Since the early 1980s, many developing countries have attempted to produce an accurate count of their civil servants as a precursor to reforms aimed at containing or reducing the number of public employees and/or reforming public sector pay. These initiatives, variously described as censuses, enumerations, headcounts, staff audits, payroll verifications, and payroll reconciliations, have been conducted in widely varying contexts, to meet a number of different objectives, using a range of methodologies. Many of these reforms have been funded by the World Bank.

A recently conducted survey of international experience with civil service censuses suggests that many such exercises have had only limited success. While perhaps generating short-term savings, they often fail to lead to sustainable improvements in civil service management information or the strengthening of establishment controls.  This page summarizes the findings of the survey and explores how these efforts can be made more effective.

The survey mentioned above reviewed 31 censuses conducted between 1978 and 2000 –- 19 in Africa, 5 in Europe and Central Asia, 3 in South Asia, 2 in the Middle East, 1 in East Asia, and 1 in Latin America. Case studies were developed for eight of these countries. Although the survey was not comprehensive, this selection is considered broadly representative of the regional distribution of such exercises, which in recent decades have been particularly concentrated in Africa.

A sample of civil service censuses by region, 1978–2000

Almost everywhere, there appears to be an inverse correlation between the frequency of such censuses and the quality of public administration. (A possible exception is Europe and Central Asia, where such censuses have been conducted to define the scope and scale of the civil service as part of the transition to a market economy.) Africa, with the majority of such exercises, has the weakest civil service controls. Many censuses were conducted in Latin America in the 1960s and 1970s, but they have become far less common as civil service controls have improved and governments have been able to rely on management information systems for data on the size and composition of the civil service.

Civil service censuses have been conducted for a variety of reasons, but three objectives are common:

Cost-cutting: Some of the earliest civil service censuses were conducted to reduce the number of “ghost workers” (fictitious, duplicate or otherwise erroneous entries) and thereby cut payroll costs relatively quickly and painlessly. Censuses were also conducted to provide rough data for downsizing efforts –- typically voluntary retirement schemes. Cutting costs was the primary objective of the censuses in Cameroon, Chad and Nigeria, for example.

Restructuring: Later civil service censuses were conducted not just to cut costs but also to enhance productivity by restructuring departments and functions, redeploying staff, building capacity, and improving human resource management. These censuses focused on establishing a detailed profile of public employees –- such as data on the mix of skills, ages and genders –- to facilitate departmental or functional reviews. This approach reflects the shift in civil service reforms from simple downsizing toward enhancing public sector productivity. Censuses of this variety have been conducted in Guinea, Romania, and Sierra Leone.

Establishing a baseline as part of transition: Some recent censuses have been conducted to establish a baseline profile of the civil service in transition economies, to facilitate planning and budgeting. Typically, the focus is on defining the scope and scale of the civil service and of functions within the civil service in systems where the distinction between public and private sectors and social and commercial activities has previously not been clear. Such censuses have been conducted in Georgia, Kosovo and Kazakhstan.

A further distinction can be made between censuses designed to obtain “snapshot” data more accurate than the data in payroll and personnel records or as a cross-check against such data, and censuses designed to lay the foundations for a new, regularly updated, permanent system for collecting data and setting controls.

For more country examples, see Country Examples by Primary Objective.

The survey found that three main approaches were used to conduct censuses.

Headcounts can involve significant costs and logistical challenges, and the quality of the data is often challenged. Case studies suggest that consultants and government officials tend to underestimate the logistical challenges of these censuses, particularly in remote or rural areas. In Uganda, for example, early censuses required that civil servants appear with photocopies of key documents –- even though copying facilities were unavailable in many areas.

Questionnaires. A questionnaire is the main alternative to a physical headcount, and is used most often when more detailed data on human resources are  needed for restructuring or baselining efforts. This approach typically involves the distribution of questionnaires to employees or employers, who must submit the information back up the line, taking responsibility for its accuracy. The data are then compiled in some type of database. Questionnaire-based censuses have been used in Georgia, Tanzania and the Indian state of Orissa.

Questionnaire-based censuses are often akin to a decentralized headcount. Key to the success of this approach is achieving the cooperation of the local data provider. The census-taker must be clear about the information being sought and keep data requests short and to the point. In Uganda, the 7 original pieces of data specified grew to more than 220. In Orissa, a standard form was not used, creating confusion about the questions being asked and resulting in many inconsistent, inaccurate, and illegible responses.

Payroll reconciliation. A third approach involves reconciling the payroll against alternative data sources –- such as individual personnel files or service books, the nominal roll or established register, or other databases. (See Cain and Thurston for further details.) This appears to be the least favored methodology, probably because of the difficulty of identifying credible alternative data sources. Reconciliation has been used in Yemen, where a headcount was considered politically impractical, and has been used as a supplementary check in countries including Uganda and Zimbabwe.

Reconciliation can be extremely time-consuming and becomes more complicated as the number of personnel databases grow. The Gambia, for example, has departmental personnel databases as well as a central system for human resource information and the payroll. A reconciliation exercise is only as meaningful as the data sources on which it is based.

These three methodologies are not mutually exclusive. For example, Ghana and Tanzania combined elements of the headcount approach and the questionnaire approach. 

For more country examples, see Country Examples by Methodology.

The survey suggests that civil service censuses often have mixed results. Censuses have helped to eliminate ghost workers in some countries, resulting in moderate or even significant cost savings. Still, such audits have been costly and, in the absence of routine establishment controls, there often has been no mechanism to ensure that ghosts do not get back on the payroll. Censuses have contributed to efforts to restructure civil services, but such exercises are typically incremental processes and need to be informed by regularly updated information on personnel management.

What has been learned? First, censuses are costly and must be planned strategically. Objectives must be clear –- whether they are short-term savings or a broader initiative to establish a comprehensive HR database. Moreover, censuses must be structured relative to the local context and capacity.  Information requirements should be analyzed carefully and balanced against the capacity to verify, analyze and use data gathered in a timely and effective way. Census enumerators need to be well trained, and there must be effective mechanisms for resolving disputes about data and rectifying errors. Care should be taken to ensure that the exercise is comprehensive or the results will have limited value.

The second key lesson is that, whenever possible, censuses need to be devised as part of a long-term institutional investment in basic payroll and personnel systems. Such exercises are risky and more complex than simple headcounts. (In one West African country, for example, it took nearly five years to establish a computerized human resource database.)  The most serious census problems have occurred when methodologies more suited for simple headcounts have been used with a view toward solving longer-term problems of record maintenance. If costly repeats are to be avoided, it is essential to establish ways of dealing with census returns before the census is conducted –- including through information technology systems and trained personnel –- and to start updating the database as soon as the census is complete. At the same time, efforts should be made to identify and fix failures or circumventions of earlier management systems.

The third lesson is the need for sanctions and incentives to ensure compliance. Staff and managers should be encouraged to ensure that payroll and personnel data are complete and accurate. Such efforts may involve, for instance, stopping salary payments for staff on the payroll who are not enumerated. (Although, where such tactics are used, it is important to allow sufficient time for people on leave and those carrying out legitimate business away from work to return and be counted.) If managers know that they will be held accountable for inaccuracies and irregularities, external auditors conducting random checks could have as much impact as costly complete staff audits. (It is common, for example, to see a pronounced drop in the number of inaccurate personnel entries just before a census is conducted.) 

The fourth lesson is the importance of coordinated action, top-level support and local ownership of the exercise’s design and delivery. Many efforts to improve personnel data and establishment controls have failed because of inadequate coordination between key ministries, usually those responsible for managing personnel and those responsible for managing the payroll. Many efforts to eliminate ghost workers have failed because of lack of coordinated follow-up between central and line departments. In unsuccessful censuses, the exercise was usually perceived as an end in itself, conducted to fulfill a condition imposed by external donors. In more successful census exercises, local ownership was repeatedly cited as a reason for success.

• Cain, Piers and Thurston, Anne. “Personnel Records: A Strategic Resource for Public Sector Management.” A Public Service Theoretic Series, Commonwealth Secretariat, March 1998.
• McCallum, Neil, Vicky Tyler. International Experience with Civil Service Censuses and Civil Service Databases. Annex 4. International Records Management Trust. London, UK. May 2001.
• De Merode, Louis, and Charles S. Thomas. “Implementing civil service pay and employment reform in Africa: the experience of Ghana. The Gambia, and Guinea.” Rehabilitating Government: Pay and Employment Reform in Africa. Eds., David L. Lindauer, Barbara Nunberg. World Bank Regional and Sectoral Studies. The World Bank. Washington DC 160-194
• Nunberg, Barbara and John Nellis. "Civil Service Reform and the World Bank." World Bank Discussion Paper 161. The World Bank. Washington DC. 1995.
• Nunberg, Barbara. “Experience with civil service pay and employment reform: an overview.” Rehabilitating Government: Pay and Employment Reform in Africa. Eds., David L. Lindauer, Barbara Nunberg. World Bank Regional and Sectoral Studies. The World Bank. Washington DC 119-159.

This page is based on a PREM Note prepared by Robert Beschel and Ed Mountfield (East Asia PREM), drawing on research by Tripti Thomas (PREM Public Sector Anchor) and the International Records Management Trust.  It was submitted on 11 December 2001.