1997). The NPCS assumes that true profit
comes at the end of the year, but the SDI assumes that true profit comes
in the middle of the year.
For example, suppose a DFI starts a year with equity of 100. It posts a
true profit of 10 by the end of the year, and it gets no more fresh funds in
the year. In a one-year time frame and with m set at 10 percent, the NPCS
is 0 because true profit is just enough to compensate for the opportunity
cost of public funds in net worth:
T                       T
NPC  =        (1-)Eo + X  (6      -        T )   FFt-TP
t=1         t    ~~~~~~t=1 1
= (1 - 81) *E0 + ( 1 -  81) * FF1 - 81 TP1          38
= (1 - 0.9091). 100 + (0.9535 - 0.9091) 0 - 0.9091 - 10
= 9.09 - 9.09 = 0 .
The measure of subsidy in the framework of the SDI is 0.5:
S =m . E + A . (m - c) + K - P
=0.10 [(100 + 100 + 10)/2] + 0 + 0 - 10    (3.9)
10.5 - 10 = 0.5.
If the measure of average equity excludes true profit, then the SDI does
equal zero, because equity remains unchanged during the year. In this
case, the SDI is the same as the one-year NPCs:
S  m . E + A . (m - c) + K - P,
0.10- [(100 + 100)/2] + 0 + 0 - 10      (3.10)
10 - 10 = 0.
Is the Net Present Cost to Society Better
Than the Subsidy Dependence Index?
Both the SDI and the NPCS work well in short time frames. In practice,
many people understand ROE, and the measure of subsidy in the SDI can
be transformed into a Subsidy-Adjusted ROE. The SDI is useful if the
user wants a quick, crude estimate, if the time frame is short, if inflation
is low, and if the user understands ROE but not the NPCS. Both the SDI
and the SDIL (which uses the NPCs) can be seen as measures of matching
grants provided to members of a target group through a DFI.



58                           DEVELOPMENT FINANCE INSTITUTIONS
Unlike the SDI, the NPCS works in long time frames and is recognized
as the best tool to judge projects (Brigham and Gapenski 1993). Choices
may be sub-optimal investments if based on the SDI instead of on the
NPC,, especially in long time frames and in particular when an SDI that
indicates subsidy independence takes place only after years of SDIs that
indicate subsidy dependence.
What Are the Net Present Cost to Society
and the Subsidy Dependence Index in the Long Term
for the Example Development Finance Institution?
For the example DFI, the NPC5 and SDIL for three time frames that start
at birth (the start of Year 00) are in table 3.1. The first time frame ends at
the end of Year 01, the second at the end of Year 02, and the third at the
end of Year 03. Table 3.1 also shows the one-year NPCS and the SDI based
on the one-year NPCs.
Time Framefrom Birth to the End of Year 01
With the opportunity cost of public funds to society ml set at 10 percent, the
discount rate S1 from the point of view of the birth of the DFI for a flow at the
end of Year 01 is 1/(1 + 0.1) - 0.9091 (line Fc). The discount rate S1 -05 for flows
in the middle of the year is [1/(1 + 0.1)]05- 0.9535 (line Fd). Start equity Eo
for the example DFI is 0 (line Fo). Direct grants DG, were 1,700 and paid-in
capital PC1 was 300. Revenue grants RG, were 400. The discount on public
debt Al - (ml - cl) was 10. Discounts on expenses DX, were 100. Total fresh
hmds FF, were 1,700 + 300 + 400 + 10 + 100 = 2,510 (line Fj). True profit TP,
was -310 (ine Fn). The NPCS0-1 from the start to the end of Year 01 was:
NPC°-= (1- 8)  Eo+ (60 - ) . FF, - S  TPi
-(1 - 0.9091) * 0 + (0.9535 - 0.9091) * 2,510 - 0.909 - (-310) (3.11)
-111.44 + 281.82 - 393.28.
Subsidy in the SDI for Year 01 is 420 (line Cq of table 2.4). Because aver-
age equity includes true profit and because true profit is negative (the
most common case), the SDI is more than the NPCs.
Revenue from loans LP1, i1 is 420. The SDI for the example DFI in Year
01 is 1.00 (line Cx of table 2.4), but the SDI( - 'is (line Fv of table 3.1):
NPCO -        393.27
SDL-==- 1.03.                                    (3.12)
S1 - LP1 * il  0.9091 * 420



Table 3.1 Net Present Cost to Society
Line                                                                 12/31/01       12/31/02    12/31/03
For one-year time frames
Fa  Disc. flow at end of year         1/(1+Cd)                        0.9091        0.9091       0.9091
Fb  Disc. flow middle of year         [1/(1+Cd)]05                    0.9535        0.9535       0.9535
For time frames that start at birth
Fc  Disc. flow at end of year         Fct _ 1/ (1+Cd)                 0.9091        0.8264       0.7513
Fd  Disc. flow middle of year         Fct _ I [1/(1 + Cd)105          0.9535        0.8668       0.7880
Fe  Change direct grants, DG          Df                               1,700           300         300
Ff  Change paid-in capital, PC        Di                                300            345         265
Fg  Revenue grants, RG                BI                                400            400         400
Fh  Discount public debt, A. (m - c)    Cl                               10             30          50
Fi  Discounts on expenses, DX         Bn                                100            100         100
Ul    Fj  Fresh funds in year, FF           Fe + Ff + Fg + Fh + Fi           2,510          1,175       1,115
Fk  Accum. discounted fresh funds    Fkt -1 + Fd . Fj                  2,393          3,412       4,290
Fl  Accum. fresh funds                Flt 1 + Fj                       2,510         3,685        4,800
Fm  True profits, TP                  Dn                               (310)         (275)         385
Fn  Accum. true profit                Fnt1 - + Fm                      (310)         (585)        (200)
Fo  Start equity at birth, Eo         Aoo                                 0              0           0
Fp  Start equity this year, Eo        Ao°  1                              0          2,200        3,100
Fq  One-year NPCS                     (1 - Fa) Fp +Fj * (Fb - Fa) - Fm  Fa 393         502         (19)
Fr  NPC5 from birth                  (1 - Fc) Fo + Fk - Fc . Fl - Fn Fc    393         850         834
Fs  Revenue from lending, LP.i        Ba                                420           1,080       1,700
Ft  Accum. rev. from lending          Ftt -1 + Fs                       420           1,500       3,200
Fu  One-year SDI with NPCS            Fq/(Fa. Fs)                       1.03          0.51       (0.01)
Fv  Long-run SDI                      Fr/(Fc. Ft)                       1.03          0.69         0.35
Note: Monetary figures in constant units.
Source: Example of authors.



60                             DEVELOPMENT FINANCE INSTITUTIONS
The DFI could have been privately profitable with 103 percent more
revenue from loans. With the size of the loan portfolio held constant and
with the actual yield at 0.40, this implies a change in yield of 0.40- 1.03
0.41 and a subsidy-free yield of 0.40 + 0.41 = 0.81.
Time Framefrom Birth to the End of Year 02
With m2 at 10 percent, the discount rate52 from birth to the end of Year 02
is [1/(1 + 0.1)12  0.8264 (line Fc of table 3.1). The discount rate 82 - 0.5
for flows in the middle of the year is [1/(1 + 0.1)1'5 - 0.8668 (line Fd).
Direct grants DG2 are 300, paid-in capital PC2 is 345, revenue grants RG2
are 400, the discount on public debt A2 - (m2 - c2) is 30, and discounts on
expenses DX2 are 100. Total fresh funds FF2 are 300 + 345 + 400 + 30 + 100
= 1,175 (line Fj). True profit TP2 was -275. Thus, the NPCO -2 for the first
two years of the example DFI was:
2                        T
NPC°-2 = (1 - 62). *  +        (Et-+ _E+  2 )* FF - 5 2 *   TP
= (1 - 0.8264) * 0
+ (0.9535 - 0.8264) * 2,510 + (0.8668 - 0.8264)  (3.13)
- 1,175 - 0.8264 * (-310 - 275)
- 319.02 + 47.47 + 483.44 = 849.93.
Revenue from loans LP2* i2 is 1,080. The SDI°-2 is:
NPCO°-2               849.93
DIO-2= -      224                                  0.69.   (3.14)
E I  LPt* it    0.8264 .(420 + 1,080)
t=l
Time Frame from Birth to the End of Year 03
With m3 at 10 percent, the discount rate 833 from birth to the end of Year 03
is [1/(1 + 0.1)]3 - 0.7513 (line Fc of table 3.1). The discount rate d3 - 05 for
constant flows is [1/(1 + 0.1)12.5  0.7880. Direct grants DG3 are 300, paid-
in capital PC3 is 265, revenue grants RG3 are 400, the discount on public
debt A3- (M3 - C3) is 50, and discounts on expenses DX3 are 100. Total fresh
funds FF3 are 300 + 265 + 400 + 50 + 100 = 1,115. True profit TP3 is 385. The
NPCO-3 is:



SOCIAL COST OF A PUBLIC DFI IN THE LONG TERM                  61
3                       3
(t - .5 _8)  F8 
NPCO-3 = (1 - 63). Eo +     (    -3    t- 3  1 TPt
= (1 - 0.7513) . 0 + (0.9535 - 0.7513) . 2,510
+ (0.8668 - 0.7513) * 1,175 + (0.7880 - 0.7513)  (3.15)
* 1,115 - 0.7513. (-310 - 275 + 385)
- 507.52 + 135.71 + 40.92 + 150.26 = 834.41.
Revenue from loans LP3. i3 is 1,700. The SDIO- 3 is:
NPCO-3                834.41
SDIL-3=      3                                       - 0.35. (3.16)
63    LPt it   0.7513- (420 + 1,080 + 1,700)
The example DFI had an actual yield i over the three-year time frame
of 0.40. All else constant, an increase in the yield in each year of 0.40 . 0.35
= 0.14 would have led to subsidy independence. The subsidy-free yield
for the lifetime of the DFI would be 0.40 + 0.14 = 0.54. If the yield on loans
i had been 54 percent in all three years instead of 40 percent, then the
NPCS for the three-year time frame would have been zero.
The example DFI was subsidy-independent in Year 03 (SDI of 0.00)
even though it was not subsidy-independent from birth through Year 03
(SDIL of 0.35). Thus, measurement of the social cost of public DFIs should
include both the short-term SDI and the long-term SDI. The two mea-
sures cannot be compared directly because the SDI uses undiscounted
values and the SDIL uses discounted values.



4
What Are the Pitfalls
When Calculating
the Subsidy Dependence Index
or Net Present Cost to Society?
The use of the SDI and the NPC5 in practice presents two key challenges.
The first is the need to pick a meaningful social opportunity cost. The
second is the need to cope with the constraints of accounting data within
an economic framework.
What Is the Social Opportunity Cost?
The social opportunity cost is defined as the return to public funds in the
marginal public project. Because it is difficult to measure and a wide
range of reasonably defensible estimates of social opportunity costs
appear in the literature, chapter 1 discussed five proxies. Regardless of
the proxy, the SDI and the NPCS are useful inasmuch as they show orders
of magnitude and trends. When the proxy is lower than the true social
opportunity cost, then the SDI and NPCS also are lower bounds on true
social costs. Results in Schreiner (1997) suggest that long-term measures
such as the SDIL are not very sensitive to the social opportunity cost
because the vast bulk of cash flows take place long after the start of the
time frame.
Two other important points about social opportunity costs from chap-
ter 1 are repeated here. First, the social opportunity cost is not necessari-
ly the cost to the DFI of public funds. Second, the social opportunity cost
is not necessarily the opportunity cost of a private entity, that is, the cost
to replace public funds with private funds. It is not uncommon to confuse
social and private opportunity costs.
What Can Be Done to Cope with Accounting Data?
The most important caveat for the SDI and NPC5 is that they use account-
ing data that were not designed for economic (present-value) analysis.
For example, accrued revenue in the income statement may never be col-
lected. Even if the DFI eventually collects all accrued revenue or if it pro-
62



PITFALLS WHEN CALCULATING THE SDI OR THE NPCS                 63
vides for all expected losses from unpaid accrued revenue, the financial
statements still overstate the present value of accrued revenue. In gener-
al, items in the balance sheet are not recorded in terms of their present
values; for example, debt and fixed assets are recorded at their cost to the
DFI. The income statement of the DFI often does not distinguish between
cash items and accrued items, nor does it distinguish between flows at
different times within a reported period.
Of course, the SDI and the NPCS are limited by the data and assump-
tions fed to them. As in all financial analysis, good outputs require good
inputs. This is not a weakness of the measures described here but rather
a standard caveat of all analysis, especially when accounting data are
stretched to fit economic purposes.
Like any disciplined attempt to measure performance, the SDI and the
NPCS are like canaries in a coal mine that serve to unearth deviations
from GAAP and International Accounting Standards. In this case, any
financial indicator based on these weak data would probably be of low
quality. Measurement helps to discover these weaknesses so that they can
be addressed.
By far the two most important problems with the accounting data of
DFIs are the failure to provide properly for loan losses and the failure to
adjust for inflation. Either failure can result in financial statements that
do not accurately reflect financial performance. When failures are sub-
stantial, any financial analysis is either meaningless or misleading.
Why Should a Development Finance Institution Provide
for Loan Losses?
Financial statements should reflect business performance. The business of
DFIs is to produce financial services such as deposits and loans, and, in
general, defaults and loan losses are a normal part of doing business. A
loan may not turn sour for a long time, but a DFI should record the expect-
ed expense of the loss at disbursement. This conservatively reflects that
some loans will go bad, even though, at disbursement, the DFI does not
know which ones (Christen 1997). Ex post write-offs of bad loans under-
state profit in the year of the write-offs and overstate profits in past years.
Thus, a DFI should incur expenses for provision for loan losses con-
stantly as it makes loans. These expenses build the reserve for loan loss-
es, a contra-asset account. The net loan portfolio is the value of loans out-
standing that the DFI expects to recover. It is the gross loan
portfolio-which includes all loans outstanding, some of which will not
be repaid-minus the reserve for loan losses.
If the DFI does not provide enough for loan losses, then it deflates
expenses and inflates profits and net worth. The reserve for loan losses is



64                             DEVELOPMENT FINANCE INSTITUTIONS
too small, and the net portfolio is too big. Other authors discuss how to
estimate the amount of provision for loan losses (Christen 1997; Von
Pischke and others 1988; Bolnick 1988).
Often DFIs do not provide enough for loan losses. This distorts all
financial ratios, including the SDI and the NPCS. An important part of the
business of a DFI is to make loans and to collect them, so any measure of
performance must be based on financial statements that reflect the true
risk of loans in the portfolio.
The example DFI did not provide for loan losses at all (line Bi of table
2.3). Most DFIs, however, cannot recover all their loans. Suppose that in
each of the three years the DFI made 100 in loans that later turned bad.
Thus, in each year the DFI should have incurred expenses of 100 as pro-
vision for loan losses. It is assumed that the DFI did not accrue revenue
from interest. If it had, then the recognition of the bad loans would also
require an adjustment to decrease accrued revenue from loans.
In the adjusted balance sheet, the reserve for loan losses changes by
-100 each year (line Jc of table 4.1). The net portfolio (line Jd) and total
assets (line Jg) shrink in step. Retained earnings (line Jn) fall because
accounting profit falls.
The ripple effects of provision for loan losses are shaded in the adjust-
ed income statement (table 4.2). The expense for loan-loss provisions
increases from zero to 100 in each year (line Ki), and this changes the
operating margin (line Kk) and accounting profit (line Km).
Adequate provision for loan losses also changes the SDI. Without ade-
quate provision, the SDI was 1.00, 0.50, and 0.00 (line Cx of table 2.4). With
adequate provision, the SDI is 1.18, 0.52, and -0.01 (line Gb of table 4.3).
These are small differences, but the effects of proper provisions would be
much larger for many DFIs.
Inadequate provision for loan losses leads to an inaccurate SDI and to
inadequate financial ratios in general. Without provisions, ROE is 0.18,
0.10, and 0.24 (line En of table 2.6). With provisions, ROE falls to 0.10,
0.06, and 0.23 (line Ge of table 4.3). Provisions cause the SAROE to change
from -0.28, -0.10, and 0.10 (line Eo of table 2.6) to -0.37, -0.13, and 0.11
(line Gf of table 4.3). ROA and SAROA follow the same pattern. Table 4.3
shows the SDI, the NPCS, and the SDIL with provisions for loan losses.
Yaron (1992b) provides more discussion of provisions for loan losses.
Why Should a Development Finance Institution
Adjustfor the Effects of Inflation?
Inflation wreaks havoc with financial statements prepared under the
assumption that monetary figures keep a constant value (Goldschmidt,
Shashua, and Hillman 1986). Adjustments help to ensure that the data



Table 4.1. Balance Sheet with Loan Losses
Line                                                                   12/31/01       12/31/02     12/31/03
Assets
Ja   Cash                             Data                                600            700         800
Jb   Loan portfolio (gross)            Data                             2,100          3,300       5,200
Jc   Reserve for loan losses           Data                             (1(X)          (100)        (100)
Jd   Loan portfolio (net), LP I'         + Jc                           2,000          3,200       5,100
Je  Investments, I                     Data                               200            400         600
Jf   Fixed assets (net)                Data                               100            200         200
Jg   Total assets                     Ja + Jd + Je + Jf                 2,900          4,500       6,700
Liabilities
Jh   Deposit liabilities               Data                               200            400         600
Ji   Private debt                      Data                               200            300         400
Jj   Public debt, A                    Data                               400            800       1,200
Jk   Total liabilities                Jh + Ji + Jj                        800          1,500       2,200
Equity
JI   Paid-in capital, PC               Data                               300            645         910
Jm   Direct grants, DG                 Data                             1,700          2,000       2,300
Jn   Retained earnings                Jnt1 + Km                           100            255        1,090
Jo   Total equity                     Jl + Jm + Jn                      2,100          2,900       4,300
jp   Total equity and labilities      Jk + Jo                           2,900          4,400       6,500
Note: Monetary figures in constant units.
Source: Example of authors.



Table 4.2. Income Statement with Loan Losses
Line                                                                 12/31/01      12/31/02     12/31/03
Ka  Revenue from loans, LP  i        Data                               420          1,080       1,700
Kb  Revenue investments, I j         Data                                 5            15          25
Kc  Total revenue from operations    Ka + Kb                            425          1,095       1,725
Kd  Expenses int. deposit liabilities  Data                               5            15          25
Ke  Expenses int. private debt       Data                                10            25          35
Kf  Expenses int. public debt, A  c  Data                                10            30          50
Kg  Total int. exp.                  Kd + Ke + Kf                        25            70         110
Kh  Financial margin                 Kc - Kg                            400          1,025       1,615
Ki  Exp. admin.                      Data                               600          1,170       1,080
Kl  Revenue from grants, RG          Data                               400           400         400
t " ''(mit' 3',P A~zcowt". t';"0S00ing prof00N PM Kk , KI^O 100                       155         835
Memo item:
Kn  Discounts on expenses, DX        Data                               100           100         100
Note: Monetary figures in constant units.
Source: Example of authors.



Table 4.3. Summary with Loan Losses
Line                                                                12/31/01       12/31/02    12/31/03
Ga  Subsidy, S                       Not shown                        495            565        (25)
Gb  Subsidy Dependence Index, SDI   Not shown                         1.18          0.52       (0.01)
Gc  ROA                              Not shown                        0.07          0.04        0.15
Gd  Subsidy-Adjusted ROA             Not shown                       (0.27)        (0.09)       0.07
Ge  ROE                              Not shown                        0.10          0.06        0.23
Gf  Subsidy-Adjusted ROE             Not shown                       (0.37)        (0.13)       0.11
Gg  One-year NPC5                    Not shown                         465           527         (41)
Gh  NPCS from birth                  Not shown                        465            944         910
Gi  One-year SDI with NPCS           Not shown                        1.22          0.54       (0.03)
Gj  Long-run SDI                     Not shown                        1.22          0.76        0.38
Note: Monetary figures in constant units. Average equity includes profit in current period.
Source: Example of authors.



68                           DEVELOPMENT FINANCE INSTITUTIONS
measure what they intend to measure. Just as with provisions for loan
losses, the problem is with the meaningfulness of the data, not with the
measures that use the data.
IAS 29 suggests a few simple adjustments to use if the effects of infla-
tion might affect the results of the analysis. The goal is a set of adjusted
financial statements with the same meaning as unadjusted statements
when prices are stable. Yaron (1992b) also discusses the need to adjust for
inflation. Goldschmidt (1992) discusses IAS 29, and Goldschmidt and
Yaron (1991) outline shortcut methods with numerical examples.
Christen (1997) adjusts for inflation (and for the effects of subsidized
funds) directly and elegantly in the financial statements of an example
DFI. Once adjusted, common financial measures such as ROE are mean-
ingful.
The NPCS requires inflation-adjusted data unless inflation is zero for
the whole time frame. Otherwise, monetary figures from different times
are in different units and cannot be added together. Even if annual infla-
tion is low, inflation adjustments are important in long time frames. If the
SDI and NPCS are applied to inflation-adjusted figures, then opportunity
costs should be in real terms because nominal rates with inflation-adjust-
ed data would count costs twice (Yaron 1992b). The example DFI is
assumed to be in an economy without inflation.
What Are Other Pitfalls and Caveats?
How Can Average Stocks Be Computed?
Given only year-end balance sheets, average stocks are half the sum of
the start and end stocks. This monograph uses this method, but such two-
point averages can mislead if actual cash flows are seasonal, lumpy, or
otherwise nonuniform.
This is a data problem. Balance sheets are snapshots at a moment, and
income statements sum revenues and expenses regardless of when they
took place. A better average requires more frequent data from monthly or
quarterly financial statements.
Is Exemptionfrom Reserve Requirements a Subsidy?
A deposit-taking DFI that is exempt from reserve requirements gets a
subsidy (Benjamin 1994; Yaron 1992b). Reserve requirements are funds left
on deposit with the central bank. They tax financial intermediation by
reducing the return on deposits.
Exemption from reserve requirements lowers the cost to the DFI not
only of deposits but also of equity and liabilities. Let k be the reserve



PITFALLS WHEN CALCULATING THE SDI OR THE NPCS              69
requirement, K the interest rate eamed on required reserves (often zero),
and Dep the average deposit liability. The subsidy for a DFI exempt from
reserve requirements is (Benjamin 1994; Yaron 1992b):
S = m . E + A. (m - c) + RG + DX - P
k. [E (m - a) + (A + Dep) (m - ic)]           (4.1)
1 -k
Without a reserve requirement, k is zero, and the last term vanishes.
Suppose that the example DFI is exempt from a reserve requirement k of
20 percent and that required reserves earn no interest (K = 0). Subsidy in
Year 01 is then:
S = 0.1 * 1,100 + 200- (0.1 - 0.05) + 400 + 100 - 200
+ 0.2 [1,100. (0.1 - 0) + (200 + 100) . (0.01 - 0)]  (4.2)
1 - 0.2
= 110 + 10 + 300 + 0.2- (110 + 30)/0.8 = 420 + 35 = 455.
The SDI without the adjustment for the exemption from reserve
requirements is 420/420 = 1.00 (line Cx of table 2.4). With the exemption,
the SDI is 455 /420 - 1.08.
How Can Exemption from Taxes on Profits Be Handled?
Most DFIs do not pay taxes on profits. This is a subsidy because a tax cut
is like a cash gift. For simplicity, this monograph ignores taxes, but the
frameworks of the SDI and the NPCs can be adjusted to handle taxes
(Schreiner 1997).
How Can Protection from Foreign Exchange Risk
Be Handled?
Some DFIs hold debt denominated in foreign currencies but do not bear
the risk that the exchange rate will change before payment is due. If a
public entity absorbs the risk, then there is a subsidy defined as the dif-
ference in the payment with versus without protection, minus any pre-
mium paid by the DFI for insurance for exchange rate risk. One way to
compute this-analogous to the discount on public debt as A. (m - c)-is
to assume that the DFI would replace foreign exchange with domestic
currency (A) and then compute the subsidy per unit of foreign exchange
as the price of equivalent domestic funds (m) minus the actual cost of for-
eign exchange (c).



70                           DEVELOPMENT FINANCE INSTITUTIONS
How Can Guarantees of Debt Be Handled?
Some DFIs have private debt backed by public guarantees. This debt is
subsidized because the DFI would have to pay more for an equivalent
unguaranteed loan. The subsidy is determined by the difference between
the interest rate with and without the guarantee. Benjamin (1994, see
appendix) provides a framework to estimate the cost of debt in the
absence of guarantees.
How Can Nonfinancial Services Be Handled?
DFIs often produce both financial and nonfinancial services (e.g., busi-
ness training or agricultural extension). In most cases, each line of busi-
ness should be analyzed by itself. Most of the work for the analyst is to
divide the accounts, unless the DFI does it itself. Helms (1998), Christen
(1997), and Yaron (1992b) discuss the issue and give example formats to
help make the division. The fact that public funds are often earmarked for
one line of business may help to simplify the division.
How Can Apex Development Finance Institutions
and Their First-Tier Customers Be Handled?
Apex DFIs make loans to first-tier DFIs that then re-lend to final borrowers
(Gonzalez-Vega 1998). Care is required to make sure that all subsidies in
the chain are counted once and only once. There are three basic guidelines.
First, although apex DFIs often charge the prime rate or some other
"market" rate to DFIs, this rate is subsidized because it is still below the
cost of funds from private sources. If public debt from the apex costs
more than private debt, then the retail DFI would borrow on the market
and skip the hassle of the apex DFI. The market price for a loan to a DFI
is not the prime rate charged to blue-chip private firms but rather the
price that covers all expected costs-including the cost of risk-of a loan
to the DFI.
Second, the analyst must not double-count costs by adding social cost
as seen at the level of the apex to social cost as seen at the level of the first-
tier DFI. To see why not, suppose an apex DFI lends two dollars, one to
each of two first-tier DFIs, and that the first-tier DFIs get no other funds
from anywhere else in the year. If both first-tier DFIs go broke in one year,
then society loses two dollars. The sum of the social cost of two dollars
for the apex and the social cost of two dollars for the first-tier DFIs is four
dollars, and society cannot lose more than it had loaned in the first place.
Third, it does not make sense to analyze only the apex or only the first-
tier DFI. The whole system matters because the price charged by the apex



PITFALLS WHEN CALCULATING THE SDI OR THE NPCS                     71
is like an arbitrary transfer price between two subsidiaries with the same
owner (society). The apex can set' its price high or low to shuffle the rev-
enues and expenses-and the measure of social cost-between the two
tiers. Because public funds are used in both tiers, pricing policy should
not affect the measure of social cost.
As an example, suppose an apex DFI has two identical first-tier cus-
tomers, no debt, and 100 of paid-in capital from public sources through
the year. The apex earns 6 per year on two loans of 50 to the first-tier DFIs
at 6 percent interest. Because revenues are 6 and expenses are assumed to
be zero, profit for the apex is 6. Given K = 0, m = 0.1, and A = 0, subsidy
is 0.1 * 1(100 + 100 + 6)/2] + 0. (0.1 - 0) + 0 - 6 = 4.3 (equation 2.10). The
SDI is 4.3/6 - 0.72 (equation 2.7).
Now suppose that each of the first-tier DFIs has no expenses except for
the 3 paid for their apex debt. Each first-tier DFI has 100 of paid-in capi-
tal from public sources through the year. With 50 of debt and 100 of net
worth, each DFI gets revenue of 1 by lending 150 at an interest rate of
two-thirds percent. Each posts a net return of 1 - 3 = -2. Given that K = 0,
m = 0.1, and A = 50, subsidy for each first-tier DFI is 0.1 * [(100 + 100 -
2)/2] + 50- (0.1 - 0.03) + 0 - (-2) = 9.9 + 3.5 + 2 = 15.4. The SDI is 15.4/1
= 15.4.
The sum of the three measures of subsidy is 4.3 + 15.4 . 2 = 35.1.
Because the revenue from loans to final borrowers is 2, the SDI for the
system would be 35.1/2 = 17.55. But this subsidy is not the social cost of
the system, nor is this SDI the change in revenue from loans needed to
make social cost equal to zero.
To see why, suppose that nothing changes except that the apex
decreases its interest rate to 1 percent. Its profit falls to 1, and subsidy is
0.1 [(100 + 100 + 1)/2] + 0- (0.1 - 0) + 0 - 1 = 9.05. For the first-tier DFIs,
profit increases to 1 - 0.5 = 0.5, and subsidy S is 0.1 * [(100 + 100 + 0.5)/2]
+ 50- (0.1 - 0.01) + 0 - 0.5 = 10.025 + 4.5 - 0.5 = 14.025. The sum of the
three measures of subsidy has changed to 9.05 + 14.025 * 2 = 37.1, and the
SDI is now 37.1/2 = 18.55.
By now the problem is clear. In both cases, all the DFIs were owned by
society, 150 * 2 = 300 was lent to final borrowers, and revenue from loans
to final borrowers was 2. Nothing changed except the transfer price
between the DFIs, yet the supposed measure of social cost changed from
35.1 to 37.1.
The correct approach is to consolidate the financial statements of all
DFIs in the system and then to compute subsidy. This removes the depen-
dence on the transfer price (Stickney and Weil 1994). In this example, con-
solidated net worth is 300, the sum of net worth in each DFI. The debt lia-
bilities of the first-tier DFIs cancel with the loan assets of the apex. This
leaves 300 in consolidated assets as loans to final borrowers. Expenses are



72                           DEVELOPMENT FINANCE INSTITUTIONS
zero, revenues from loans are 2, and profit is 2 - 0 = 2. Subsidy is 0.1 -
[(300 + 300 + 2)/2] + 0 . (0.1 - 0.01) + 0 - 2 = 28.1, and the system SDI is
28.1/2 = 14.05.
How Can Compensating Balances Be Handled?
Some loan contracts require borrowers to maintain a minimum deposit
with the DFI until the loan is repaid. This decreases the effective loan
portfolio LP. For example, a DFI with 100 in loans as assets and 10 in com-
pensating balances as liabilities would have an effective loan portfolio
not of 100 but of 90. Compensating balances are not subsidies because the
(private) borrower accepts them as part of the price of the loan (IADB
1994). No public funds flow.
All else constant, the smaller effective loan portfolio LP due to com-
pensating balances does not affect the SDI nor the NPCs because it does
not affect the revenues, expenses, or net worth of the DFI. It does, how-
ever, increase the yield on loans and thus increase the subsidy-free yield;
revenue from loans LP * i is unchanged, but the effective loan portfolio LP
decreases, so the yield on loans i must increase.
Suppose that all the deposit liabilities of the example DFI (line Ah
in table 2.2) are compensating balances. The average loan portfolio LP
decreases from (0 + 2,100)/2 = 1,050 (line Ct of table 2.4) to (0 + 2,100
- 200)/2 = 950. Revenue from loans stays at 420 (line Ba of table 2.3).
The yield on loans i, however, increases from 420/1,050 = 0.40 to
420/950 - 0.4421 (equation 2.9). The SDI stays at 1.00 because neither
LP nor i appear in it except through revenue from loans LP - i,
unchanged at 950 . 0.4421 - 420. The increase in i, however, increases
the subsidy-free yield from 0.80 (line Caa of table 2.4) to 0.44  (1 +
1.00) = 0.88 (equation 2.17).
Does It Make Sense to Regress the Subsidy Dependence Index
on Itemsfrom the Financial Statements?
It does not make sense to regress the SDI against items from the financial
statements. Regressions assume a stochastic relationship between depen-
dent and independent variables, but the SDI has an exact, known relation-
ship to all items in the financial statements. This requires not statistics but
algebra.
In contrast, it may make sense to regress the SDI on factors not in its
formula. For example, Benjamin (1994) regressed the SDI on the age of a
nonrandom sample of microfinance DFIs. He found that the SDI
decreased with age.



PITFALLS WHEN CALCULATING THE SDI OR THE NPCS                 73
What Are the Key Caveats?
The SDI and the NPCs are subject to six often-misunderstood caveats.
* The SDI does not say that all DFIs should raise interest rates until sub-
sidy is zero.
* Neither the SDI nor the NPCS pretends to answer all questions asked
about financial performance from all points of view. Standard financial
analysis is still useful as long as it uses meaningful data.
* Comprehensive analyses should consider not only the SDI itself but
also the level of subsidy, the actual yield, the subsidy-free yield, and
the absolute change in the yield that would make subsidy zero.
* Neither the SDI nor the NPC5 pretends to measure benefits.
* The SDI and the NPCS require the analyst to find meaningful data and
opportunity costs and to use the results to suggest ways to improve
performance.
• The SDI measures subsidy dependence as seen by society, not private
profitability as seen by a private entity
The SDI and the NPCS are useful as measures of the social cost of DFIs
and thus as part of the process that allots public funds. They are useful
even in the absence of measures of benefits, although the existence of
measures of cost should not be used to advocate for the irrelevance of
benefits. An example is the issue of agricultural extension, often provid-
ed free to clients by agricultural DFIs. It is expensive to measure the ben-
efits of extension. In contrast, it is inexpensive to measure the costs. Once
costs are known, the pursuit of efficiency and improved social welfare
can focus on down-to-earth questions. Is there a new technology that
needs extension to speed its spread? Could farmers pay for it? Should
only rich farmers be asked to pay for it? Should fees be phased in? Would
subcontractors cost less and provide better service than the employees of
the DFI?



5
Recent Proposed Changes
to the Subsidy Dependence
Index
The importance of the social cost of DFIs has prompted several attempts
to refine the SDI or to use other standards to judge performance. This
chapter presents critiques of three recent proposals. They fix what is not
broken, or they tweak the SDI to answer unimportant questions. This
chapter is based on Schreiner and Yaron (1999).
The Subsidy Dependence Ratio of Khandker
In several papers on DFIs in Bangladesh, Khandker proposes the Subsidy
Dependence Ratio (SDR) as an alternative to the SDI (Khandker and
Khalily 1996; Khandker, Khalily, and Khan 1995). Similar measures have
also been proposed by Holtmann and Mommartz (1996), SEEP (1995),
and the IADB (1994).
These authors are concerned that the SDI compares subsidy only with
revenue from loans even though DFIs also get revenue from investments
in nonloan assets such as treasury bills. In principle, a DFI could decrease
its subsidy dependence through increased revenues either from loans or
from investments.
The SDR compares subsidy with revenue both from loans and from
investments. Fixing the fact that the SDR of Khandker, Khalily, and Khan
omits K, if j is the yield on investments and if I is the average investment
so I j is revenue from investments, then the SDR is:
S
SDR =          - + I j               (5.1)
Both the SDR and the SDI have subsidy S in the numerator. Like the
SDI, the SDR is negative if and only if an SAROE exceeds the social
opportunity cost. Thus, the SDR and the SDI do not differ in their most
important aspect, the measurement of subsidy. They differ only in what
they compare with subsidy.
74



RECENT PROPOSED CHANGES TO THE SDI                         75
What Question Does the Subsidy Dependence Ratio Answer?
The SDR tells how much more revenue from loans and investments
would be needed to reach subsidy independence. This is not a very use-
ful question. While most DFIs have some degree of local monopoly and
some freedom to set the price of their loans, DFIs are probably price tak-
ers in the investment market. If a DFI could get a higher rate of return on
investments without more risk, then presumably it would have already
done so (IADB 1994). More importantly, the mission of a DFI is not to
invest in nonloan assets but to make loans to members of a target group.
In general, it is true that a DFI can decrease social cost via any
increased revenue or decreased expense, so it is indeed useful to compare
subsidy not only with revenue from loans but also with other revenue
and expense items. But the biggest, most malleable item is revenue from
loans, and lending is the main purpose of a DFI. DFIs do invest in order
to maintain liquidity and to meet demand from clients for loans and
withdrawals of deposits, but investment is not their main line of busi-
ness.
The numerator of both the SDR and the SDI is subsidy. The denomi-
nator of the SDI is revenue from loans, while the denominator of the SDR
is revenue from loans and from investments. Thus, the SDR is always less
than or equal to the SDI. In almost all cases, the need to maintain liquid-
ity means that investments are nonzero, and so the SDR makes a DFI look
less subsidy-dependent than the SDI. If investments are large compared
with loans-as is the case in some years for some of the DFIs studied by
Khandker-then the SDR is much smaller than the SDI. This misleads
because a DFI cannot increase the rate of return on its investments at will
unless it also assumes more risk and because the purpose of DFIs is to
make loans to the target group. (An SDR of 100 percent implies that a DFI
could become subsidy-independent by doubling the yield on both loans
and investments. Even if the DFI could double the yield on loans, how-
ever, it could not double the yield on investments without incurring
much more risk. Thus, the elimination of subsidy would imply more than
a doubling of the yield on loans, suggesting that the SDR understates
subsidy and overstates subsidy independence.)
For example, the SDR gives an unfair assessment of Grameen Bank,
the best-known DFI in the world (Yaron, Benjamin, and Piprek 1997, p.
146):
[The SDRI results in an understatement of Grameen's dependence
on subsidies, particularly during its initial years of operation, when
a larger share of its financial resources was invested in the capital



76                          DEVELOPMENT FINANCE INSTITUTIONS
market. The measure therefore also underestimates the subsequent
progress Grameen made in reducing its dependence on subsidies as
the share of funds invested in the capital market declined relative to
the share of funds loaned to clients. Following [the logic of the
SDR], a microfinance institution could appear increasingly inde-
pendent of subsidies simply by reducing its loans outstanding.
How Is the Subsidy Dependence Ratio Motivated?
Khandker, Khalily, and Khan justify the SDR as follows (1995, p. 46):
As part of a prudent risk-reducing policy, a financial institution may
diversify its financial resources to maximize expected return and
profit. This needs to be taken into account while calculating the SDI.
Otherwise, even if everything else remains the same, a portfolio mix
can yield a higher profit for a program that diversifies resources
compared to a program that only lends, and consequently, [the] SDI
differs by program.
WNhile more loans may indeed mean more losses if the rush to make
more loans leads to more default, the above claim is weak on two counts.
First, the variance of the SDI across DFIs is not a weakness but a strength.
A measure that did not vary would be useless. Second, the SDI does
account for the diversification of assets because the measure of subsidy in
the numerator includes profit and thus, by definition, all revenue from all
sources, including investments.
Khandker, Khalily, and Khan also offer a second motivation of the SDR
(1995, p. 47):
To the extent that a program always minimizes its income risk
through portfolio diversification, the SDR appears more consistent
than the SDI with such a practice, and consequently is subject to less
variation over time and across programs.
We disagree with this claim on two counts. First, few DFIs minimize
income risk. Indeed, Khandker, Khalily, and Khan (1995) suggest that
DFIs "maximize expected return and profit" (p. 46), which would require
anything but to minimize risk. Second, variation in how funds are split
between investment and lending over time and across programs has the
same effect on the numerator of both the SDR and the SDI. The fact that
the denominator of the SDR is always greater than or equal to the denom-
inator of the SDI means that the SDR will be less than or equal to the SDI.



RECENT PROPOSED CHANGES TO THE SDI                          77
While this does indeed imply that the SDR has less variation than the
SDI, the reduced sensitivity also means that the SDR dampens differences
and so is less useful as a tool to assess performance.
Finally, Khandker, Khalily, and Khan (1995) claim that the SDI pre-
scribes higher yields on loans as the only way to reduce subsidy depen-
dence. This is not true (Yaron 1992a, 1992b). Increased yields on loans
may indeed often be the easiest, quickest, and most practical way to
decrease subsidy dependence, but a DFI that pursues efficiency will also
use economies of scale, high recuperation, decreases in operating costs,
and increases in deposit mobilization. For the example DFI and for the
sample of DFIs in Benjamin (1994), subsidy independence resulted not so
much from increased interest rates as from improved efficiency with age
and growth.
What Is the Subsidy Dependence Ratio
for the Example Development Finance Institution?
The SDR has the same numerator as the SDI but a bigger denominator:
SDR  =m.E+A  .(m-c)+K-P
01       LP. i+lj
0.10 - 1,100 + 200. (0.10 - 0.05) + 500 - 200  (5.2)
1,050 . 0.40 + 100. 0.05
= 420/425  0.988.
The SDI was 420/420 = 1.00 (line Cx of table 2.4), saying that the
example DFI could be subsidy-independent if the yield on loans
increased by 100 percent. In contrast, the SDR says that the example DFI
could be subsidy-independent if the yields both on loans and on invest-
ments increased by 99 percent. Most DFIs are price makers for their loans
to their specific target groups and price takers for their investments.
Thus, a DFI could probably increase the yield on loans but not the yield
on investments.
To see the weakness of the SDR, suppose that the example DFI got an
extra direct grant DG of 1,000 at the start of Year 01 and invested it at a yield
j of 5 percent. If the new direct grant does not increase expenses, then
accounting profits grow by 1,000 .0.05 = 50. Average equity grows by 1,025,
the 1,000 granted at the start of the year plus half of 50, the extra profit from
the investment in the year. The DFI used more public funds but did not
produce any more loans to the target group. The SDI reflects this downturn
in performance because it increases by 0.13, from 1.00 to 1.13:



78                           DEVELOPMENT FINANCE INSTITUTIONS
SD',  m -E + A   (m -c) +K -P
LP i
0.10. (1,100 + 1,025) + 200 (0.10 - 0.05) + 500 - (200 + 50) (53)
1,050 0.40
= 472.5/420 -1.13.
The SDR, in contrast, increases only 0.007, from 0.988 to 0.995:
SDI~j = m  E + A  (m -c) + K -P
LP i + I . j
0.10- (1,100 + 1,025) + 200 * (0.10 - 0.05) + 500 - (200 + 50) (54)
1,050 * 0.40 + (100 + 1,000) * 0.05
= 472.5/475 - 0.995.
Social cost increased from 420 to 472.5, and the DFI produced the same
loans to the target group. How did the performance of the DFI change?
The SDI suggests that it worsened a lot. In contrast, the SDR suggests that
it barely changed.
If (m - j)/j < SDR, then investments of extra direct grants will decrease
the SDR even though the SDI increases. In the example above, m = 0.10, j
= 0.05, and investment of extra direct grants increased the SDR slightly
because (0.10 - 0.05)/0.05 = 1 > SDR - 0.988. Usually, however, (m - j)/Ij
< SDR. For example, if the return on investments j increased from 0.05 to
0.06, then (0.10 - 0.06)/0.06 - 0.667 < SDR - .988. Investment of extra
direct grants still increases the SDI, from 1.00 to 1.10:
,,  m-E+A-(m-c)+K-P
SDI0l =       LP * i
0.10- (1,100 + 1,030) + 200 . (0.10 - 0.05) + 500 - (200 + 60) (55)
1,050 . 0.40
= 463/420 -1.10.
The SDR, however, decreases, from 0.988 to 0.953:
SDI",  m mE +A.(m-c) +K-P
LP * i + I j
0.10 * (1,100 + 1,030) + 200 . (0.10 - 0.05) + 500 - (200 + 60) (5.6)
1,050. 0.40 + (100 + 1,000) . 0.06
= 463/486 -0.953.



RECENT PROPOSED CHANGES TO THE SDI                          79
Investment of extra public funds increased social cost from 420 to 463.
The SDI increased to reflect this, but the SDR decreased, suggesting that
subsidy dependence decreased even though more public resources were
used to produce the same output for the target group. Hence, the SDR is
not a useful measure of subsidy dependence.
The Profitability Gap of Sacay
Three concerns prompted Sacay (1996) to propose the Profitability
Gap (PG) as an alternative to the SDI. First, Sacay wanted to compare
subsidy with equity. Second, Sacay wanted to account for the subsi-
dies implicit when a government allows a DFI to fall below minimum
legal standards for capital adequacy. Third, Sacay said that the SDI
assumes that subsidy can be decreased only by increases in the yield
on loans.
These concerns are unfounded (Belli 1996b). First, a function of the
measure of subsidy in the SDI is already equivalent to an SAROE.
Second, most DFIs meet legal capital requirements. For those DFIs that
do not, the PG proposed by Sacay counts some subsidies twice. Third, the
SDI does not claim that the only way to remove subsidy is to increase the
yield on loans.
What Question Does the Profitability Gap of Sacay Answer?
The PG tells how far from a target SAROE is a DFI that gets subsidies
from an exemption from legal capital standards, but this is not a very use-
ful question. Given a target SAROE of m, the PG of Sacay is:
P - A (m - c) - max (0, Emin  E)     (5.7)
PG  = m  -                  '                 (5.7)___ __ __
E + max (0, Emin - E)
where E min is the minimum equity required by law and
r      0 if 0Ž Emin - E
max (0, Emin_E) =
E) Emir - E if Ermm - E > 0.
Sacay calls max (0, E min  E) the capital deficiency. If capital exceeds the
legal minimum, then the deficiency is zero. Otherwise, it is the minimum
less actual equity.
With no capital deficiency, Emin - E < 0 and so max (0, Emin - E) = 0.
The PG is then:



80                          DEVELOPMENT FINANCE INSTITUTIONS
PG     =rn- P- A. (rn-c)-O0
PGno deficiency     E m  0
E m+A*(m-c)-P                     (5.8)
E
The numerator of the PG with no capital deficiency, except for the lack
of K, is the same as subsidy in the SDI. Without K, donors could force the
PG as low as they like with profit grants. We adjust the PG to prevent this:
G'E  = m Er+A- (m-c)+K-P   S
PGno deficiency         E             E9)
With no capital deficiency, the PG compares subsidy with equity
rather than with revenue from loans. Like the SDI, the PG is negative if
and only if a Subsidy-Adjusted ROE exceeds the opportunity cost m:
PG�O¢rn . F - TP                                     TP
E                                         E
For a capital-deficient DFI, the PG proposed by Sacay (with K added) is:
PG'            m -  P - A . (m - c) - K - (Emin - E)
deficiency Sacay     E + (Emin - E)
F + (Emm -           ~~~~~(5.11)
Emn-m + A.- (mn - c) + K - [P - (Emir - E)
Emin
The PG proposed by Sacay would adjust capital up to its legal mini-
mum, taking the needed capital from profit and making it unavailable to
compensate for subsidies. While it does make sense to charge an oppor-
tunity cost m against the full minimum capital requirement Emin, it does
not make sense to take Emin - E from profit P. This would impute a social
cost of 1 + m for each dollar of capital deficiency, m for the use of the dol-
lar for the year, and 1 because the dollar was used up. But the dollar was
not used up, so the correct PG with capital deficiency should replace E
with Emin but not take the difference from profit:
PG     ~Ernin.rn+A.(rn-c)+K-P                  (12)
PG deficiency =                K                  (5.12)
Emim



RECENT PROPOSED CHANGES TO THE SDI                         81
None of the six example DFIs in Sacay (1996) had capital deficiencies.
Whether the level of capital is adequate or deficient, the social opportu-
nity cost of funds used by a DFI should be adjusted to reflect the risk due
to its leverage (Benjamin 1994).
Decreased Subsidy Dependence through an Increased Yield
on Loans
The SDI does not assume that an increased yield on loans is the only way
to decrease subsidy dependence. Among a host of factors, the SDI
depends on loan recuperation, deposit mobilization, and administrative
costs. The classic statement of the SDI repeatedly insists that a DFI can
decrease its subsidy dependence in many ways (Yaron 1992b, pp. 5, 7, 23).
The Average Subsidy Dependence Index of Hulme
and Mosley
Two important works compute four-year averages of SDIs for 10 DFIs
around the world (Mosley and Hulme 1998; Hulme and Mosley 1996, p.
44). The broad conclusions of these works depend on the average SDIs
because they help to determine which DFIs are analyzed as ones with a
focus on growth and sustainability.
The average SDI of Hulme and Mosley has two problems. First, it can-
not be interpreted as the percentage increase in revenue on loans that
would make subsidy zero. Second, its formula in the one-year case does
not seem meaningful.
The Ratio of Averages and the Average of Ratios
The ratio of averages is not the same as the average of ratios:
(  2a )    c +d)                       (5.13)
c + d      2
2
The SDI is a ratio. Huhne and Mosley computed the average SDI as the
average of ratios, the right-hand side of equation 5.13. But only the ratio
of averages-the left-hand side of equation 5.13-keeps the meaning of
the SDI as the percentage increase in lending that, all else constant, would
make the sum of subsidy through the years zero.



82                            DEVELOPMENT FINANCE INSTITUTIONS
For the first two years of the example DFI, the average SDI computed
as the average of ratios (right-hand side of equation 5.13) is:
a   b~~~    S,  s ~420    540
(+2  ) LPI. +i  LP2  i2    420   1,080  = 0.75. (5.14)
2              2                  2
A 75 percent increase in the yield on loans would increase profit in the
first year by 0.75 * 420 = 315. Using start equity Eo and not average equi-
ty E, this leaves a subsidy of 420 - 315 = 105. In the second year, profits
would increase by 0.75 . 1,080 = 810. This leaves a subsidy of 540 - 810 =
-270. The sum of subsidy in the two years is not zero but 105 - 270 = -165.
In contrast, the ratio of averages (left-hand side of equation 5.13) is:
a 2 b            240   4
(2Sb)                       420 + 540  = 0.64.       (5.15)
(c + d)  LP1 i1 + LP2 i2   420 + 1,080
\ 2 /
A 64 percent increase in the yield on loans would increase profit in the
first year by 0.64 * 420 = 268.8. Using start equity Eo and not average equi-
ty E, this leaves a subsidy of 420 - 268.8 = 151.2. In the second year, prof-
its would increase by 0.64 * 1,080 = 691.2. This leaves a subsidy of 540 -
691.2 = -151.2. The sum of subsidy in the two years is now zero.
In any case, the SDI should not be averaged across years because it is
meaningful only in short time frames. In long time frames, a full picture
of subsidy dependence requires a measure that discounts flows by when
they take place (Schreiner 1997). If, as in Hulme and Mosley, the SDI is
averaged through a long time frame anyway, then the analyst should
divide the sum of subsidy in all years by the sum of revenue from loans
in all years. This would preserve the interpretation of the SDI.
The Loan Portfolio LP as a Proxyfor Average Public Debt A
Public debt A is a liability of a DFI, and the loan portfolio LP is an asset.
In general, the two are not equal. In fact, they differ markedly when the
DFI mobilizes savings or when the DFI has a high ratio of equity to assets.
Hulme and Mosley (1996, p. 92), however, replace A with LP in their mea-
sure of subsidy dependence. They also change the expression (m - c) in
the discount on public debt to (c - m):



RECENT PROPOSED CHANGES TO THE SDI                        83
Subsidy in Hulme and Mosley = m E + LP- (c - m) + K - P.  (5.16)
LP i
The formula in Hulme and Mosley (1996) follows neither the spirit nor
the letter of the SDI. In private correspondence, Hulme and Mosley state
that they deliberately replaced A with LP, but they do not say whether
the switch of c and m (which makes the discount on public debt negative)
is a typographical error. The replacement of A with LP does not make
sense because the social opportunity cost applies to the public funds used
by a DFI, not to the funds loaned to the target group. Otherwise, a DFI
that did not lend would have less subsidy than one that did. In later work
(1998, p. 789), Hulme and Mosley write out the standard SDI formula,
although they do not elaborate on the shift in the tool used to measure
subsidy in DFIs.



Appendix
A Framework to Approximate
the Opportunity Costs
of Private Entities
This appendix presents a framework to approximate the opportunity costs
of debt and equity for private entities. It is based on Benjamin (1994).
What Is the Price of Private Debt?
The price of private debt depends on what kind of debt it is. The two
options are deposits and market debt.
When Will Deposits Replace Public Debt?
If a DFI takes deposits, it is assumed that deposits will replace public
debt. The base cost is taken as the rate the DFI pays on deposits plus a
mark-up of three percentage points (Benjamin 1994; Yaron 1992b). This
assumes that a DFI could attract more deposits at the same rate it pays
now. In practice, the assumed mark-up for administrative costs would be
based on actual expected costs.
The example DFI paid 5 percent on deposits (line He of table A.1). With
the mark-up, private deposits would cost 8 percent (line Hg).
When Will Market Debt Replace Public Debt?
If a DFI does not take deposits, then it is assumed that private debt
replaces public debt. The cost of private debt is taken as the local prime
rate plus a premium for risk. Most DFIs are far riskier than blue-chip,
prime-rate borrowers. In the example, the prime rate is 9 percent (line Hh
of table A.1).
In some cases, a DFI that lost public support might replace some equi-
ty with private debt. But most DFIs are too weak to borrow on the mar-
ket even with equity propped up by subsidized funds. Furthermore,
lenders are unlikely to adjust interest rates more than a few percentage
points to compensate for extra risk. Thus, most DFIs would not replace
public equity with private debt.
84



Table A.1. Private Opportunity Costs
Line                                                                    12/31/01       12/31/02     12/31/03
Ha  Start deposit liabilities          Aht                                  0            200          400
Hb  End deposit liabilities            Ah                                 200            400          600
Hc  Average deposit liabilities, Dep   (Ha + Hb)/2                        100            300          500
Hd  Exp. int. deposit liabilities      Bd                                   5             15           25
He  Rate paid deposit liabilities, d   Hd/Hc                              0.05           0.05        0.05
Hf  Adj. for extra administrative costs  Data                             0.03           0.03        0.03
Hg  Opportunity cost public
debt for deposits, M            He + Hf                            0.08           0.08        0.08
Hh  Prime rate                         Data                               0.09           0.09        0.09
Hi  Age of DFI in years                Data                                 1              2            3
Hj  Premium for age                    2/100/Hi                           0.02         0.010        0.007
Hk  ROE                                En                                 0.18           0.10        0.24
HI  Premium for profitability          See text                           0.00           0.01        0.00
Hm  Opportunity cost, debt for debt, m  Hh + Hj + Hl                      0.11           0.11        0.10
Hn  Start total liabilities            Akt_1                                0            800        1,500
Ho  End total liabilities              Ak                                 800           1,500       2,200
Hp  Average total liabilities          (Hn + Ho)/2                        400           1,150       1,850
Hq  Average equity, E                  Cc                                1,100         2,650        3,850
Hr  Leverage, L                        Hp/Hq                              0.36           0.43        0.48
Hs  Opportunity cost of equity, M      Hm   (1.1 + 0.1 Hr)                0.13           0.13        0.11
Note: Monetary figures in constant units. Average equity includes profit in current period.
Source: Example of the authors based on Benjamin (1994).



86                            DEVELOPMENT FINANCE INSTITUTIONS
How DOES EXPERIENCE AFFECT THE PRICE OF PRIvATE DEBT? Less-experienced
DFIs pay more for private debt. Benjamin (1994) assumes that the experi-
ence premium to be added to the prime rate is 2/100/n, where n is the age
of the DFI in years. All else constant, young DFIs are riskier than old DFIs
because lenders do not know them as well and because they are more like-
ly to go bankrupt. The example DFI adds 0.02 in the first year, 0.01 in the
second year, and about 0.007 in the third year (line Hj of table A.1).
As with the mark-up for administrative costs to handle extra deposits,
this crude assumption is meant to capture the spirit of risk premia for
experience. In most cases, these numbers will provide a consistent base
for comparison. In some cases, however, the analyst can pick risk premia
matched to a specific DFI.
How DOES PROFITABILITY AFFECT THE PRICE OF PRIVATE DEBT? Profitable
DFIs pay less for private debt because they are less risky. Benjamin (1994)
illustrates this with a rule: If the DFI has an ROE of less than zero, then
add 0.03 to the prime rate. If ROE is more than zero but less than the
prime rate, then add 0.02. If ROE is more than the prime rate but less than
twice the prime rate, then add 0.01. Otherwise, add nothing.
For the example DFI, the adjustment for profitability is zero in Years 01
and 03 (line Hl of table A.1). In Year 02, the adjustment is 0.01.
The sum of the prime rate, the adjustment for profitability, and the
adjustment for experience is m, the assumed private opportunity cost of
public debt replaced with private debt. In the example, m is 11 percent in
the first two years and 10 percent in the third year (line Hm of table A.1).
What Is the Price of Private Equity?
Equity costs more than debt because equity is riskier. Benjamin (1994)
estimates the price of private equity M by adding a premium for risk to
the price of private debt m.
Leverage L is the ratio of liabilities to equity. As a DFI has more lever-
age, owners demand a higher Return on Equity (Modigliani and Miller
1958). More debt with a constant amount of equity means more fixed
obligations and thus a higher risk to equity if revenues fall short. A bank-
rupt firm pays creditors before shareholders, so shareholders bear more
risk as a DFI uses more debt. The example DFI does not have much lever-
age, ranging from 0.36 to 0.48 (line Hr of table A.1).
Based on historical data on leverage and ROE in the United States,
Benjamin (1994) related M, the private opportunity cost of equity, to m,
the private opportunity cost of public debt, and to L, leverage:
M = m * (1.1 + 0.1 L).               (A.1)



APPENDIX                                                         87
For example, a DFI without debt would need to pay 1.1 m to attract
private capital. A DFI with a debt: equity ratio of 9: 1 would need to pay
2 . m to attract private capital. For the example DFI, M in its first three
years is 13 percent, 13 percent, and 11 percent (line Hs of table A.1). This
is about two percentage points higher than m.
This framework provides general guidelines and does not substitute
for the judgment and knowledge of the analyst. For most public DFIs, the
risk of the loss of public support and the pure business risk of its untest-
ed financial and organizational technology may swamp the risk due to its
financial leverage.
For the example DFI, table A.2 computes an SDI using the private
opportunity cost. Thus, it measures costs not to society but to private
investors. This private SDI is a measure of the increase in the yield on
loans that, all else constant, would allow a DFI to show a profit and to
compensate for the private opportunity cost of funds, assuming that all
public funds were replaced by private funds. For society, the SDI was
1.00, 0.50, and 0.00 (line Cx of table 2.4). For a private entity, the SDI is
1.07, 0.57, and 0.02 (line Ix of table A.2).



Table A.2. Subsidy Dependence Index with Private Opportunity Costs
Line                                                               12/31/01     12/31/02    12/31/03
Ia  Start equity                    Alt  1 + Amt - 1 + Ant 1           0         2,200       3,100
lb  End equity                      Al + Am + An                   2,200         3,100       4,600
co   Ic   Average equity, E               (la + lb)/2                    1,100         2,650       3,850
Id  Opportunity costs of
private entities, M           Hs                              0.13          0.13        0.11
le  Subsidy on equity, E - M        Ic* Id                           138           333        427
If  Start public debt               Ajt _ 1                            0           400        800
Ig   End public debt                Aj                               400           800       1,200
Ih   Average public debt, A         (If + Ig)/2                      200           600       1,000
Ii  Exp. int. public debt, A, c     Bf                                10            30         50
Ij   Rate paid for public debt, c   Ii/Ih                           0.05          0.05        0.05
Ik  Opportunity cost public debt, m    Hm                           0.11          0.11        0.10
Il   Discount public debt, A  (m - c)   Ih * (lk - Ij)                12            36         47
Im   Revenue grants, RG             BI                               400           400        400
In   Discounts on expenses, DX      Bn                               100           100        100



lo   K                                 Im + In                             500            500          500
Ip   Accounting profit, P              Bm                                  200            255          935
Iq   Subsidy, S                        Ie + Il + Jo - Ip                   450            614           39
Ir   Start loan portfolio (net)        Adt 1                                 0           2,100       3,300
Is   End loan portfolio (net)          Ad                                2,100           3,300       5,200
It   Average loan portfolio (net), LP  (Ir + Is)/2                        1,050          2,700       4,250
Iu   Revenue from loans, LP . i        Ba                                  420           1,080       1,700
Iv   Yield on lending, i               Iu/It                               0.40           0.40        0.40
Iw   Revenue from lending, LP . i      It Iv                               420           1,080       1,700
Ix   Subsidy Dependence Index, SDI    Iq/Iw                                1.07           0.57        0.02
ly   Yield on lending, i               Iv                                 0.40            0.40        0.40
Iz   Change in yield                   ly * Ix                             0.43           0.23        0.01
Iaa  Subsidy-free yield                ly + Iz                             0.83           0.63         0.41
Note: Monetary figures in constant units. Average equity includes profit in current period.
Source: Example of authors.



References
Adams, Dale W. 1988. "The Conundrum of Successful Credit Projects in
Floundering Rural Financial Markets." Economic Development and
Cultural Change 36 (2): 355-68.
Adams, Dale W, and J. D. Von Pischke. 1992. "Microenterprise Credit
Programs: Deja Vu." World Development 20 (10): 1463-70.
Adams, Dale W, Douglas H. Graham, and J. D. Von Pischke. 1984.
Undermining Rural Development with Cheap Credit. Boulder, Colo.:
Westview.
Amin, Sajeda, Ashok S. Rai, and Giorgio Topa. 1999. "Does Microcredit
Reach the Poor and Vulnerable? Evidence from Northern
Bangladesh." Center for International Development Working Paper
28. Harvard University, Cambridge, Mass.
Ballard, Charles, John Shoven, and John Whalley. 1985. "General
Equilibrium Computations of the Marginal Welfare Cost of Taxes in
the United States." American Economic Review 75 (1): 128-38.
Barltrop, Chris J., and Diana McNaughton. 1992. Banking Institutions in
Developing Markets. Vol. 2: Interpreting Financial Statements.
Washington, D.C.: World Bank.
Belli, Pedro. 1996a. Handbook on Economic Analysis of Investment
Operations. Washington, D.C.: World Bank.
.1996b. "Measurement of Sustainability." Memorandum to Orlando
Sacay, October 9, World Bank, Washington, D.C.
Benjamin, McDonald P., Jr. 1994. "Credit Schemes for Microenterprises:
Motivation, Design, and Viability." Ph.D. diss. Georgetown University,
Washington, D.C.
Besley, Timothy. 1994. "How Do Market Failures Justify Intervention in
Rural Credit Markets?" World Bank Research Observer 9 (1): 27-47.
90



REFERENCES                                                    91
Binswanger, Hans P., and Shahidur R. Khandker. 1995. "The Impact of
Formal Finance on the Rural Economy of India." Journal of Development
Studies 32 (2): 234-62.
Bolnick, Bruce R. 1988. "Evaluating Loan Collection Performance: An
Indonesian Example." World Development 16 (4): 501-10.
Bolnick, Bruce R., and Eric R. Nelson. 1990. "Evaluating the Economic
Impact of a Special Credit Programme: KIK/KMKP in Indonesia."
Journal of Development Studies 26 (2) : 299-312.
Brent, Robert J. 1996. Applied Cost-Benefit Analysis. Cheltenham: Edward Elgar.
Brigham, Eugene F., and Louis C. Gapenski. 1993. Intermediate Financial
Management. Fort Worth: Dryden.
Buckley, Graeme. 1997. "Microfinance in Africa: Is It Either the Problem
or the Solution?" World Development 25 (7): 1081-93.
Carter, Michael R. 1989. "The Impact of Credit on Peasant Productivity and
Differentiation in Nicaragua." Journal of Development Economics 31: 13-36.
Carter, Michael R., and Pedro Olinto. 1996. "Getting Institutions Right for
Whom? The Wealth-Differentiated Impact of Property Rights Reform
on Investment and Income in Rural Paraguay." University of
Wisconsin-Madison. Processed.
Chaves, Rodrigo A., and Claudio Gonzalez-Vega. 1996. "The Design of
Successful Rural Financial Intermediaries: Evidence from Indonesia."
World Development 24 (1): 65-78.
Christen, Robert Peck. 1997. Banking Services for the Poor: Managing for
Financial Success: An Expanded and Revised Guidebook for Development
Finance Institutions. Boston: Acci6n International.
Christen, Robert Peck, Elisabeth Rhyne, Robert C. Vogel, and Cressida
McKean. 1995. "Maximizing the Outreach of Microenterprise Finance:
An Analysis of Successful Microfinance Programs." Program and
Operations Assessment Report 10. United States Agency for
International Development, Washington, D.C.
Coleman, Brett E. 1999. "The Impact of Group Lending in Northeast
Thailand." Journal of Development Economics 60: 105-41.



92                            DEVELOPMENT FINANCE INSTITUTIONS
Conrad, Douglas A. 1984. "Returns on Equity to Not-for-Profit Hospitals:
Theory and Implementation." Health Services Research 19 (1): 42-63.
. 1986. "Returns on Equity for Not-for-Profit Hospitals: A
Commentary and Elaboration." Health Services Research 21 (1): 17-20.
Dasgupta, Ajit K., and D. W. Pearce. 1978. Cost-Benefit Analysis: The Theory
and Practice. London: Macmillan.
David, Cristina C., and Richard L. Meyer. 1983. "Measuring the Farm
Level Impact of Agricultural Loans." In J. D. Von Pischke, Dale W
Adams, and Gordon Donald, eds., Rural Financial Markets in Developing
Countries. Baltimore: Johns Hopkins University Press.
Devarajan, Shantayanan, Lyn Squire, and Sethaput Suthiwart-Narueput.
1997. "Beyond Rate of Return: Reorienting Project Appraisal." World
Bank Research Observer 12 (1): 35-46.
The Economist. 1997. "Valuing Companies: a Star to Sail By?" August 1
53-55.
Feder, Gershon, Laurence J. Lau, Justin Y. Lin, and Xiapeng Luo. 1990.
"The Relationship between Credit and Productivity in Chinese
Agriculture: A Microeconomic Model of Disequilibrium." American
Journal of Agricultural Economics 72 (5): 1152-57.
Gale, William G. 1991. "Economic Effects of Federal Credit Programs."
American Economic Review 81 (1): 33-52.
Garber, Alan M., and Charles E. Phelps. 1997. "Economic Foundations of
Cost-Effectiveness Analysis." Journal of Health Economics 16: 1-31.
Gittinger, J. Price. 1982. Economic Analysis of Agricultural Projects.
Baltimore: Johns Hopkins University Press.
Goldschmidt, Yaaqov. 1992. "International Accounting Standard 29:
Formulation and Clarification of Income Measurement in
Hyperinflationary Economies." International Journal of Accounting 27:
137-50.
Goldschmidt, Yaaqov, and Jacob Yaron. 1991. "Inflation Adjustment of
Financial Statements: Application of International Accounting
Standard 29." Working Paper 670. World Bank, Washington, D.C.



REFERENCES                                                    93
Goldschmidt, Yaaqov, Leon Shashua, and Jimmye S. Hillman. 1986. The
Impact of Inflation on Financial Activity in Business with Applications to the
U.S. Farming Sector. Lanham, Md.: Rowman and Littlefield.
Gonzalez-Vega, Claudio. 1998. "Microfinance Apex Mechanisms:
Concepts, Synthesis of Lessons, and Recommendations." Report to the
Consultative Group to Assist the Poorest. Ohio State University,
Columbus.
Gonzalez-Vega, Claudio, Mark Schreiner, Richard L. Meyer, Jorge
Rodriguez-Meza, and Sergio Navajas. 1997. "BancoSol: The Challenge
of Growth for Microfinance Organizations." In Hartmut Schneider,
ed., Microfinancefor the Poor? Paris: OECD (Organisation for Economic
Co-operation and Development).
Hashemi, Syed M. 1997. "Building up Capacity for Banking with the
Poor: The Grameen Bank of Bangladesh." In Hartmut Schneider, ed.,
Microfinancefor the Poor? Paris: OECD.
Heckman, James J., and Jeffrey A. Smith. 1995. "Assessing the Case for
Social Experiments." Journal of Economic Perspectives 9 (2) : 85-110.
Helms, Brigit S. 1997. "Anatomy of a Micro-Finance Deal: The New
Approach to Investing in Micro-Finance Institutions." Consultative
Group to Assist the Poorest Focus Note 9. World Bank, Washington, D.C.
_   1998. "Cost Allocation  for Multi-Service  Microfinance
Institutions." Consultative Group to Assist the Poorest Occasional
Paper 2. World Bank, Washington, D.C.
Holtmann, Martin, and Rochus Mommartz. 1996. Technical Guide for
Analyzing the Efficiency of Credit-Granting Non-Government
Organizations (NGOs). Saarbrucken: Breitenbach.
Hossain, Mahabub. 1988. "Credit for Alleviation of Rural Poverty: The
Grameen Bank in Bangladesh." Research Report 65. International Food
Policy Research Institute, Washington, D.C.
Hulme, David, and Paul Mosley. 1996. Finance Against Poverty, Vols. I and
II. London: Routledge.
IADB (Inter-American Development Bank). 1994. Technical Guide for the
Analysis of Microenterprise Finance Institutions. Washington, D.C.



94                           DEVELOPMENT FINANCE INSTITUTIONS
IMF (International Monetary Fund). Various years. International Financial
Statistics. Washington, D.C.
Iqbal, Farrukh. 1986. "The Demand and Supply of Funds among
Agricultural Households in India." In Inderjit Singh, Lyn Squire, and
John Strauss, eds., Agricultural Household Models. Baltimore: Johns
Hopkins University Press.
Jennings, Edward H. 1993. "Financial Management in Not-for-Profit
Businesses." In Eugene F. Brigham and Louis C. Gapenski, eds.,
Intermediate Financial Management. Fort Worth: Dryden.
Katz, David A., and H. Gilbert Welch. 1993. "Discounting in Cost-
Effectiveness Analysis of Healthcare Programmes." PharmacoEconomics
3 (4): 276-285.
Khandker, Shahidur R. 1996. "Grameen Bank: Impact, Costs, and
Program Sustainability." Asian Development Review 14 (1): 97-130.
Khandker, Shahidur R., and Baqui Khalily. 1996. "The Bangladesh Rural
Advancement Committee's Credit Programs: Performance and
Sustainability." World Bank Discussion Paper 324. Washington, D.C.
Khandker, Shahidur R., Baqui Khalily, and Zahed Khan. 1995. "Grameen
Bank: Performance and Sustainability." World Bank Discussion Paper
306. Washington, D.C.
Khandker, Shahidur R., Zahed Khan, and Baqui Khalily. 1995.
"Sustainability of a Government Targeted Credit Program: Evidence
from Bangladesh." World Bank Discussion Paper 316. Washington,
D.C.
Koch, Timothy W. 1992. Bank Management. Fort Worth: Dryden.
Krahnen, Jan Pieter, and Reinhard H. Schmidt. 1994. Development Finance
as Institution Building. Boulder, Colo.: Westview.
Ladman, Jerry R., and Ronald L. Tinnermeier. 1981. "The Political
Economy of Agricultural Credit: The Case of Bolivia." American Journal
of Agricultural Economics 62: 66-72.
Lapar, Ma. Lucila A., Douglas H. Graham, Richard L. Meyer, and David
S. Kraybill. 1995. "Selectivity Bias in Estimating the Effect of Credit on



REFERENCES                                                    95
Output: The Case of Rural Nonfarm Enterprises in the Philippines."
Economics and Sociology Occasional Paper 2231. Ohio State
University, Columbus.
Lipton, Michael, and Martin Ravallion. 1995. "Poverty and Policy." In
Jere Behrman and T. N. Srinivasan, eds., Handbook of Development
Economics. Vol. IIIB. Amsterdam: Elsevier.
Markandya, Anil, and David W. Pearce. 1991. "Development, the
Environment, and, the Social Rate of Discount." World Bank Research
Observer 6 (2): 137-52.
McKeman, Signe-Mary. 1996. "The Impact of Micro-Credit Programs on
Self-Employment Profits: Do Non-Credit Program Aspects Matter?
Essays on Micro-Credit Programs: The Grameen Bank Experience."
Ph.D. diss. Brown University, Providence, R.I.
McKnelly, Barbara, and Christopher Dunford. 1998. Impact of Credit with
Education on Mothers' and Their Young Children's Nutrition: Lower Pra
Rural Bank Credit with Education Program in Ghana. Davis, Calif.:
Freedom from Hunger.
Mishan, E. J. 1988. Cost-Benefit Analysis. New York: Praeger.
Modigliani, F., and M. Miller. 1958. "The Cost of Capital, Corporation Finance,
and the Theory of Investment." American Economic Review 49 (2): 261-97.
Moffitt, Robert. 1991. "Program Evaluation with Nonexperimental Data."
Evaluation Review 15 (3) : 291-314.
Montgomery, Mark, Terry Johnson, and Syed Faisal. 2000. "Who Succeeds
at Starting a Business? Evidence from the Washington Self-Employment
Demonstration." Grinnell College, Grinnell, Iowa. Processed.
Morduch, Jonathan.1998. "Does Microfinance Really Help the Poor? New
Evidence from Flagship Programs in Bangladesh." Princeton
University, Princeton, N.J. Processed.
. 1999. "The Role of Subsidies in Microfinance: Evidence from the
Grameen Bank." Journal of Development Economics 60: 229-48.
. 2000. "The Microfinance Schism." World Development 28 (4)
617-29.



96                           DEVELOPMENT FINANCE INSTITUTIONS
Mosley, Paul. 1996. "Metamorphosis from NGO to Commercial Bank: The
Case of BancoSol in Bolivia." In David Hulme and Paul Mosley, eds.,
Finance Against Poverty. Vol. II. London: Routledge.
Mosley, Paul, and David Hulme. 1998. "Microenterprise Finance: Is There
a Conflict between Growth and Poverty Alleviation?" World
Development 26 (5): 783-90.
Mould, Maurice C. 1987. Financial Information for Management of a
Development Finance Institution. World Bank Technical Paper 63,
Industry and Finance Series. Washington, D.C.
Norgaard, Richard B., and Richard B. Howarth. 1992. "Sustainability and
Intergenerational Environmental Rights: Implications for Benefit-Cost
Analysis." In John M. Reilly and Margot Anderson, eds., Economic
Issues in Global Climate Change: Agriculture, Forestry, and Natural
Resources. Boulder, Colo.: Westview.
Pauly, Mark V. 1986. "Returns on Equity for Not-for-Profit Hospitals."
Health Services Research 21 (1): 1-16.
Paxton, Julia. 1996. A Worldwide Inventory of Microfinance Institutions.
Washington, D.C.: World Bank.
Pitt, Mark M., and Shahidur R. Khandker. 1998. "The Impact of Group-
Based Credit Programs on Poor Households in Bangladesh: Does the
Gender of the Participants Matter?" Journal of Political Economy 106
(5): 958-96.
Quirk, James, and Katsuaki Terasawa. 1991. "Choosing a Government
Discount Rate: An Alternative Approach." Journal of Environmental
Economics and Management 20: 16-28.
RESULTS International. 1996. "The Microcredit Summit Declaration and
Plan of Action." Journal of Developmental Entrepreneurship 1 (2): 131-76.
Richardson, David C. 1994. "PEARLS: Financial Stabilization,
Monitoring, and Evaluation." Research Monograph Series 4. World
Council of Credit Unions, Madison, Wis.
Rogaly, Ben. 1996. "Micro-Finance Evangelism, 'Destitute Women,' and
the Hard Selling of a New Anti-Poverty Formula." Development in
Practice 6 (2): 100-12.



REFERENCES                                                    97
Rosenberg, Richard. 1994. "Beyond Self-Sufficiency: Licensed Leverage
and Microfinance Strategy." USAID (U. S. Agency for International
Development) Microenterprise Development Brief 17, Washington,
D.C. Processed.
.1996. "Microcredit Interest Rates." Consultative Group to Assist the
Poorest Occasional Paper 1. World Bank, Washington, D.C. Processed.
Rosenberg, Richard, Robert Peck Christen, and Brigit Helms. 1997.
"Format for Appraisal of Micro-Finance Institutions." Consultative
Group to Assist the Poorest Technical Tools Series 4. World Bank,
Washington, D.C. Processed.
Sacay, Orlando. 1996. "Measuring Subsidy and Sustainability of
Microfinance Institutions." World Bank, Washington, D.C. Processed.
Sacay, O., B. Randhawa, and M. Agabin. 1996. "The BAAC Success Story:
A Specialized Agriculture Bank under Government Ownership."
World Bank, Washington, D.C. Processed.
Schmidt, Reinhard H. 1997. "Corporate Governance: The Role of Other
Constituencies." In Alice Pezard and Jean-Marie Thiveaud, eds.,
Corporate Governance: Les Perspectives Internationales. Paris: Association
d'Economie Financiere.
Schmidt, Reinhard, and Claus-Peter Zeitinger. 1996. "The Efficiency of Credit-
Granting NGOs in Latin America." Savings and Development 20 (3): 353-84.
Schreiner, Mark. 1995. "Meta-rules." Economics and Sociology
Occasional Paper 2268. Ohio State University, Columbus.
. 1997. "A Framework for the Analysis of the Performance and
Sustainability of Subsidized Microfinance Organizations with
Application to BancoSol of Bolivia and Grameen Bank of Bangladesh."
Ph.D. diss. Ohio State University, Columbus.
. 1999a. Aspects of Outreach: A Framework for the Discussion of the
Social Benefits of Microfinance. Center for Social Development Working
Paper 99-3. Washington University in St. Louis.
. 1999b. "A Cost-Effectiveness Analysis of the Grameen Bank of
Bangladesh." Center for Social Development Working Paper 99-5.
Washington University in St. Louis.



98                           DEVELOPMENT FINANCE INSTITUTIONS
. 2000a. "A Framework for Financial Benefit-Cost Analysis of
Individual Development Accounts at the Experimental Site of the
American Dream Demonstration." Center for Social Development,
Washington University in St. Louis. Processed.
____. 2000b. "Resources Used in 1998 and 1999 to Produce Individual
Development Accounts in the Experimental Program of the American
Dream Demonstration at the Community Action Project of Tulsa
County." Center for Social Development, Washington University in St.
Louis. Processed.
SEEP. 1995. Financial Ratio Analysis of Micro-Finance Institutions. New
York: Pact.
Sherraden, Michael. 1991. Assets and the Poor: A New American Welfare
Policy. Armonk, N.Y: M. E. Sharpe.
Sherraden, Michael, Lissa Johnson, Margaret Clancy, Sondra Beverly,
Mark Schreiner, Min Zhan, and Jami Curley. 2000. "Savings Patterns in
IDA Programs-Downpayments on the American Dream Policy
Demonstration, a National Demonstration of Individual Development
Accounts." Center for Social Development, Washington University in
St. Louis.
Sial, Maqbool H., and Michael R. Carter. 1996. "Financial Market
Efficiency in an Agrarian Economy: Microeconometric Analysis of the
Pakistani Punjab." Journal of Development Studies 32 (5): 771-98.
Silvers, J. B., and Robert T. Kauer. 1986. "Returns on Equity for Not-for-
Profit Hospitals: Some Comments." Health Services Research 21 (1)
21-28.
Singh, Inderjit, Lyn Squire, and John Strauss. 1986. "A Survey of
Agricultural Household Models: Recent Findings and Policy
Implications." World Bank Economic Review 1 (1) : 149-79.
Smith, Stephen C., and Sanjay Jain. 1998. "Village Banking and Maternal
and Child Health: Theory and Preliminary Evidence from Honduras
and Ecuador." George Washington University, Washington, D.C.
Processed.
Stickney, Clyde P., and Roman L. Weil. 1994. Financial Accounting: An
Introduction to Concepts, Methods, and Uses. Fort Worth: Dryden.



REFERENCES                                                      99
Stiglitz, Joseph E. 1993. "The Role of the State in Financial Markets." In
Proceedings of the World Bank Annual Conference on Development
Economics 1993. Washington, D.C.: World Bank.
- 1998. "The Private Uses of Public Interests: Incentives and
Institutions." Journal of Economic Perspectives 12 (2) : 3-22.
Stiglitz, Joseph E., and Andrew Weiss. 1981. "Credit Rationing in Markets
with Imperfect Information." American Economic Review 71 : 393-410.
Tollison, R. 1984. "Politics without Romance: A Sketch of Positive Public
Choice Theory and Its Normative Implications." In R. Tollison and J.
Buchanan, eds., The Theory of Public Choice-Il. Ann Arbor: University of
Michigan Press.
Tully, Shawn. 1993. "The Real Key to Creating Wealth." Fortune,
September 20, pp. 143, 160-2.
- 1994. "America's Best Wealth Creators." Fortune, November 28,
pp. 38-50.
Tweeten, Luther. 1992. Agricultural Trade: Principles and Policies. Boulder,
Colo.: Westview.
Von Pischke, J. D. 1991. Finance at the Frontier: Debt Capacity and the Role of
Credit in the Private Economy. Washington, D.C.: World Bank.
-  1996. "Measuring the Trade-Off between Outreach and
Sustainability of Microenterprise Lenders." Journal of International
Development 8 (2) : 225-39.
. 1998. "Measuring the Performance of Small Enterprise Lenders."
In Mwangi S. Kimeny, Robert C. Wieland, and J. D. Von Pischke, eds.,
Strategic Issues in Microfinance. Brookfield: Avebury.
Von Pischke, J. D. and Dale W Adams. 1980. "Fungibility and the Design
and Evaluation of Agricultural Credit Projects." American Journal of
Agricultural Economics 62: 719-24
Von Pischke, J. D., Robert C. Vogel, Peter Flath, and Maurice C. Mould.
1988. "Measurement of Loan Repayment Performance." Economic
Development Institute Course Note 030/086. World Bank,
Washington, D.C.



100                           DEVELOPMENT FINANCE INSTITUTIONS
Weinstein, Milton C., and William B. Stason. 1977. "Foundations of Cost-
Effectiveness Analysis for Health and Medical Practices." New England
Journal of Medicine 296 (13): 716-21.
Wheeler, John R. C., and Jan P. Clement. 1990. "Capital Expenditure
Decisions and the Role of the Not-for-Profit Hospital: An Application
of a Social Goods Model." Medical Care Review 47 (4): 467-86.
Women's World Banking. 1995. "The Missing Links: Financial Systems
that Work for the Majority." New York. Processed.
World Bank. 1989. World Development Report 1989: Financial Systems and
Development. New York: Oxford University Press.
. 1994. "Mexico: Agricultural Sector Memorandum." World Bank,
Natural Resources and Poverty Division and Latin America and
Caribbean Regional Office, Country Department II, Washington, D.C.
Yaron, Jacob. 1992a. Successful Rural Finance Institutions. World Bank
Discussion Paper 150. Washington, D.C.
. 1992b. Assessing Development Finance Institutions: A Public Interest
Analysis. World Bank Discussion Paper 174. Washington, D.C.
. 1994. "What Makes Rural Finance Institutions Successful?" World
Bank Research Observer 9 (9) : 49-70.
Yaron, Jacob, McDonald Benjamin, and Gerda Piprek. 1997. Rural Finance:
Issues, Design, and Best Practices. Environmentally Sustainable
Development Studies and Monographs Series 14. World Bank,
Washington, D.C.






The term developmentJfinance institutions, or DFIs, refers to public or private
financial intermediaries that use resources from government or donors to
improve social welfare. As funds for development become increasingly scarce,
measuring the social cost of public DFIs becomes more important. If the
social benefit of a DFI exceeds its social cost, then it can be said that the
public funds are well spent.
Developmnent Finance Institutions: Measurin7g Their Subsidy presents two
means for measuring social cost: the Subsidy Dependence Index and the Net
Present Cost to Society. Both measures advance on common financial ratios
because they shift the paradigm from reported (accounting) costs-many of
which are routinely subsidized-to opportunity (economic) costs, leading to
a more efficient determination of the sustainability of a DFI. Sustainability
improves social welfare if the consequent long-term increase in the length,
breadth, scope, and quality of outreach compensates for the short-term
increase in costs shifted to the target group. As presented in this book, use
of the Subsidy Dependence Index and the Net Present Cost to Society to
measure social cost constitutes a first step toward the better informed use
of public funds.
THE WORLD BANK
1818 [H Street, N.W.
Washington, D.C. 20433 LU.S.A.
Telephone: 202-47-1234                             H
Facsimile: 202-47-6391
Internet: www.worldbank.org
E-mail: feedback(@worldbank.org                           ISBN 0-8213-4984-8