SH8HS The World Bank JANUARY NUMBER 62 PUBLIC SECTOR W1ternational experience with civil 0 servoce censuses Civil service censuses have often failed to deliver lasting improvements in devel- oping countries. To be more effective, such censuses should have clearer objectives, use more appropriate methodologies, and pursue long-term improvements in pay- roll and personnel systems. Since the early 1980s many developing coun- censuses have been conducted to define the How can civil service ties have tried to count and contain the num- scope and scale of the civil service as part ber of civil servants as part of public sector of the transition to a market economy.) censuses be made reforms-including reforms funded by the Africa, with the majority of such exercises, World Bank. These initiatives, variously has the weakest civil service controls. Many more effective? described as censuses, enumerations, head- censuses were conducted in Latin Amer- counts, staff audits, payroll verifications, and ica in the 1960s and 1970s, but they have payroll reconciliations, have been conducted become far less common as civil service con- in widely varying contexts, pursuing many trols have improved and governments have different objectives and using a range of been able to rely on management infor- methodologies. But a recent survey by the mation systems for data on the size and com- World Bank and the International Records position of the civil service. Management Trust shows that many such exercises have had limited success. How can Census objectives these efforts be made more effective? Civil service censuses have been conducted for a variety of reasons, but three objectives Regional distribution of are common: censuses * Cutting costs. Some of the earliest cen- The survey reviewed 31 civil service censuses suses were conducted to reduce the num- conducted between 1978 and 2000-19 in ber of "ghost" workers (fictitious, Africa, 5 in Europe and Central Asia, 3 in duplicate, or otherwise erroneous pay- South Asia, 2 in the Middle East, 1 in East roll entries) and so quickly and relatively Asia, and 1 in Latin America (table 1). painlessly cut payroll costs. Such exer- Although the survey was not comprehen- cises were also conducted to provide sive, this selection is considered represen- rough data for downsizing exercises- tative of the regional distribution of such typically voluntary retirement schemes. exercises, which in recent decades have Cutting costs was the primary objective been concentrated in Africa. of the censuses in Cameroon, Chad, and Almost everywhere, there appears to be Nigeria, for example. an inverse correlation between the fre- * Restructuring. Later civil service censuses quency of such censuses and the quality of were conducted notjust to cut costs but public administration. (A possible excep- also to enhance productivity by restruc- tion is Europe and Central Asia, where such turing departments and functions, rede- FROM THE DEVELOPMENT ECONOMICS VICE PRESIDENCY AND POVERTY REDUCTION AND ECONOMIC MANAGEMENT NETWORK TABLE 1 A SAMPLE OF CIVIL SERVICE CENSUSES BY REGION, 1978-2000 Europe aind Africa Central Asict South Asia Middle East East Asia l.atin Ameicat Beniin Georgia Nepal Lebanon Cambodia Argenitinia Bllurkina Faso Kazakhstani Orissa (India) Yemen Clmeil-oon Kosovo Pakistan (Central African Rep. l.ithuania C.had Romania Ethiopia Garribia, The Ghan a CGuinea Keniva Civil service Niger Niger-ia censuses are Rvandla Senegal conducted for sierrLeone Uganda different reasons Zaiibia ZiLnbahwc and using various SWIPCe-u SLII cV 1W Ilie Worlid Baik ainti Internatio ial Records NMaiiagement Trist. approaches ploying staff, building capacity, and shot" data, which is more accurate thall the improving human resource maniage- data in payroll andc personnel records or ment. Such censuses focused on estab- can be used to cross-check such data, and lishinlg a detailed profile of public censuses designed to lay the foundations employees-such as data on the mix of for a new, regularly updated, permanent sys- skills, ages, and geniders-to facilitate tern for collecting data and setting controls. departmen tal or fuLnctionial reviews. This approach r eflects the shift in civil service Methodologies used r eforms frorm simple downsizing toward The survey found that three main enhlancinig public sector productivity, approaches were used to conduct censuses. SUCh1 censIses hiave been conducted in Guinlea, Roumaniia, and Sierr a Leone. Physical headcounts Establishing a baseline as part of transition. A physical headcount, sometimes called a Some recenit censuses have been con- staff audit, focuses on determining the lum- ducted to establish a baseline profile of ber of staff employed (as opposed to the the civil service in transition economies, number of positions established) and on to facilitate planning and budgeting. Typ- whether the names on a payroll belong to ically, the focus is on defining the scope genuine employees. This approach is used ancl scale of the civil service and of most often when the goal is to cut costs by functions Wit]thin the civil service in sys- eliminating ghost workers. A headcount typ- tems where the distinction between ically involves trained teams traveling to var- public and private sectors and social and ious parts of the census area, where commercial activities has not previously individual employees are required to pre- been clear. Such censuses have been sent themselves, often with identification conducted in Georgia, Kazakhstan, and and sometimes with documentation (such Kosovo. as photocopies of letters of appointment or A further- distinction can be made birth records). These data are then checked between censuses designed to obtain "snap- off, usually against the payToll. In some cases PREMNOTE 62 JANUARY 2002 photographs or fingerprints are taken. Phys- because of the difficulty of identifying cred- ical headcounts have been used in Cam- ible alternative data sources. Reconciliation bodia, Cameroon, and Zambia. has been used in Yemen, where a headcount Headcounts can involve significant costs was considered politically impractical, and and logistical challenges, and the quality of has been used as a supplementary check in the data is often challenged. Case studies countries including Uganda and Zimbabwe. suggest that consultants and government Reconciliation can be extremely time- officials tend to greatly underestimate the consuming and becomes more complicated logistical challenges, particularly in remote as the number of personnel databases grows. or rural areas. In Uganda, for example, early The Gambia, for example, has departmen- censuses required that civil servants appear tal personnel databases as well as a central with photocopies of key documents-even system for human resource information and though copying facilities were unavailable the payroll. A reconciliation exercise is only To succeed, in many areas. as meaningful as the data sources on which it is based. censuses must be Questionnaires These three methodologies are not mutu- A questionnaire is the main alternative to a ally exclusive. For example, Ghana and Tan- planned strategically physical headcount and is used most often zania combined elements of the headcount when more detailed data on human approach and the questionnaire approach. and structured resources are needed for restructuring or baselining efforts. This approach typically Lessons from international relative to the local involves the distribution of questionnaires experience to employees or employers, who must sub- The survey suggests that civil service cen- context and capacity mit information back up the line, taking suses have often had mixed results. Cen- responsibility for its accuracy. The data are suses have helped eliminate ghost workers then compiled in some type of database. in some countries, resulting in moderate or Questionnaire-based censuses have been even significant savings. But such audits have used in Georgia, Tanzania, and the Indian been costly and, in the absence of routine state of Orissa. civil service controls, there has often been Questionnaire-based censuses are often no mechanism to ensure that ghost work- akin to a decentralized headcount. Key to ers do not get back on the payroll. Censuses the success of this approach is achieving the have contributed to efforts to restructure cooperation of the local data provider, who civil services, but such exercises are typically may otherwise accord the exercise low pri- incremental processes and need to be ority. Census-takers must be clear about the informed by regularly updated information information being sought and keep data on personnel management. requests short and to the point. In Uganda What has been learned? First, censuses the 7 original pieces of data specified grew are costly and must be planned strategically. to more than 220. In Orissa no standard Objectives must be clear-whether they are form was used, creating confusion about short-term savings or a broader initiative to the questions being asked and resulting in establish a comprehensive database on many inconsistent, inaccurate, and illegible human resources. Moreover, censuses must responses. be structured relative to the local context and capacity. Information requirements Payroll reconciliation should be analyzed carefully and balanced A third approach involves reconciling the against the capacity to verify, analyze, and payroll against alternative data sources- use data gathered in a timely and effective such as individual personnel files or service way. Census enumerators need to be well books, the nominal roll or established reg- trained, and there must be effective mech- ister, and other databases. This appears to anisms for resolving disputes about data and be the least favored methodology, probably rectifying errors. Care should be taken to PREMNOTE 62 JANUARY 2002 ensure that the exercise is comprehensive, counted.) If managers know that they will or its restults will be of limited value. be held accountable for inaccturacies and The second lesson is that whenever pos- irregularities, external auditors conducting sible, censuises need to be devised as part random checks could have as much impact of a long-ter-m institutional investment in as costly complete audits. (It is common, for basic payroll ancl personnlel systems. This example, to see a pronounced drop in the reqtiries that the census be part of a strat- number of inaccurate personnlel entriesjtist egy fo)r maintaining accul-ate information before a census is condutcted.) in tlie future. Suclh exercises are mtich The fourth lesson is the importanice of riskier, halrder, and mor-e complex than sim- coordinated action, top-level support, and ple heaclcotiints. (In one West African coun- local ownership of the exercise's design and try, for example, it took nearly five years to delivery. Many efforts to improve person- establish a coinptuterized human resource nel data and civil service contr-ols have failed clatabase.) The Imlost serious census prob- because of inadequate coordination lems have occurr-ed whlen methodologies between key ministries, usually those respon- more suited for simple headcounts have sible for managing personnel and those beenl used w,vith a view toward solving longer- responsible for managing the payroll. Many term problems of record maintenance. If efforts to eliminate ghost workers have failed costly repeats are to be avoided, it is essen- because of lack of coordinated follow-up tial to establish ways of dealing with census between central and line departments. In retur-ns before the census is conducted- unsuccessful censuses the exercise was usu- including throtuglh infor-miiationi technology ally perceived as an end in itself, conducted and trained personniel-and to start updat- to fulfill a condition imposed by external ing the database as soon as the census is donors. In more successful census exercises, complete. At the same time, efforts should local ownership was repeatedly cited as a be made to idenitify and fix failures or cir- reason for success. cutriventions of earlier management systems. This note was written b Robert Beschel (Senior The tlhird lesson is the need for incen- Public Sector Specialist, South Asia Region) and tives to ensuL-e compliance. Staff and man- Edward Mountfield (Economist, East Asia and agers shotild be encouLraged to ensure that Pacific Region) drawing on research by Tripti payroll and personnel data are complete Thomas (Research Analyst, Public Sector Group, and accuLrate. Such efforts may involve, for PREM Network) and the International Records instanice, stopping salary payments for staff Management Trust (www.irmt.org). oni the payroll who are not enumerated. Ifyouareinterested insimilartopics, conside7- (Thouglh wher-e such tactics are used, it is joining the Administrative and Civil Service importan t to allow enough time for people Reform Thematic Group. Contact Nick Manning on leave and those carrying out legitimate (x31202) or click on Thematic Groups on businiess away fi-om work to return and be PREMnet. This note series is intended to summarize good practice and key policy find- |Ei b|| ings oni PREM-related topics. The views expressed in these notes are those of the aLithors and do not necessarily reflect the views of the World Bank. PREM- rnotes are distributed widely to Bank staff and are also available on the PREM wvebsite (http://prem). If you are interested in wNTiting a PREMnote, email your idea to Sarah Nedolast. For additional copies of this PREMnote please contact PuvenRebioandcmMdanaleme thePREMpAdvisoryService at x87736. Prepared for World Baink staff