REGULATORY EFFECTIVENESS
AND THE EMPIRICAL IMPACT OF
VARIATIONS IN REGULATORY GOVERNANCE
ELECTRICITY INDUSTRY CAPACITY
AND EFFICIENCY IN DEVELOPING COUNTRIES
By
John Cubbin, City University, and Jon Stern, London Business School
ABSTRACT
This paper assesses for 28 developing countries over the period 1980-2001 whether
the existence of a regulatory law and higher quality regulatory governance are
significantly associated with superior electricity outcomes. The analysis draws on
theoretical and empirical work on the impact of independent central banks and of
developing country telecommunications regulators. The empirical analysis concludes
that a regulatory law and higher quality governance are positively and significantly
associated with higher per capita generation capacity levels. In addition, this positive
impact continues to increase for at least three years and probably for over 10 years as
experience develops and regulatory reputation grows. The results are robust to
alternative dynamic specifications and show no sign of any significant endogeneity
biases.
World Bank Policy Research Working Paper 3535, March 2005
The Policy Research Working Paper Series disseminates the findings of work in progress to encourage
the exchange of ideas about development issues. An objective of the series is to get the findings out
quickly, even if the presentations are less than fully polished. The papers carry the names of the
authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in
this paper are entirely those of the authors. They do not necessarily represent the view of the World
Bank, its Executive Directors, or the countries they represent. Policy Research Working Papers are
available online at http://econ.worldbank.org.
1
1. Introduction
Over the last 10-15 years, a lot of attention has been given to the role of institutions in
economic growth. This has, in large part, been driven by economic policy priorities
such as how to develop effectively functioning market economies in Central and
Eastern Europe and the former Soviet Union post-1989; and how to foster economic
growth in lagging world regions such as Sub-Saharan Africa. In parallel, and partly in
response, there have been major explorations of the role of institutions in the
functioning of market economies both by economists (e.g. the literature arising out of
Williamson's transaction cost economics approach) and by economic historians (e.g.
North (1990) and others).
In recent years, there has also been a substantial empirical literature on the relative
roles of institutions, policy, geography and trade openness on growth performance
across countries. This literature currently indicates that institutional quality is the
dominant determinant of variations in long-term growth performance1. Good
institutions embody a heritage of past good policy decisions and themselves generate
a flow of superior policy decisions that support sustained investment and productivity
growth2. In his recent survey on growth strategies, Rodrik (2003) argues that,
although it is quite possible to achieve short-term growth accelerations (e.g. of 10
years or more) with very limited institutional change; the main requirement to ensure
sustained growth and convergence with the living standards in advanced countries "...
is the acquisition of high quality institutions". In particular, he argues that there is a
requirement for a "... cumulative process of institution building to ensure that growth
does not run out of steam and that the economy remains resilient to shocks" 3.
Infrastructure industries are not just a microcosm of the aggregate economy. The
arguments above on aggregate growth apply with extra force to utility service
industries. This is because they are not only highly capital intensive, but also most of
their assets are very long-lived and (in economic terms) sunk assets. Hence, an
effective institutional framework is essential to sustain growth in output, efficiency
and capacity for commercialized utility service industries such as electricity,
telecommunications, water and others - particularly if these industries have significant
amounts of private investment (physical and/or financial).
This paper was prepared as part of the research program on Industrial Organization Policy for
Development at the Development Research Group of the World Bank, under the direction of
Ioannis Kessides. We are grateful for comments on this paper from seminar participants at the
University of Cambridge, at City University and at the ISNIE 2004 Conference. We would
particularly like to thank Dimitrios Asteriou, Richard Gilbert, Jean-Michel Glachant and
Witold Henisz for helpful discussions, Peter Burridge for econometric advice and Evens
Salies for initial assistance with the data collection. The authors alone are, however, solely
responsible for the analysis and the views expressed for which none of the above take any
responsibility and which should not be attributed to the World Bank nor to any of its staff or
members.
1 See Rodrik , Subramanian and Trebbi (2002) for a recent survey of the literature on studies of
cross-country growth performance.
2 Rodrik et al (2002), pp 20-21.
3 Rodrik (2003), p.25
2
The standard institutional solution is to introduce an independent regulatory agency,
operating within a clearly defined legal framework4. The regulatory agency is
intended to provide the "high quality institution" which permits and fosters sustained
growth in capacity and efficiency in the utility service industries particularly the
network elements. Hence, whether country X has a high or a low quality institution is
determined primarily by the quality of governance of the regulatory agency
(conditional on the governance quality for the country as a whole). As with the
aggregate economy, developing countries with high quality regulatory agencies (as
measured by their regulatory governance) should attract more investment on a
sustained basis into their utility service industries and at a lower cost of capital, as
well as having higher efficiency levels and growth rates in the regulated utilties.
We would expect this outcome to arise because regulatory agencies with better
governance should (a) make fewer mistakes and (b) have their mistakes identified
and rectified better and more quickly so that (c) good regulatory practice is more
readily established and maintained. It may well be possible to obtain a major short-
to-medium term increase in investment without an effective regulatory framework,
but the considerations outlined above suggest that this will not be sustained long-term.
The collapse of the Asian IPP boom of the early 1990s and the difficulties with many
of the Latin American infrastructure reforms and concession contracts in the late
1990s provide some evidence to support this conjecture.
The perspective outlined above is at the heart of the recent literature on regulatory
governance for utility service industries, particularly the literature that focuses on
developing and transition economies. This perspective is set out in Levy and Spiller
(1994) which draws explicitly on North (1990) as well as in a number of
subsequent papers5. However, until recently, there has been very little systematic
empirical testing of the hypothesis that better regulatory governance (a) reduces
unserved demand by encouraging investment or (b) increases efficiency. There have
been many case studies and these can be very illuminating but do not allow reliable
generalizations but, until the last 2-3 years, little formal econometric or other
statistical testing.
This is now changing. More developing country utility regulators have been in place
for 5 years or more and data are now becoming available on them that can be related
to industry outcomes on a comparable basis, most obviously for telecoms. Hence,
there have been a number of studies of the impact of a regulatory agency on capacity
growth and efficiency in telecoms. All the major recent studies show that having a
4 An independent regulatory agency is not the only way of providing the necessary institutional
support either in theory or in practice. In addition, an independent regulator may be combined
with a high or a low degree of reliance on contracts and courts.
There is a major issue of whether or not low income developing countries have the human and
other resources to sustain independent regulatory agencies, particularly regulatory agencies
with a significant degree of discretion. Nevertheless, an independent regulatory agency has
become the standard solution to the private investment problem for utilities in the same way as
an independent central bank has become the standard solution to handle commitment and time
inconsistency problems in monetary policy. See Section 2 below as well as the literature
discussed in Stern and Cubbin (2003).
5 See, inter alia, Smith (1997), Stern and Holder (1999), Noll (2001).
3
regulatory agency is significantly associated, either directly or indirectly, with higher
mainline capacity per capita and higher labor productivity.
In this paper, we carry out a similar exercise for electricity supply industries in
developing countries. Specifically, we provide an econometric analysis of the
relationship between the quality of regulatory governance and (a) the level of
generation capacity per capita and (b) some efficiency measures for a sample of 28
Latin American, Caribbean, Asian and African countries over the period 1980-2001.
The plan of the paper is as follows. In Section 2, we discuss the underlying economic
issues and the main institutional design considerations. This includes a summary
review of recent relevant literature and its relevance for our analysis. In Section 3, we
set out our modelling approach, including the modelling objectives, our econometric
approach, data issues and potential econometric concerns. Section 4 presents
descriptive statistics. Section 5 presents the estimation results. Section 6 discusses
the results and their implications and provides some short concluding comments.
4
2. Underlying Economic Issues, Institutional Design and Implications
for Empirical Analysis
The main issue on which we focus is the inability of governments to make credible
and binding commitments about utility pricing to sustain private investment while
retaining decision-making powers over these issues.
The discussion of utility service regulation concentrates on commercialized utilities
facing genuine budget constraints, particularly where private investment and/or
private finance is important. The focus of the discussion (and of our empirical work)
is on regulatory governance (e.g. autonomy, accountability, etc) rather than on
regulatory content (e.g. methods of price, investment and related aspects of
regulation)6.
2.1 Time Inconsistency Problems and Utility Service Industry Investment
The underlying economic issue for utility regulation as for monetary policy is that
governments, particularly at certain times, have a strong incentive to behave in a
shortsighted and populist manner that reduces welfare summed over a medium to
long-term period. Hence, both in general but particularly at times of pressure, they
place a very high weight on retail electricity prices over the next year relative to the
medium to long term. In consequence, in the utilities industry context, authoritarian
governments facing serious protests (and democratic governments facing difficulties
in imminent elections) have a strong incentive to hold down electricity prices below
economic cost even if this jeopardizes future investment and consumption.
For utility service industries, long-term contracts without a regulatory agency may be
sufficient in some circumstances to provide the necessary institutional surety (e.g. for
toll roads, water and sewage and similar). However, a regulatory agency may well
help improve the sustainability of contracts even in those industries7. For electricity,
although contracts may play a large part, they do not seem to be able to substitute for
regulation in providing a sound basis for private investment in generation, let alone in
transmission and distribution8.
In consequence, we assume in what follows that an independent regulatory agency is
the first-best method of ensuring that private investment in the electricity and similar
industries can be sustained and at the lowest possible cost of capital. Similarly, an
independent regulatory agency seems to be the best way of providing effective but
reasonable incentives for efficiency and high productivity and strong growth in
these. The question then is what are the appropriate measures of governance to
ensure the effectiveness of the regulatory agency in terms of these objectives.
The answers to these are typically given as a combination of:
6 We looked, in passing, at methods of price/profits regulation in our empirical work but this
issue was a subsidiary concern for this paper. See Section 4 for the results.
7 See Guasch, J.L., Laffont, J.J. & Straub, S., (2003) for a discussion of renegotiation of water
and transport concession contracts in Latin America.
8 See Stern (2003) for a discussion of these issues in the context of the development of the UK
electricity industry pre-1940.
5
(i) some formal legal requirements both (a) to underpin the regulatory
agency and (b) to set out the powers and duties of the regulatory
agency; and
(ii) regulatory processes that promote consistent and reliable decision-
making.9
In the empirical work that we discuss below, our measures of regulatory governance
will, perforce, be limited to four indicators of the formal legal aspects of the
regulatory agencies10.
2.2 Output Measures for Utility Regulatory Agencies
For utility service industries, there is a major issue in defining appropriate output
measures for utility regulation. For all countries, rich and poor, a relevant output is
higher levels of (and faster growing) technical and efficiency as well as quality of
service. However, whereas virtually all developing countries need significant
increases in capacity to meet demand at least in electricity supply, the same is by no
means always true for rich countries. One of the main drivers of the liberalization
plus privatization plus independent regulation OECD electricity reform model has
been the desire to reduce unnecessarily high capacity reserve margins as well as to
reduce investment costs.
This issue is important since significantly higher investment (and private investment)
was the single most important reason cited over the last 15 years by the World Bank
and similar policy institutions for the promotion of independent regulatory agencies in
electricity and similar utility service industries11. This view goes back to the
underlying time-inconsistency problem and the question of how, given limited tax
resources, developing countries can increase capacity and reduced unserved demand
particularly for countries with poor reputations as regards their treatment of private
investment. Hence, an independent regulatory agency has been advocated as the way
in which private investors can be assured that they will be able to earn a reasonable
rate of return.
In consequence, on this hypothesis, it is to be expected that sizeable increases in
private investment flows (domestic and foreign) will arise in developing country
electricity and similar industries following the establishment of an independent
regulatory agency. It is, however, worth noting that the speed at which the regulatory
credibility is established is very unclear. It is likely to take some time, so that one
might well expect lags of some years between the establishment of the new regulatory
agencies and any significant increase in investment.
9 See among others Levy and Spiller (1994), Smith (1997), Stern and Holder (1999), Noll
(2001). For a full discussion, see Stern and Cubbin (2003).
10 See Cubbin and Stern (2004) for a discussion of the implications of having only these as
measures of governance.
11 The World Bank's 1994 World Development Report "Infrastructure for Development" is a
good example. See Chapter 3.
6
The implications of the above are that, in estimating the impact of regulatory
governance variables on outcomes, we concentrate on:
(i) Electricity capacity levels in developing countries, excluding transition
economies as well as OECD countries; and
(ii) Efficiency measures in developing countries, insofar as they are
available.
We discuss the precise statistical measures of these in the next section. However, the
key point to note here is that we have chosen our sample so that it includes only
countries where there is good reason to believe that there are significant amounts of
unsatisfied electricity demand because of capacity constraints.
.
2.3 Results from Studies of Regulation on Developing Country
Telecommunications Outcomes
The approach outlined above is echoed in a rapidly growing literature on the impact
of regulation on telecom outcomes.
The main empirical papers in this area (e.g. Fink, Mattoo and Rathindran (2003),
Wallsten (2002) and Gutierrez (2003)) estimate the effects of regulation on
(i) mainline penetration rates (a standard measure of capacity) and
(b) efficiency (e.g. mainlines per employee).
They typically estimate panel data models (primarily fixed effects models) with one
or other of the outcome measures as the dependent variable, and include regulatory
variables as independent variables along with competition and privatisation variables,
as well as standard control variables. We follow this approach in estimating the
impact of regulation on electricity industry outcomes.
The standard model estimated in these papers (e.g. by Gutierrez) is
Yit Xit + Dit + i i =1, ..., N; j = 1, ..., T
= + it, (Equ 1)
where X is a vector of exogenous variables,
D is a vector of dummy variables,
iis a country specific fixed effect and
it, is an error model.
The X vector includes both regulatory variables and standard control variables
The approach of Gutierrez (2003) is particularly relevant to this paper. He constructs
a regulatory governance index for his sample of 22 Latin American and Caribbean
countries. This 7-element index (derived from the Stern-Holder typology) is
7
calculated from examination of each country's telecom laws and changes in the laws.
In our model for electricity outcomes, we adopt a similar approach and use a
`snapshot' 4-element index for 2000. (See Section 4 below for further details of our
index and the data.)
Gutierrez (2003) finds statistically and positive direct effects of his regulatory index
both on tele-density and on efficiency. This result occurs both in static and dynamic
models and after testing for the endogeneity of regulation. The estimated effect of a
1-point increase in the index on mainlines per 100 inhabitants varies somewhat
depending on the precise model specification but is, in general, of the order of 20%.
The Gutierrez study and its estimates provide a useful benchmark for our modelling
of the effects of regulation on developing country electricity industry outcomes12.
2.4. Results from Studies of Regulation on Developing Country Electricity
Industry Outcomes
As yet, there are only a very few and very preliminary empirical studies e.g. Zhang,
Kirkpatrick and Parker (2002) and a part of Pargal (2003). For data availability
reasons, the capacity variable for these studies is generation capacity only. This is
measured in physical units (ie in Gigawatts) Data on this is available on an annual
basis from the US Department of Energy's EIA website for almost all countries from
1980. Unfortunately, there is nothing similar available for capacity in transmission or
distribution13.
These papers find only weak effects of regulation, if any, and there are major
problems in disentangling the effects of regulation from those of liberalisation.
However, the studies are much more preliminary than those for telecoms, particularly
in data terms.
In this paper, we have had access to much better data on regulatory governance and its
variation across countries. However, our estimation of models for capacity, like those
of Zhang et al and Pargal, is also limited to generation capacity.
12 See Stern and Cubbin op cit, p. 38-43 for further details of these studies
13 Pargal uses the Calderon Serven infrastructure investment data set for 9 Latin American
countries 1980-98. This divides electricity investment into public and private but appears,
again, only to cover generation. See Calderon and Serven (2002) for a description of these
data.
8
3. Model Specification and Modelling Issues
Our modelling work is primarily concerned with whether better regulatory
governance in developing countries:
(i) increases rated generation capacity per capita; and
(ii) increases efficiency e.g. by increasing capacity utilisation in generation
and/or reducing transmission and distribution losses.
3.1 Underlying Economic Rationale
On capacity, we start from the basis that developing countries have serious capacity
constraints which lead to significant unserved demand arising, among other reasons.
from many years of low levels of investment. In developing countries, it has typically
been the case that electricity supplies were inadequate and intermittent. Supply was
insufficient to cope with the level of demand as a result of a variety of interconnected
factors.
Rectifying the issue of inadequate levels of capacity and investment has been a major
policy objective and a justification for electricity sector reform shared by developing
country governments and development agencies, national and international, including
the World Bank and the international regional banks.
The World Bank and others have argued that the establishment of good regulatory
governance (e.g. via the development of well-founded independent regulatory
agencies) has been a key element in their reform strategy over the last 15 or more
years. Hence, estimating whether regulatory agencies have significant impacts on
electricity capacity levels over time is important for the effectiveness of the policy.
This also provides a test of the theoretical case for the importance of time
inconsistency arguments as a useful framework for considering investment in the
electricity industry.
Of course, inadequate supply levels are not due just to inadequate investment. In
many developing countries, rated capacity has been much higher than available
capacity. However, the same factors (e.g. revenue shortages and inadequate returns)
also lead to low levels of maintenance. This is a major reason for expecting that
improvements in regulatory governance will increase efficiency and raise capacity
utilisation rates.
3.1.1 Regulation and Capacity Levels
The effect of electricity reform and the introduction of explicit regulation is to focus
the policy of the electricity industry on providing sufficient supplies.
In some cases, this has been done by harnessing the forces of private ownership
and/or competition. In others, it has to provide a workable financial framework
within which the electricity industry could develop by loosening the ties with
9
government for example, by enacting an electricity law giving various powers and
duties to a Ministry regulator thereby requiring publicly owned electricity companies
to operate in a more commercial way which would, among other things, allow state
owned electricity companies to borrow from banks or debt markets on standard,
commercial terms.
Investment is encouraged once effective regulation is available to support a workable
financial framework. If the electricity industry is in private ownership the owners
have the prospect of earning a reasonable return on their investment; if publicly
owned, the industry can become independent of tax revenue or continually increasing
loans. In addition, the existence of an effective regulatory framework can also
encourage the growth of private investment and/or private finance within state
systems, as has been happening in recent years in India and China.
These considerations suggest that the presence of an effective regulatory framework
should, in general, lead to increased investment in the electricity sector, including the
balanced development of generation, transmission, and distribution ceteris paribus.
Unfortunately, comparable time-series capacity data across countries only exists for
generation and it is on this aspect that the present study focuses.
In an unconstrained market economy, per capita generation capacity will adjust to the
level of demand, which will depend upon the level of per capita income, the price of
electricity, and environmental factors such as climate. The price of electricity will be
determined in part by the efficiency of the sector. The latter may depend upon
regulatory factors, but also availability of energy sources such as hydro, gas, oil, and
coal. (This is most evident in cross-U.S. comparisons of prices.) However, many
developing countries with a traditional, vertically integrated and state-owned
electricity sector will be constrained not so much by market demand but by the
availability of continuing subsidy.
The capacity constraints arise because of either inadequate government revenues for
electricity investment or subsidy payments and/or insufficient revenue flow to support
viable private investment. A simple diagrammatic version of such a model is shown
in Figure 1 below
10
Figure 1 Chronic Supply Constrained Electricity Shortage (loss making
public enterprise)
Supply without subsidy
Price
Subsidy Supply with subsidy
Pre-
reform Demand
price
Private Unsatisfied demand
expenditure
Quantity
Capacity
In this model, the level of capacity in the unreformed industry depends on the sum of
private and public expenditure on investment in electricity which, in turn, will be
determined primarily by the level of national income per capita. It is also well-
established that the demand for electricity (and hence for electricity capacity) has an
elasticity close to 1 with respect to GDP. Hence, we would expect equilibrium
electricity demand and supply for electricity to be related to GDP growth.
For both these reasons, we include per capita GDP in our model, with an expected
long-run elasticity not significantly different from 1. We also consider other control
variables that have been found as statistically significant in previous studies eg the
share of industry in value added, country debt levels and country economy-wide
governance indicators.
The effect of an effective regulatory framework should be to reduce the constraint on
the operation of the market, increasing supply and moving the outcome closer to the
market equilibrium. The better the governance of the regulator, the greater the
expected increase in capacity and increase in electricity supply.
We measure the quality of governance primarily by an index of regulatory
governance which has 4 elements:
(i) Whether the country has an electricity or (energy) regulatory law;
(ii) Whether the country has an autonomous or a Ministry regulator for
electricity;
11
(iii) Whether the country's electricity regulator is funded from licence fees
(or equivalent) or out of the government budget; and
(iv) Whether the staff in the electricity regulator can be paid as appropriate
given skill needs and labour markets or whether staff have to be paid
on civil service pay scales.
These are all measured by 0/1 dummies. The highest governance ranking (a score of
4 on the index) is represented by having enacted an electricity regulatory law, plus an
autonomous regulator, plus funding from licence fees and the staff not being confined
to civil service pay scales. The dating of the switch from 0 to 1 on the appropriate
variables (subsequently maintained at a constant level) is derived from the date of
enactment of the law (except for cases where other information was available to
provide a known, superior alternative). Hence, we can investigate the effect of age of
the regulatory agency as well as its existence.
Given the economic arguments set out above and in Section 2, we would expect the
coefficients on the index and on each of its components to be positive. We might also
expect the effect of regulation to increase with the age of the regulator, particularly
for the first few years.
In terms of the typology in Section 2.2, the regulatory variables in our index are all
measures of formal attributes of regulation. Unfortunately, no comparable data is
currently available on the informal, practical qualities of electricity regulation and the
necessary omission of data on these characteristics must be borne in mind when
considering the results, including potential biases to the estimates and to estimated
standard errors. In addition, unlike Gutierrez (2003), we have no time dimension on
changes in formal governance attributes subsequent to the enactment of the
electricity/energy regulatory law.
These considerations suggest a capacity model of the following form:
Log(ELCAPPC)it = (a0 + ai) + a1 log(GDPPC)it + a2 Industryit + a3 Debtit + a4
RegIndexit + a5 Xit + vit
(2)
Where Log ELCAPPC is the log of per capita electricity generation capacity
in Gigawatts;
a0 is a constant term;
ai is a time-invariant country specific fixed effect
GDPPC is real per capita national income in $US 199514;
Industry is the log of industry value added as a percentage of GDP;
14 Hence, GDP is on an exchange rate rather than a PPP basis.
12
Debt is the share of government debt service as a percentage of gross
national income;
RegIndex is our regulatory governance index (or individual
components of it);
X is a vector of other potentially relevant variables (e.g. rule of law
and corruption measures, age of regulatory agency, method of price
regulation, etc); and
uit is an error term
In all cases, the variables exist for i = 1, ..., I countries over t = 1, ...., T time periods.
The regulatory index takes the value of 0,1, 2, 3 or 4 where zero is ascribed to
countries with a Ministry regulator, no electricity regulatory law, government budget
funding and civil service pay scales.
The X vector for this equation might well include domestic fuel/hydro source
availability and a variety of other country specific economic and/or institutional
variables. However, these variables can be expected to stay fairly constant over the
period of estimation - as do country governance rankings.
Following the literature on the impact of telecom reform in developing countries, we
also explore the role of (i) privatisation and (ii) competition on generation capacity
growth. We investigate both direct and indirect effects (e.g. interactions between
these variables and the regulatory index).
Although we start by estimating an OLS version of the model above, most of the
results reported in Section 4 are for a fixed effects model15. Differencing the equation
above eliminates the constant term and the country-specific fixed effects. If fixed
effects are significant, the error term the equation above will not be normally
distributed with zero mean when estimated by OLS. (See Section 3.2.2 below for a
fuller discussion of econometric issues.)
The fixed effects are likely to include country variables with little or no time variation
over the sample period. This affects not just fuel source availability, but also many
constant or slowly changing institutional variables. The estimated fixed effects may
therefore capture key aspects of the rule of law and corruption as country rankings on
these indicators tend to be relatively stable over 10-20 year periods.
The equation above is a static representation of the model, which provide evidence on
long-run equilibrium effects. We also consider:
(i) dynamic variants e.g. incorporating a lagged dependent variable as in
Gutierrez (2003); and
15 See Section 3.3.2 .1 below for a fuller discussion of heterogeneity issues.
13
(ii) error correction models which allow for more explicit examination of
long-run equilibrium effects as opposed to short-run adjustment effects
(iii) IV (instrumental variable)models that control for the potential
endogeneity of our regulatory governance index.
3.1.2 Regulation and Efficiency
As regards efficiency, we concentrated on the impact of regulation on two readily
measurable characteristics of electricity supply industries for which comparable time-
series data existed:
(i) Utilisation of generation capacity; and
(ii) Technical losses in transmission and distribution.
The first was measured as:
(Total Annual Generation in TWh)/(Generation Capacity in TW/365 * 24).
This measure provides a good proxy for the availability of generation plant. Many
developing countries have rated capacity levels that are considerably higher than
available capacity and higher utilisation rates should closely reflect improvements in
availability e.g. from the impact of better regulatory governance on maintenance
expenditure.
Technical losses were measured as transmission and distribution losses as a
percentage of total generation.
In both cases, we deliberately estimated a simple and parsimonious fixed effects
model with the regulatory index as the main explanatory variable and real per capita
GDP as a control variable. This was, not least, because there was no obvious well-
defined theoretical model on which to base a more sophisticated approach.
We would very much have liked to estimate models for quality of supply (e.g. supply
interruptions, coverage of system) and also for commercial losses. Empirical studies
of electricity reform have shown that a major impact has been to improve quality and
to reduce non-technical losses, particularly at the distribution level16. Unfortunately,
no data currently exists for these variables that would allow the estimation of cross-
country panel data models to test for the impact of improved regulation on quality.
16 See, for instance, Bacon and Besant-Jones (2001)
14
3.2 Modelling Approach
The purpose of the investigation was:
a) to undertake a preliminary analysis for the electricity industry of the effect of
independent regulators and aspects of their governance on improving the
overall performance of the sector; and
b) to identify priority areas where enhanced data was required to allow a better
analysis of these effects.
Under a), the key questions we have tried to answer are:
i) Does the existence of an independent regulator appear to have any effects on
measurable aspects of electricity industry performance (generation capacity,
utilisation and technical losses)?
ii) If so, how big is the effect?
iii) By how much is the size of any effect influenced by measurable aspects of the
governance of the regulatory institutions?
iv) What effects do private ownership and competition have in enhancing the
aspects of performance we have measured, independently and in combination
with regulation?
Unfortunately, data limitations prevented us from seriously addressing the impacts of
privatisation and competition17.
On b), the quality and precision of the answers to these questions should help us to
identify priorities for improvements on currently measured data. Consideration of the
potential impact on the results of omitted variables and the resulting potential biases
should help identify priorities for collecting data on variables for which data is not
currently available.
3.2.1. Data
We have collected data on 28 developing countries over a 21 year period (1980-
2001). Of the 28 countries in the sample, 15 were in Latin America, 6 in the
Caribbean, 4 were in Asia and 5 were in Africa. The list of countries includes large
countries (e.g. Brazil and India), small countries (e.g. Jamaica); middle income
countries (e.g. Chile and Mexico) and poor countries (e.g. Ethiopia and Sudan). The
full list of countries for which we have data is in the Appendix.
Although much of the regulatory activity took place in the last half of the data set, the
earlier period is important in effectively establishing benchmark levels of the
dependent variables, and also in reducing some of the biases that can potentially arise
17 See Section 5.1.2.2 below.
15
in the use of short panels. In fact, 20.7% of the total number of country-sample years
were years with an autonomous regulator and 31% with an electricity or energy
regulatory law.
The key data sources used are:
· US Energy Information Agency for data on generation capacity by
country (GW) 1980-2001 (Noted that the EIA series does not
distinguish between publicly and privately owned generation capacity.)
· World Bank Development Indicators - for Per capita GDP in $US1995;
electric power transmission and distribution losses and other control
variables
· The Preetum Domah 2001 survey of electricity regulators for data on
electricity regulatory governance, privatisation and competition
(supplemented by the authors' own research).18
The Domah survey data (covering 50 developed, transition and developing countries)
are the best data currently available to estimate the impact electricity regulators, not
least because it allows the dating of regulatory reforms, primarily because it records
the year in which key regulatory legislation was enacted.
The Domah data set is very suitable for a preliminary investigation of the impact of
regulation but is far from ideal. In particular, it suffers from the following:
1) The data on electricity market structure is relatively weak and the data on
privatisation very limited;
2) There is no data on the informal, practical aspects of regulation (e.g. security
of tenure of regulatory agency heads or commissioners, etc);
3) The data on regulatory governance, competition and privatisation has no time
dimension beyond a simple 0/1 dichotomy set at the year in which key
regulatory legislation was enacted;
4) The data on the formal aspects of regulation only allows for a 4-element index
rather than a larger index. These data weaknesses should be born in mind
when considering the econometric results.
3.2.2. Econometric Issues
Panel data generally allow major opportunities for carrying out investigations that are
not possible with single-year cross sections or single-country time series, but these
give rise to a number of issues which need taking into account for estimation. In our
case, with data on 28 countries for 21 years, we have a large and long panel. Because
of missing observations, it is an unbalanced panel.
18 See Domah, Pollitt, and Stern (2002) for full details. We are very grateful to Preetum Domah
for permission to use the information from his survey in this paper.
16
The use of panel data may have many benefits but their use also raises a number of
potential econometric problems including:
1) Coefficient heterogeneity across countries.
We have strong prior views that countries will differ consistently in their intercepts
according to persisting, largely time invariant local factors. For this reason our
maintained hypothesis is that a fixed effects model is more appropriate than a random
effects approach. In addition, the fixed effects static model avoids the potential biases
which could arise in the random effects model owing to correlation between the
included exogenous variables and omitted country attributes.
2) Dynamic structure
Static models, which assume that all adjustment to disequilibrium occur within the
period defined by observation frequency may be inappropriate. In particular,
investment in electricity is not usually completed in a year so we would expect that
scope for some adjustment process would need to be incorporated into our model.
Such processes can be modelled generally by a combination of lags on the dependent
variable (autoregressive) and on the explanatory variables (moving average).
However, the presence of a lagged dependent variable in a fixed-effects model can
result in a biased estimates for the lagged dependent variable coefficient. The size of
the bias will depend on the number of time series, N, the length of the time series, T,
and the influence of other exogenous variables in the determination of the dependent
variable.19 The problem is mainly significant in short panels. For T=21 we have
estimated the asymptotic bias (as N increases) to be of the order of 3%. This is an
upper limit given the presence of other major influences on the dependent variable.
There is also the potential problem of spurious correlation eg if both electricity
capacity and GDP per capita were both strongly trended across our countries. In fact,
they are not in our data set, but we have considered carefully how the dynamics
should be modelled and we report a selection of the key results.
3) Endogeneity and Causality
There has been much discussion of the need to take account of the endogeneity of
regulatory agencies. This has been a major theme in the ICB literature where the
introduction of an ICB (particularly the early introduction) may be interpreted as a
signal of strong commitment to anti-inflation policies. Similarly, the early
introduction of an autonomous regulator may also be a signal of a strong commitment
to commercialisation and the enforcement of property rights20.
19 See Hsiao (1986)
20 See Gual and Trillas (2002)
17
Given the relatively time-invariant rankings of countries' governance (including rule
of law, corruption, etc), it is not clear that there exists a particularly serious
endogeneity problem to the extent that there is an overall issue, it should be well-
handled by country specific fixed effects. In addition, as noted by Fink et al (2002)
and others, it is also extremely difficult to find appropriate instruments for regulatory
governance variables. Nevertheless, we do in Section 5.2.4.1 explicitly consider
endogeneity and report IV (instrumental variable) estimates that attempt to control for
it within a fixed effects modelling framework.
Discussion of endogeneity issues in institutional models frequently reflects concerns
over causality rather than endogeneity per se. With a long panel of 21 years, fixed
effects should adequately control for country-specific institutional quality variations
so that any bias in the estimates of regulatory governance impact from that source
should be small. However, even if that were so, there remains the question of whether
the estimated coefficients on regulatory governance can be taken as estimates of what
would happen if countries were to improve or reduce the quality of governance of
their existing regulatory institutions - or, more importantly, regulatory institutions of
given quality were to be introduced into a country currently without such institutions.
These, particularly the latter are the key policy questions.
We discuss both issues in Section 5.2.421.
21 We are grateful to Richard Gilbert and Jean-Michel Glachant for helpful discussions on these
issues.
18
4. Descriptive Statistics
In this section, we report some key descriptive statistics from the Domah survey.
4.1 Countries with Autonomous Electricity Regulators
Table 1 shows that by 1998 just under half the countries in our sample had an
autonomous electricity regulator22 mainly in Latin America. However, in the
following three years 3 African and 1 Caribbean country joined the set. Asia provides
an exception to the spread of autonomous regulators with only 1 country (Philippines)
having an autonomous regulator before 2001.
By 2001, a majority of countries had regulators classified (at least in legal terms) as
autonomous.
Table 1: The Trend towards Autonomous Regulators (by Continent)
1998 2001
Total Ministry Autonomous Ministry Autonomous
Africa 5 5 0 3 2
Asia 4 3 1 3 1
Carrib 6 3 3 2 4
Latin America 13 4 9 4 9
Total 28 15 13 11 17
Source: Domah 2001 survey, supplemented and updated by authors
4.2 Countries with Electricity Regulatory Laws
Even where there was no autonomous regulator, laws for the reform of the ESI
Including regulatory reform were being passed. Table 2 shows the regional
distribution of electricity reform laws for those states without autonomous regulation.
According to the Domah data, all the countries with autonomous regulators had
enacted an electricity regulatory law. By the end of our sample period only two
countries in the sample (Barbados and Indonesia) did not have any electricity
regulatory law in place.
These laws sometimes provided for IPPs or other elements of market reform, for
commercialisation and sometimes for unbundling and competition in generation and
supply23. If the laws covered regulation, they typically specified the powers and
duties of the Ministry (or designated Ministry agency/department) in carrying out
regulatory functions.
22 The Domah questionnaire used the term "autonomous" rather than "independent", not least
because it is more neutral. We treat the two terms as synonymous.
23 However, the actual introduction of competition and/or privatisation took place at some later
date, typically with several events at different times. This is why we cannot within this data
set obtain good indicators for privatisation and competition.
19
Table 2: Non-Autonomous Regulators: Existence of Law
1998 2001
Total Law No Law Law No Law
Africa 5 0 5 3 0
Asia 4 1 2 2 1
Carrib 6 1 2 1 1
Latin America 13 3 1 4 0
Total 28 5 9 10 2
Source: Domah 2001 survey, supplemented and updated by authors
4.3 Age Distribution of Autonomous Regulatory Agencies
Figure 2 shows the age distribution of energy regulatory agencies.
Figure 2: Age Distribution of Autonomous Electricity Regulators
Age distribution in 2001
10
9
8
7
6
5
Frequency 4
3
2
1
0
0 1-3 4-6 7-9 10-12 13+
Age of law
Source: Domah 2001 survey, supplemented and updated by authors
Figure 2 shows clearly how many of the DTE regulatory agencies in our sample were
very recently established. 8 (47%) were under 3-years old in 2001, including all the
20
African electricity regulators. The median age was just under 5 years. However, 5
(29%) were 10 years old or more and accounts for 42% of the total number of sample
years with an autonomous regulator. The over 13 year-old group of autonomous
electricity regulatory agencies comprises Costa Rica, Philippines and Trinidad and
Tobago.
4.4 Ministry or Autonomous Regulator and Per Capita GDP
Figure 3 shows, very interestingly, that at least within this sample - there is little
relationship between the existence of an autonomous regulator and per capita GDP.
Both autonomous and Ministry regulators are scattered through the income range.
The mean income for countries with an autonomous electricity regulator was $3,500.
For those with a ministry regulator it was $3,300. The difference was not significant.
However, low income countries with an autonomous regulator have younger
regulators e.g.. the two Sub-Saharan African regulators established since 1998 (Kenya
and Uganda).
Figure 3: Type of Regulator and Per Capita Income (in Real $ 1995)
Independent regulators and National Income per
capita
Independent
regulator
Ministry
18000.00
16000.00
14000.00
12000.00
capita 10000.00
per 8000.00
6000.00
GDP 4000.00
2000.00
0.00
1 4 7 10 13 16 19 22 25 28
Rank by GDP
Source: Domah 2001 survey, supplemented and updated by authors
4.1.5 Generation Capacity and per Capita Income
Figures 4 and 5 plot generation capacity by real GDP for our 28 country panel at the
start and end of the period. The plots show an upward sloping but by no means either
uniform or linear relationship.
21
Figure 4: Generation Capacity and Income 1980
Generation capacity and Income 1980
0.8
0.7
0.6
0.5
generation 0.4
capacity 0.3
0.2
Percapita 0.1
0
0 2000 4000 6000 8000 10000 12000
Per capita GDP
Source: US Energy Information Agency and World Bank Development Indicators
The outlier in the bottom right is Ecuador- as it was again in 2001. In 2001, only 55%
of the population had access to mains electricity, even though the country is Latin
America's largest oil exporter.
Figure 5: Generation Capacity and Income in 2000
Generation Capacity and Income 2000
1.6
1.4
1.2
1
Generation 0.8
Capacity0.6
Capita 0.4
0.2
Per 0
0.00 5,000.00 10,000.00 15,000.00 20,000.00
Per capita GDP
Source: US Energy Information Agency and World Bank Development Indicators
The country at the top left of Figure 5 is Paraguay. It is a major exporter of hydro-
power, and meets 25% and 40% respectively of Brazil and Argentina's electricity
22
demands.24 Interestingly, Ecuador has a (relatively young) autonomous regulator but
Paraguay has a Ministry regulator.
For the pooled sample, the elasticity of generation capacity with respect to real per
capita GDP was 0.89 (with a standard error of 0.02). The cross-section elasticities for
selected years were as follows:
Table 3: Generation Capacity Real Income Elasticities
Year Estimated Generation Capacity/ Real GDP
Elasticities
(Standard error in parentheses)
1980 0.78
(0.09)
1990 0.90
(0.11)
2000 0.94
(0.11)
Table 3 shows a trend towards the simple elasticity rising towards unity i.e. faster
growth in generation capacity than in real income over the period. It remains to be
seen whether this is in any way related to the spread of electricity regulatory reform.
4.1.6 Generation Capacity Utilisation 1980 and 2000
Beginning and end-of-period generation capacity utilisation rates are shown in Figure
6 below. In general, there has been a noticeable increase in capacity utilisation, but
there are country exceptions (e.g. Colombia).
Figure 6: Utilisation of Generation Capacity 1980 and 2000
Utilisation 1980 and 2000
16
14 Utilisation 2000
12 Utilisation 1980
10
8
equency 6
Fr
4
2
0
<0.1 0.1- 0.2- 0.3- 0.4- 0.5- 0.6- 0.7- 0.8- 0.9-
0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Utilisation rate
24See IEA country analysis brief :Paraguay at www.eia.doe.gov
23
4.5 Correlation between Indicators of Regulatory Governance
As discussed above, our regulatory index includes 4 indicators. These are classified
positively for: (i) the enactment of an electricity regulatory law; (ii) the existence of
an independent/autonomous regulator; (iii) funding from licence fees (or equivalent)
and (v) staff salaries not necessarily confined to civil service pay scales.
Although the majority of our results are based on the index, we also try to estimate
their separate effects. However, the degree to which we are able to do so depends on
the levels to which they are correlated with one another. Not surprisingly, they are
highly inter-correlated as shown in Table 4 below.
Table 4: Correlation Matrix between Regulatory Governance Variables
ElAct Funding Orgtype Cserv
ElAct 1
Funding 0.848968 1
Orgtype 0.783066 0.703489 1
Cserv 0.783066 0.551221 0.442631 1
The correlations are highest within the law/funding/autonomy grouping.
For a visual depiction of the generation capacity and GDP data, see the Data
Appendix in Cubbin and Stern (2004)
24
5 Econometric Results
In what follows, we report various results. In Table 5 below, we report the core
results for our static model of per capita generation capacity. Tables 6 reports some
results from simple LDV dynamic models of generation capacity. Table 7 reports the
results from more sophisticated error correction dynamic models and Table 8 reports
some results for generation capacity utilisation and technical losses.
5.1 Econometric Results for Models of Generation Capacity and
Investment
We start by reporting the results of an OLS equation as a baseline. All subsequent
equations are modelled using a fixed effects estimator.
Given the nature of the underlying model, we would expect a fixed effects model to
be more appropriate than a random effects model. For some of the equations, we
tested this assumption using the Hausman test and the random effects model was
consistently rejected in favour of a fixed effects model.
5.1.1 Basic Static Generation Capacity Model Results
The key results are shown in Table 5:
· The fixed effects model clearly dominates the OLS model as shown in the
standard error of estimate for the regressions.
· The estimated coefficient on the regulatory index is significantly different
from zero at the 1% level in Equations 1 and 2.
· The implications of Equation 2 (our basic fixed effects model) is that, in the
long-run, each unit increase in the regulatory governance index is associated
with 4.3% higher per capita generation capacity. Hence, a country with best
regulatory governance practice and an index score of 4 could expect to have
17.2% higher generation capacity per capita in the long run.
· The impact of regulation clearly increases with age of regulator. Equation 3
suggests a long-run effect of regulators (Ministry or autonomous) aged over 3
years of 35% on per capita generation capacity. Equation 4, which assumes a
quadratic effect of age of regulator, implies that the impact of having a
regulator peaks at 15 years.25
· The coefficient estimates for log(real GDP) are 0.7 - 0.8, with t-values of 8 or
more.
· Neither the debt nor the industry value added variables were significant at the
5% level except in the OLS equation a result consistently replicated.
25 The implicit decline in effectiveness after 15 years is not well-founded as only one of our
regulatory agencies (Costa Rica) has a regulator in place for more than 15 years
25
· The equations all have very low Durbin-Watson statistics which suggest that t-
values may be upward biased. (We explore this further in Section 5.2.2,
where we report estimates from error correction models.)
Table 5: Static Models for Generation Capacity
Dependent Variable = Age-of -
Log(Electricity Generation Basic FE regulator Quadratic in age
capacity per capita) OLS model model dummies of regulator
Explanatory variables 1 2 3 4
Constant -8.286
(-52.162)
Real GDP per capita (log) 0.772 0.805 0.697 0.699
(31.071) (9.970) (8.343) (8.522)
Debt payments as a 4.14E-12 2.49E-12 -4.65E-13 1.07E-12
proportion of national
income
(0.838) (0.556) (-0.104) (0.244)
Industry value added as 0.024 -0.002 0.000 0.001
proportion of GDP
(7.981) (-0.607) (-0.003) (0.232)
Index of regulatory 0.056 0.043 -0.026 -0.011
governance 0-4
(2.982) (3.444) (-1.067) (-0.638)
Regulator under 1 year 0.090
(1.465)
Regulator 1-3 years 0.187
(2.398)
Regulator aged over 3 0.353
years
(4.370)
Age of regulator 0.055
(4.132)
(Age of regulator)2 -0.002
(-2.635)
Fixed Fixed Fixed effects
Estimation method OLS effects effects
Adjusted R-squared 0.764 0.952 0.954 0.954
S.E. of regression 0.605 0.272 0.267 0.266
F-statistic 465.943 372.079 352.169 365.770
Durbin-Watson 0.043 0.161 0.163 0.153
No of observations 577 577 577
Note: t statistics in
parentheses
26
5.1.2 Variants on the Fixed Effects Generation Capacity Model
A large number of variants, static and dynamic, are reported in full in Cubbin and
Stern (2004). Here, we summarise the key results.
5.1.2.1 Static Model: Individual Governance Elements
The first set of variants was the estimation of the static model of per capita generation
capacity, as in Table 5 above, but including in 4 separate equations each of the
individual governance elements in our regulatory index. The main results were as
follows:
· The largest estimated regulatory effect of the index components was, perhaps
surprisingly, from having an electricity law (18% with a t-value of 5.1) rather
than from having an autonomous regulator (10% with a t-value of 2.3).
However, these must be interpreted in the light of the correlation between
them of 0.78.
· Licence funding of the regulator also had a positive estimated effect (13.5%
with a t-value of 3.4)
· The estimated sign on non-mandatory civil service pay scales was negative
(and significant at the 1% level) ie the opposite of that predicted by regulatory
governance theory.
The strong effect of a regulatory law and of age of regulator variables derived from
the date of the law appears to reflect a signalling and commitment effect from
having a legal framework which makes even Ministry regulators significantly more
accountable for how they carry out their functions. It may be that the effect of an
autonomous regulator would be higher in a sample where more autonomous
regulators had been in operation for more than 5 years.
Given the high degree of collinearity between the regulatory variables, we used
principal components to help better identify the effects of the individual governance
elements. We first computed the principal components of the four governance
element. We then included the first 2 principal components (accounting for over 90%
of the total index variance) in a static fixed effects regression of per capita generation
capacity with per capita GDP as the other explanatory variable. The estimated impact
of each of the 2 principal components was positive but only the coefficient estimate of
the first principal component was statistically significant at the 5% level, with a t-
value of 3.8.26
Interestingly, the loadings of the individual components in the first principal
component (accounting for 76% of the total index variance) were broadly similar.
Nevertheless, the loading from the electricity law element was the highest, providing
26 Including the first 2 principal components in a dynamic equation with a lagged dependent
variable produced very similar results.
27
some corroboration for the results and the associated conjectures above arising from
the separate equations for the individual governance elements.
5.1.2.2. Static Model: Privatisation and Competition
The equations estimated showed no statistically significant effect of either of these on
generation capacity. However, the privatisation dummy is deficient on dating and the
competition dummy is a weak proxy as well as poor on dating.
· The coefficient on our competition proxy variable was consistently negative
but not significantly different from zero.
· The coefficient on the privatisation variable was only significant at the 10%
level or better when interacted with the regulatory dummy and the regulatory
index was excluded27.
5.1.2.2. Static Model: Country Governance Indicators
We included as explanatory variables the Kaufmann indexes for (i) rule of law and (ii)
corruption by country for 1998. The key results were:
· The corruption index was never statistically significant in the fixed effect
regressions at the 5% level or better, either as a separate variable or when
interacted with the regulatory variables.
· The country rule of law index approached significance at the 5% level when
interacted with the regulatory index and an age of regulator was also included
as an explanatory variable.
· The Kaufman rule of law index was, however, highly significant in an OLS
equation and led to non-significance of the electricity regulatory variable.
The last result (and the pattern of residuals) is a major reason why we believe that the
estimated fixed effects capture most of the country-specific institutional differences.
We also found:
- No statistically significant correlation between the fixed effects and
the Kaufmann rule of law index; but
- A sizeable and statistically significant interaction term between our
regulatory index and the Kaufmann rule of law index in a random
effects specification (a coefficient of 0.07 with a t-value of 2.3).
27 Cubbin and Stern (2004), Table 7 and discussion for more details
28
These results provide interesting pointers that the estimated fixed effects capture
wider country-specific institutional quality issues but are clearly far from conclusive.
5.2 Dynamic Models for Generation Capacity
In this section we discuss the results
(a) For a simple dynamic models for per capita generation capacity, adding a
lagged dependent variable to the static, fixed effects model; and
(b) For more sophisticated error correction models.
Given the nature of the generation investment planning and construction process, we
would expect quite long lags, which will be picked up from the simple formulation in
(a). However, well-fitting simple LDV models may reflect spurious correlations
rather than a systematic relationship so the more sophisticated models were estimated
to test for spurious correlation as well as to improve our understanding of the
dynamics and the lags, including the build-up of regulatory reputation effects
5.2.1 Simple Dynamic Models of Generation Capacity
These models were estimated by adding a lagged dependent variable (LDV). The
estimated coefficient on the LDV was very high over 0.85 implying, as one would
expect, a slow rate of adjustment of generation capacity and with very high
estimated t-values, over 60.
The regression results are shown in Table 6 below. The key results are:
· The implicit long-run coefficient on the regulatory governance index was 6.1
per unit on the index, implying a long-run effect of 24% on per capita
generation capacity for a maximum score on the index as compared to 17% in
the static model. (However, the estimated coefficient was only significantly
different from zero at the 10% level.)
· The implicit long-run effect in Equation 13 on per capita generation capacity
from a regulatory agency (Ministry or autonomous) with at least 3 years of
existence is 26% as compared to 35% in the static model. The estimated
coefficient was significantly different from zero at the 5% level.
· The elasticity of per capita generation capacity wrt. real GDP was very close
to 1 (0.998 and 1.03).
29
Table 6: Simple Dynamic LDV Models for per Capita Generation Capacity
Log(Electricity Log(Electricity
Generation Generation
capacity per capacity per
Dependent Variable capita) capita)
Explanatory variables
Lagged dependent variable 0.885 0.879
(66.186) (66.094)
Real GDP per capita (log) 0.119 0.121
(4.558) (4.610)
Index of regulatory governance 0-4 0.007
(1.835)
Regulator aged 1-3 years 0.010
(0.893)
Regulator aged over 3 years 0.032
(2.247)
Implied long run multiplier 1/(1- ) 8.732 8.249
Estimation method Fixed effects Fixed effects
Adjusted R-squared 0.996 0.996
S.E. of regression 0.083 0.083
F-statistic 4366.527 4399.462
Durbin-Watson 1.850 1.834
No of observations 576 584
Note: t statistics in parentheses
In all the main generation capacity equations reported above, the R2 statistics are high
around 0.95 in the static fixed effects models and over 0.99 in the dynamic model.
The latter in particular raises questions as to whether, given the fixed effects, the
empirical results are dominated by the purely statistical relationship of one highly
trended variable (per capita generation capacity) with another (real per capita GDP).
In fact, neither of these series is dominated by an obvious trend28 but the issue
requires more formal investigation. See 5.2.2 below.
5.2.2 Error Correction Models and Lags
5.2.2.1 Time Trends and Lags
Our first test was to establish whether the coefficients on the regulatory governance
variables remained positive and significant when we (a) included a time-trend and (b)
lagged the index by 3 years.
The results were as follows:
28 See Data Annex in Cubbin and Stern (2004).
30
· The time trend was statistically significant at the 1% level in a static
formulation but negative and far from significant in an LDV model. Its
estimated value in the static model was only 1.7% p.a.
· The estimated coefficient on the lagged regulatory index was positive in both
models. In the LDV model, it was statistically significant at the 1% level and
the magnitude of the estimated long-run coefficient was very similar to that in
the LDV model without a time-trend. However, the magnitude of the
estimated regulatory coefficient in the static model was about half the
magnitude of that in Table 1, Equation 2 and only significant at the 10%
level29.
(Note that lagging the regulatory variable by 3 years implies that all regulators
established after 1997 are excluded from the sample.)
5.2.2.2 Error Correction Models
If we wish to be sure the fixed effects levels equations are not just spurious
regressions, we can check to see whether there appears to be a plausible adjustment
process.
The levels equation is: Yit = i + Git + Rit + it (3)
which can be estimated as: Yit = fi + bGit + c Rit + uit (4)
where Yit = log(electricity generation capacity per capita)
Git = log(GDP per capita)
Rit is a regulatory governance variable; and
fi is the fixed effect for country i
From this, we can calculate the implied the steady state, equilibrium, or long term
value, which can be written as:
Y*it = i + Git + Rit (5)
We now postulate a partial adjustment error correction mechanism under which the
actual value of capacity changes by a constant proportion of last year's deviation from
the long term value ie
Yit = Yit-Yit-1 = - (Yit-1Y*it-1) (6)
where (Yit-1 Y*it-1) is last year's deviation from equilibrium.
If we wish to estimate (6), we can take the residuals uit from the levels equation in (4)
and calculate the regression equation
29 See Cubbin & Stern op cit, Table 9 for more details.
31
Yit = - uit-1 + eit (7)
Alternatively, we could estimate
Yit = - (Yit-1 i - Git-1 - Rit-1) + it (8a)
= - Yit-1 + Git-1 + Ri-1t + it (8b)
More specifically, since we are particularly interested in the size and significance of
the regulatory variable, R, we can impose the estimate of from the long term levels
relationship (2) and estimate
Yit = (Yit-1 - bGit-1 ) + Rit-1 + it (9a)
From this we derive the following which we estimate, including fixed effects:
.
Yit = (Yit-1 - 0.78 Git-1 ) + Rit-1 + eit (9b)
We have alternative estimates of from equations (7) and (9b). These can be
compared. In addition, we have alternative estimates of : firstly, from the levels
equation (4); and, secondly, the differences equation 7(b) from which we can calculate
the implied value of as / from the estimated coefficients.
We tested for stationarity using the Pesharan-Shin W-statistic. Applying this test to
the differenced equation 7(b), with the regulatory index as our measure of Rit, the test
clearly rejects the presence of a unit root in the residuals with a t-statistic of 8.05.
Even in the corresponding levels equation, the Pesharan-Shin W-statistic does not
appear to suggest non-stationarity in the residuals, implying that our generation
capacity variable, GDP and our regulatory variables are co-integrated30.
The results are shown in Table 7 below. We report estimates for the levels equation
(2), the partial adjustment equation (5) and the differenced model (9b). We report
these (a) where the regulatory variable is the 4-element regulatory index and (b)
where it is a dummy for all regulators over 3 years old. Equations estimated with
alternative definitions of the regulatory variable yielded estimates with similar
positive magnitude and statistically significant at the 5% level or better.
30 Very similar results on the unit root test applied when we took alternative definitions of the
regulatory variables: regulator over 3 years old and a quadratic in the age of the regulator.
32
Table 7: Generation Capacity Error Correction Models
Log (Per Cap Log (Per Cap Log (Per Cap Log (Per Cap Log (Per Cap Log (Per Cap
Generation Generation Generation Generation Generation Generation
capacity) capacity) capacity) capacity) capacity) capacity)
Levels Differences Differences Levels Differences Differences
Dependent Variable 1 2 3 1 2 3
Explanatory variables
Real GDP per capita (log) 0.7913 0.7846
(10.787) (10.830)
Index of regulatory governance (0- 0.0490
4)
(4.454)
Index of regulatory governance (t-1) 0.0118
(3.230)
Lagged Residuals from 1 0.1195 0.1188
(8.941) (8.876)
Error Correction Term 0.1181 0.1188
(8.938) (8.867)
Regulator aged over 3 years 0.2325
(5.673)
Regulator aged over 3 years (t-1) 0.0286
(2.000)
33
Long-run Coefficient on Index (per 0.099
unit) (/)
Long-run Coefficient on Regulator 0.244
Aged over 3-years (/)
Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects Fixed Effects
Estimation method
Adjusted R-squared 0.955 0.156 0.159 0.955 0.150 0.149
S.E. of regression 0.265 0.084 0.084 0.266 0.084 0.084
F-statistic 448.7 4.676 4.784 450.3 4.676 4.51
Durbin-Watson 0.165 1.79 1.80 0.168 1.78 1.78
No of observations 608 582 582 610 583 583
Note: t statistics in parentheses
34
The key results are:
· The derived estimate of the long-run effect on generation capacity of having a
regulator (Ministry or autonomous) aged 3 years or more in the differenced
equation of column 6 is 24%. This is very similar to that in the LDV model
but lower than in the static model.
· The derived estimate of the long-run effect of a unit increase in the regulatory
governance index in the differenced equation of column 3 is almost 1%,
implying that the maximum score on the index is associated with almost 40%
higher generation capacity than under a Ministry regulator with no electricity
law. This is a lot higher than in the LDV model (26%) and also above the
level in the static model (35%)31.
· The estimates of are similar to each other and very similar to the implied
adjustment speed in the LDV model. A more sophisticated two-term error
correction model showed slightly faster adjustment, particularly after 5 years.
· The estimated response to changes in GDP and regulatory governance is slow.
Only 12% of the difference between actual and equilibrium long-run per
capita capacity levels is made up in the first year. The estimated adjustment
processes with both one and with two-term error correction factors are shown
below.
Alternative convergence paths
of tfel 1.5
Two EC
onit ce 1 terms
an
rb 0.5 One EC term
oporrP suid 0
1 3 5 7 9 11 13
Years
31 Note that a simple multiplicative factor applied to a linear index is likely to exaggerate the
maximum effect if there are diminishing returns to increases in regulatory governance quality.
35
These results provide strong support for the hypothesis that the regulatory governance
impact on generation capacity in developing countries is positive and sizeable but
takes time to build up.
5.2.3 Econometric Results for Models of Generating Capacity
Utilisation and Technical Losses
Table 8 presents some results relating to the impact of regulatory governance on
efficiency. For the reasons set out in Section 3.1.2, we deliberately estimated simple
models. Unfortunately, we were unable to find any reliable time-series data for our
countries on commercial losses, or quality of service or productivity.
The results were reasonably positive for capacity utilisation in generation but we
never found any positive or significant effect of any regulatory variable for (technical)
transmission and distribution losses.
Table 8: Utilisation and Technical Losses
Technical
losses in
transmission
Utilisation of and
generation distribution
Dependent Variable capacity* (%)
Explanatory variables 17 18
Real GDP per capita (log) 0.729 -0.841
(2.279) (-0.441)
Index of regulatory governance 0-4 0.079 0.219
(2.330) (1.016)
AR(1) 0.713 0.648
(23.365) (17.786)
FE + Prais FE + Prais
Estimation method Winsten Winsten
Adjusted R-squared 0.743 0.840
S.E. of regression 0.449 2.697
F-statistic 56.196 92.624
Durbin-Watson 2.138 2.032
No of observations 574 472
Note: t statistics in parentheses
*Utilisation = generation/(capacityx24x365)
36
The positive effect in the utilisation equation of the regulatory index (significantly
different from zero at the 1% level) was found in some but not all other equations
estimated.
In the equation reported, a 1 point increase in the regulatory index is associated with a
0.8% increase in utilisation so that utilisation with the maximum index score of 4
implies a 3.2% increase relative to countries with an index score of zero. Utilisation
rates also appear to be positively (and significantly) associated with higher GDP
within and between countries.
5.2.4 Endogeneity and Causality in Generation Capacity Models
5.2.4.1 Endogeneity
Much of the literature on regulatory effectiveness expresses concerns over the
endogeneity of:
(a) countries choosing to have an independent/autonomous regulatory
agency; and
(b) the quality of governance of that agency32.
This discussion echoes similar debates about the endogeneity of independent central
banks and how best to measure the impact of central bank independence eg on
inflation and growth. The concern is essentially that countries with better
(unobservable) governance have better functioning regulatory agencies eg because
they have socio-economic characteristics that better support the rule of law, contracts
and commercialisation. The problem is that it is very difficult to find good
instruments ie instruments that are both correlated with the suspected endogenous
variable and uncorrelated with the error term so that they can be treated as exogenous.
However, Edwards and Waverman (2004) have adopted a novel approach to this
using a rank-based instrument for their (12 element) EU telecom regulatory
governance index. This approach, taken from Evans and Kessides (1993), provides a
simple procedure, firstly, for the testing of whether or not there is evidence of
endogeneity; and, secondly, for deriving an IV estimator to control for the estimation.
The procedure is as follows. Firstly, we recalibrate our 4 element regulatory
governance index, so that all countries where entries are 1 or 2 are reclassified as 1
and all entries of 3 or 4 are reclassified as 2. Zeroes remain zero. We denote this as
the Rank Index. We then estimate the following equation:
Index(Cubbin-Stern) = a0 + a1RankIndex it+ uit (10)
32 See, for instance, Fink et al (2002), Gual and Trillas (2002) and Gutierrez (2004).
37
We then include the estimated residuals from (10) in the following equation for per
capita generation capacity:
Log(ELCAPPC)it = (a0 + ai) + a1 log(GDPPC)it + a2 ûit + vit (11)
Whether or not a2 is significantly different from zero provides a test as to whether or
not there is a potential endogeneity problem associated with our regulatory
governance measure. Similarly, including the predicted values of the Cubbin-Stern
index derived from (10) in our basic, fixed effects, static model provides an effective
instrument to estimate the effect of any endogeneity bias in practice.
As explained by Edwards and Waverman, the use of the Rank Index is, by
construction correlated with the original index but orthogonal to the error term
provided that a small change in the original index would not change the position in
the Rank Index. This can be expected to hold except for observations near the
thresholds between the close to the boundaries between the Ranks.
Adopting this procedure, we find that the coefficient on the residuals of (10) in the
basic static equation for per capita generation capacity has a t-value of 1.7, implying
that there is marginal evidence of endogeneity of the Cubbin-Stern regulatory index.
However, instrumenting the Cubbin-Stern index by using its predicted value from
(10) in place of the actual value produces virtually identical results an estimated
coefficient of 0.047 with a t-value of 4.3 in the instrumented case as opposed to an
estimate of 0.049 and a t-value of 4.0 in the non-instrumented case.
Like Edwards and Waverman (2004) and Gutierrez (2003), we find some weak
evidence of endogeneity of regulatory governance quality but very little change in
coefficient estimates from correcting for it. The test in our case is not as strong as in
Edwards and Waverman who have 12 rather than 4 initial governance levels.
However, we can with some confidence reject the proposition that our results can be
dismissed on the grounds of potential endogeneity of our indicator of regulatory
governance.
5.2.4.2 Causality
The question remains as to whether, looking forward, our regulatory governance
coefficient estimates have any causal interpretation. Even if, they are not statistical
artefacts arising from failures adequately to address dynamics or endogeneity, they
may still be merely descriptions of a past set of events that cannot be applied to future
electricity regulatory governance changes in sample countries let alone to the
introduction or development of electricity regulation in non-sample developing
countries33.
33 For the reasons stated in Sections 2 and 3, we would not wish to claim that they are applicable
to countries with an excess supply of generation capacity at any time during the period after
1980. This would exclude the Central and East European countries, the CIS and almost all
OECD countries.
38
One reason why this question might be asked is that the regulatory literature derived
from Levy and Spiller (1994) emphasises country-specific constitutional, legal,
economic, and political differences as being crucial for the success or failure of utility
regulation. Hence, a highly reduced form model that abstracts from all those issues
may well fail to reflect these local issues that seem to be so important in practice.
The answer to both these concerns lies in the importance of the country-specific fixed
effects. With 28 countries each having up to 21 years of data, we can obtain estimates
of the fixed effects which should capture most if not all of the factors identified by
Levy and Spiller and the subsequent literature. Hence, the estimated impact of eg
enacting a regulatory law plus an autonomous regulator in Chile or Sudan (both
countries in our sample) will be very different. That impact is the combination of (a)
the predicted effect of the relevant regulatory variables plus (b) each country's
predicted fixed effect. The Chilean fixed effect is strongly positive relative to the
sample average whereas that for Sudan is strongly negative.
In other words, the coefficients that we report are `highest common factor' estimates
of the impact of regulatory governance indicators where the fixed effects not just
control for but effectively "wash out" all the Levy and Spiller and similar factors.
But, this means that the regulatory governance effects that we report are not just
average cross-country sample effects but that they refer to a country with average
scores on country-specific fixed effects. Moreover, they are the impacts that one
might expect, looking forward, for a country:
With an average country specific fixed effect
Implementing an average quality law
Establishing an average quality autonomous regulator, etc.
It is for such a country that one might expect that implementing a best quality
electricity regulator would increase per capita generation capacity in the long run by
15-25%.
The policy implication of this is, firstly, that the quality of overall country governance
matters considerably for the impact of regulation on outcomes (eg as in the rule of
law); and secondly, that countries cannot expect to achieve the gains we have
estimated by enacting low quality regulatory laws or introducing autonomous
regulatory agencies with very low staffing levels. But, the corollary is that the
potential gains from introducing an electricity regulator could be significantly higher
than the average level for countries with good overall governance who deliberately try
to introduce best practice regulatory agencies and practices.
The argument above follows not just from the logic of our fixed effect modelling but
is supported by the suggestive implications of a potential interactive and/or
independent impact of overall country governance (eg as measured by the Kaufmann
index) on our regulatory governance measure. (See Section 5.1.2.2 above.)
These arguments do not, of course, apply just to this paper. They also apply to the
similar models of Gutierrez, Edwards and Waverman, etc.
39
6. Discussion of Results and Concluding Comments
6.1 Discussion of Results
The results of this study seem to provide a broadly consistent picture that the
existence of a regulatory agency with good governance characteristics not only can in
principle improve regulatory outcomes but seems actually to do so in practice. For
electricity supply industries in 28 developing countries in the 1980-2001 period, we
find that an index of regulatory governance is a consistently positive and statistically
significant determinant of per capita generation. Our results, using fixed effects
estimation methods, are similar to those found in for telecoms in developing countries
(e.g. Gutierrez, 2003).
The main positive findings are that, for developing countries:
1) Averaging over developing country regulatory agencies, the estimated long-
run impact on per capita generation capacity of a maximum regulatory
governance index score of 4 is of the order of 15-25% cet par and, in
particular, after controlling for country-specific fixed effects.
2) The estimated impact of regulation increases with experience at least for the
first 3-5 years or more. The cet par impact on per capita generation capacity
of a regulator (autonomous and/or Ministry) established at least 3 years is of
the order of 25-35%.
3) The effects on per generation capacity are robust not just to the inclusion of a
lagged dependent variable but also to the inclusion of a time trend and 3-year
lags on the regulatory variables. They are also robust to modelling via an
error correction model and to IV modelling to allow for potential endogeneity
biases.
4) The effects of the enactment of (a) a regulatory law, (b) of having an
autonomous regulator and (c) licence fee funding of the regulatory agency
were each positive and statistically significant at the 1% level.
5) There is some evidence, albeit weak, that better overall country regulatory
governance is a statistically significant determinant of generation capacity
utilization (a good proxy for availability).
6) There is some evidence, albeit weak, that the better the rule of law, the
stronger the regulatory effect.
7) There is reasonable evidence that superior regulatory governance improves
generation utilization rates.
There are, however, some negative findings. These include the following
40
1) There was no evidence of any significant, positive effect of any of our
regulatory governance measures on transmission and distribution losses.
2) There was no reliable evidence in this data set that competition or privatization
were significant determinants of generation capacity either individually or
when interacted with regulatory governance. However, the data set we used
was much stronger on regulatory variables than on competition and
privatisation variables.
On the whole, we were surprised at the strength and robustness of the positive results.
The data set we used has a number of major weaknesses in spite of being the best
currently available. Among the main weaknesses of the data set are:
The absence of any data on regulatory practice, including government (and/or
electricity company) malpractice toward supposedly independent regulatory
agencies (e.g. high within-term turnover rates of regulatory office
heads/commissioners).
The absence of any reliable cross-country data on ESI efficiency and
productivity or on service quality and revenue collection.
The limited time dimension to the regulatory data and the extremely limited
time dimension to data on privatization and competition.
Potential omitted variable biases from the inability to test for the inclusion of
many potentially significant variables.
The limited and possibly unrepresentative sample of countries.
It is to be hoped that some of the major data weaknesses can be remedied e.g. by
systematic data collection exercises of the sort that have been carried out for telecom
reform.
5.2 Concluding Comments
In this paper we have presented evidence that suggests that good regulatory
governance does have a positive and statistically significant effect on some electricity
industry outcomes in developing countries notably per capita generation capacity
levels - but we have not examined why this is so.
To examine why and how regulation operates to improve outcomes is not a task that
obviously recommends itself to econometric analysis. We suggest that, at least at this
stage, it is better pursued by case studies with econometric work being concentrated
on whether the results reported in this paper are confirmed in subsequent analyses e.g.
with superior data, particularly on regulatory practice, privatization and competition
variables.
41
Nevertheless, we are confident that the results reported here are entirely consistent
with the literature on the role of institutions in economic growth. The key point is
that regulatory agencies with better governance are:
· Less likely to make mistakes
· More likely to correct mistakes speedily
· Less likely to repeat mistakes
· More likely to develop procedures and methodologies that involve participants
and develop good practice.
All of these reduce uncertainties for commercially operating companies particularly
private and foreign companies. This is especially important to sustain and encourage
long-lived, sunk investments in highly capital-intensive industries at a reasonable cost
of capital. As such, regulatory agencies, which have and maintain good governance,
provide an effective underpinning for the operation of contracts as well as sound
regulation of monopoly elements.
The utility service industries like electricity supply may be considered as a microcosm
for considering the role of institutions in sustaining investment, efficiency and
growth. But, in fact, they are a touchstone. Given their role in supporting growth as
well as their technical characteristics, electricity and similar industries are among
those most in need of strong and effective regulatory frameworks. Hence, we suggest
that our positive results on the role of good governance support and enhance the
lessons of similar studies for independent central banks and telecom reform as well as
support the general arguments of North, Rodrik and others on the role of effective and
evolving institutions for sustained growth.
It remains to be seen whether the results reported in this paper survive in the light of
further analysis and can be replicated with better data.
42
REFERENCES
Bacon, R. and Besant-Jones, J. (2001), "Global Power Reform, Privatization and
Liberalization of the Electric Power Industry in Developing Countries", Annual
Review of Energy and the Environment, Vol. 26, pp 331-59.
Cubbin, J.S. and Stern, J., (2004) "Regulatory Effectiveness: The Impact of Good
Regulatory Governance on Electricity Industry Capacity and Efficiency in
Developing Countries", London Business School regulation Initiative Working Paper
No 57.
Domah, P., Pollitt, M. & Stern, J., (2002), "Modelling the costs of electricity
regulation: evidence of human resource constraints in developing countries", London
Business School Regulation Initiative Working Paper No. 49 and CMI Working Paper
No.11.
Edwards, G.E. and Waverman, L. (2004), "The Effects of Public Ownership and
Regulatory Independence on Regulatory Outcomes: A Study of Interconnect Rates in
EU Telecommunications" London Business School EURI Working Paper.
Evans, W. and Kessides, I. (1993), "Localised Market Power in the US Airline
Industry", Review of Economics and Statistics, Vol 75 (1), 66-73.
Fink, C., Mattoo, A. & Rathindran, R., (2002), "An assessment of
telecommunications reform in developing countries", World Bank Working Paper No.
2909.
Guasch, J.L., Laffont, J.J. & Straub, S., (2003), "Renegotiation of concession
contracts in Latin America", World Bank Working Paper No.3011.
Gutierrez, L.H., (2003), "The effect of endogenous regulation on telecommunications
expansion and efficiency in Latin America", Journal of Regulatory Economics, Vol.
23:3 pp. 257-286.
Hsiao, C. (1986), Analysis of Panel Data, Cambridge University Press.
Levy, B. & Spiller, P.T., (1994), "The institutional foundations of regulatory
commitment: a comparative analysis of telecommunications regulation", The Journal
of Law, Economics & Organisation, Vol. 10 pp.201-246.
Newbery, D. M., (1999), "Privatisation, Restructuring and Regulation of Network
Utilities", The MIT Press.
North, D. C., (1990), "Institutions, Institutional Change and Economics
Performance", Cambridge University Press.
Pargal, S. (2003) "Regulation & private sector investment in infrastructure: Evidence
from Latin America", World Bank Policy Research Working Paper No. 3037.
43
Prayas Energy Group Report, (2003), "Performance of private electricity distribution
utilities in India: need for in depth review & benchmarking", Prayas Occasional
Report, Prayas Energy Group, Pune, India.
Rodrik, D, Subramanian, A., and Trebbi F., (2002), Institutions Rule: The Primacy of
Institutions Over Geography and Integration in Economic Development", John F
Kennedy School of Government Working Paper, Harvard University.
Rodrik, D., (2003), "Growth Strategies", John F Kennedy School of Government
Working Paper, Harvard University, (Forthcoming in Handbook of Economic
Growth).
Smith, W., (1997) "Utility Regulators The independence debate", The World Bank
Public Policy for the Private Sector Note No. 127.
Stern, J. & Holder, S., (1999) "Regulatory governance: criteria for assessing the
performance of regulatory systems. An application to infrastructure industries in the
developing countries of Asia", Utilities Policy, Vol. 8, pp.33-50.
Stern, J. & Trillas, F., (2001), "Regulation of telecoms: what works and why? Lessons
from Independent Central Banks", Business Strategy Review, Vol 12, No. 4, pp 17-28.
Stern, J., (2003), "Regulation & contracts for utility services: Substitutes or
complements? Lessons from UK historical experience", London Business School
Regulation Initiative Working Paper No.54.
Stern, J. and Cubbin J.S. (2003), "Regulatory Effectiveness: The Impact of
Regulation and Regulatory Governance Arrangements on Electricity Outcomes A
Review Paper", London Business School Regulation Initiative Working Paper No 56
Zhang, Y. F., Kirkpatrick, C. & Parker, D., (2002), "Electricity sector reform in
developing countries: an econometric assessment of the effects of privatisation
competition and regulation", Centre for Regulation and Competition, Manchester
CRC Working Paper No. 31.
44
APPENDIX: LIST OF COUNTRIES IN SAMPLE
Argentina
Barbados
Bolivia
Brazil
Chile
Colombia
Costa Rica
Dominican Republic
Ecuador
El Salvador
Ethiopia
Grenada
India
Indonesia
Jamaica
Kenya
Malaysia
Mexico
Nicaragua
Nigeria
Paraguay
Peru
Philippines
Sudan
Trinidad
Uganda
Uruguay
Venezuela
45