Introduction

The objective of this note is to review briefly some of the main issues facing a Public Employment Service (PES) when it wants to exploit its administrative records as basis for information about the structure and developments in the labour markets which it is serving. The main part of this note review the quality issues which needs to be thoroughly understood in order to be able to use these records effectively, and the consequences for the use of these records when producing Labour Market Information (LMI). The following sections discuss the specific issues for the use of the PES' records on job-seekers (or unemployment benefit claimants) and vacant jobs respectively; looking both at issues related to statistics on successive stocks, and at issues related to statistics on changed situations (i.e. 'flows'). The discussion will focus on the use of the administrative records as basis for statistics, and will ignore the possibilities for using them as basis for other forms of information relevant for understanding the situation and developments in the labour market. This is a deliberate choice, reflecting the author's experiences, but it seems clear that the administrative registrations may be used to generate also other forms of information which are useful when trying to understand the labour market and the forces influencing how workers and employers behave. Further elaboration on the points raised can be found in other papers written in the ILO, e.g. Hoffmann, 1995, 1996, and 1999a, EASMAT, 1998 and Phan et al, 2000. These papers, as well as ILO, 1999 and 2000, include references to national experiences as well as to other, relevant discussions. ⁽²⁾

Quality issues

The quality issues related to the use of any administrative records as basis for statistics arise from the simple fact that the administrative data are collected as part of an administrative process. This has consequences for the coverage, timeliness, frequency of updating, validity, reliability, consistency and degree of detail of the derived statistics in relation to the requirements which follow from the descriptive and analytical use to be made of them. The mere fact that the data are recorded as part of an administrative process will have consequences for these quality elements, even if statistical considerations are allowed to influence the contents and procedures of the administrative reporting system. Experience shows that this is a big if, and even when being considered seriously in determining what the registrations should cover, the statistical concerns may not influence the reporting system's actual operation, nor the public's reaction to the rules and regulations being administered.

Coverage

Deficiency in coverage will concern the type of units actually covered as well as the degree of coverage. The persons registering with a PES as 'job-seeker' or 'unemployment benefit claimant' (the 'PES unemployed') will normally only do so if they expect this to be worth the effort of doing so: Even persons who satisfy the rules for being eligible for support from the PES may not register if they are looking for a type of work for which employers normally will not seek the recruitment assistance of the PES, or if the (subjective plus objective) costs of registering exceeds the expected benefits. ⁽³⁾ Employers will only seek the assistance of the PES in filling vacancies if they expect to find job-seekers with the preferred skills or experience by this method. Even if there is a legal requirement to register vacancies, this obligation will only be respected if the costs of doing so are less than the benefits obtained, including any sanctions in the event that the non-compliance is discovered and acted upon. In general it is the attitude of the public to the PES and its services, including the efficiency of the agency and adequacy of itsresources, which will determine the extent to which the intended coverage of the registrations corresponds to the actual coverage. In addition it is clear that the intended coverage of the registrations is determined by the rules and regulations guiding the work of the PES, and these may not correspond to the coverage which is required for certain labour market issues to be described and analyzed. The conceptual differences between PES-unemployment and 'unemployment' as measured by Labour Force Surveys (LFS) are well known, and Hoffmann, 1999 looks at the corresponding issue with respect to vacancies. However, there the issue is mostly theoretical, as there in most countries are no statistical surveys which produces statistics on vacancies which 'compete' with those (which can be) produced by the PES.

Timeliness

In general the timeliness of data from administrative registrations depends on who isresponsible for reporting to the administrative authority and on the incentive for timely reporting. The time needed by the agency to process the registrations will also significantly influence the timeliness of the statistics that can be derived from them. In a PES where the initial registrations are done in local offices while the statistics based on them are produced and released centrally, the time needed locally to prepare the reports to the central office, the time needed for the transfer and the time needed in the central office to process the reports received will be important determinants on the timeliness of the statistics produced. As will be discussed below what happens at the first stage is normally most important also for the overall timeliness of the published statistics. ⁽⁴⁾

Frequency

In principle the registrations of PES-unemployed and vacancies can normally take place continuously, i.e. on every work-day. When this happens the basic registers are always up-dated and one could in principle produce daily statistical reports. In practice most PES will publish certain statistics at monthly intervals and others less frequently, but examples exist of weekly statistical compilations. However, although contacts with 'clients' will be continuous their registration may nevertheless be in batches at the end of the week or month. In particular when there are rapid increases in work-loads without corresponding increases in the capacity of the PES this can easily lead to backlogs in the registrations, which may be carried over to the next period, thus further influencing the timeliness of the statistics. When this happens it is important to recognize that unchanged frequency in producing and releasing statistics based on the registrations which have taken place, will lead to statistics which are misleading with respect to the changes which are actually happening.

Those statistics which are thought to provide indications on short term business cycle developments are those which are demanded with the highest frequency as well as the best timeliness: 'New entrants in to PES-unemployment' and 'newly reported vacancies' are examples of such statistics which also can be produced rapidly in an effectively functioning PES statistical operation. As will be discussed below, the more common 'headline' numbers on the stock of 'PES-unemployed' persons and 'vacancies' are much more problematic with respect to timeliness and therefore also frequency.

Validity of measurements

When making the initial registrations the PES officials have to find the characteristics which are most valid and pertinent for the execution of their duties, and which reflect the laws and regulations which apply. In principles such characteristics and their representation, i.e. the value sets used, may be rather different from those variables and categories that are most valid for statistical description and analysis. In practice this does not seem to be the case for the registrations taking place in PES, with the exception of the basic definitions of 'PES-unemployment' and 'vacancies' already noted above when discussing 'coverage'.

Reliability of measurements

The reliability of the information in administrative records depends on a number of factors, the most obvious one being the incentives for clients to give correct (or incorrect) information. Some 'answers' may result in a better outcome for the client than others. ⁽⁵⁾ The reliability of the recorded information will then depend on the probability of errors being discovered and on the costs to the client if it is discovered that the information given is incorrect. Another factor is the extent to which the information is actually helpful for the execution of the official's tasks. If it does not matter for the 'outcome' of the case the official may not take care to ensure that correct information is recorded. This is also why it often happens that when it s discovered that incorrect information has been recorded, for whatever reason, the information will not necessarily be corrected. If, for example, the placement officers have learnt how to cope with incorrect or imprecise registration of "occupation" on the records of job-seekers and vacancies, they may not see the need to correct the records once a satisfactory match has been achieved, thus leaving uncorrected the record which will be used to produce statistics on the occupational distribution of job seekers or vacancies. These types of errors are likely to be systematic, not random, with obvious quality consequences for the statistics.

Consistency of measurements

The demand for statistics which are consistent both over space and time leads to a need for the initial PE-registrations, which are made in the local offices to be consistent when carried out at the same time, and also to a need for them to be consistent over time unless there is a clearly documented reason, such as changes in rules and regulations, why they should not be so. It is therefore important to establish the extent to which the local offices work with clear and comprehensive instructions and guidance from the central office, and to what extent this office can to check the reliability and consistency of the information submitted to it and to what extent it has the authority to order the steps necessary to improve data quality if deficiencies are discovered. Normally the problems of consistency over timeis more difficult: The most obvious set of problems, and the easiest ones to monitor, if not overcome, relates to changes which are made to formal rules and regulations determining the work of the responsible PES, in particular those which concerns the coverage of the financial support schemes which it administers and the funds available to it under those schemes. More insidious are all the small, "informal" modifications in actual reporting and recording practices which follow from changes in budgets, workloads and priorities for the PES as a whole and locally. Because servicing the information system is often seen as incidental to the "real" tasks of the employment service - such as helping unemployed persons to find jobs or a desperate employer to find qualified staff - serving the information system may tend to be given lower priority if budgets are reduced or squeezed through uncompensated increases in the workload. This will frequently result in unreported reductions in data quality and may result in systematic changes compared to previously generated statistics ⁽⁶⁾. Unreported improvements in data quality, for example as a consequence of new technology for recording the information, may of course be just as disruptive for the consistency of time-series, at least in principle. In addition to these organization-related issues, there may be adjustments over time in how the public behaves in relation to the PES - adjustments which may reflect an improved understanding of how to use, or misuse, the system rather than reflect a change in the underlying circumstances which the data are intended to reflect. One example is that the number of applicants for certain benefits may increase following news reports or a campaign to increase public awareness of the scheme. Similarly, persons who may be eligible for certain services or benefits may conclude that the efforts needed to obtain them are not worth the effort if the PES develops a reputation for poor service.

Degree of detail

This is one of the quality dimensions where statistics based on the administrative registrations of the PES frequently are seen as superior to those based on statistical surveys. As the statistics are based on a full count on the units which have been registered and which are considered relevant, there is no need to be concerned with the imprecisions of statistics based on samples. The degree of detail in the statistics produced will consequently be limited only by the degree of detail in the initial registrations, e.g. about the branch of industry of employers announcing vacancies, and the manner in which these registrations are processed. ⁽⁷⁾ This is obviously one of the strong quality points for statistics produced from PES records as compared with survey-based statistics, together with timeliness and frequency, and this advantage is particularly strong when it comes to the possibility of providing statistics which refer to local labour markets. For reasons which may be linked to how political and administrative constituencies are defined, as well as how social ties are established and retained, there are more concerns with geographically defined labour markets than with labour markets defined in terms of e.g. occupations, even though the costs and difficulty in obtaining the skills needed for entering a new occupation may be as high as those needed to move to a new location.

How to Use the Administrative Registrations?

Faced with the issues sketched above the interesting question is what should be done to ensure that the statistics produced on the basis of PES's administrative registrations will have a known and required minimum quality on each of the quality dimensions. Fortunately, the answers to this question can be found in the established practices and principles for the production of official statistics:

- Have detailed knowledge of the way the observations are being collected;

- Monitor the data collection process to ensure consistent data quality;

- Make every effort possible to improve the data collection process when it is found to be deficient;

- Calibrate the observations generated by the administrative system (through the use of statistical surveys if possible);

- Select carefully the numbers (parameters) to be presented in the statistical tables, and make use of statistical methods, detailed understanding of the process which have generated the observations, and supplementary data to provide estimates which may correspond better to what users need;

- Explain to the user of the statistics their proper use, given the characteristics of the basic data and the methods used.

These points are interrelated and mutually supportive, but they also make clear that the task will not always be easy and may have implications for the organization of the work with e.g. registering job-seekers and vacancies, as well as involve significant costs.

Know the Data Collection Process

The need to know intimately the way the observations have been collected should be obvious, and there isno more reason to expect that actual data collection and processing practices of local employment officials are more consistent with whatever guidelines exists than if this had been a purely statistical data collection process, which conscientious statisticians would monitor closely and seek to modify if it does not function properly. The need to know the details should be even more urgent in the case of a process which, at least partly, has different goals and priorities than those of producing statistics or other forms of labour market information. Fortunately, one of the great advantages of the PES organization as a user of statistics and other forms of summary information based on its own administrative registrations is the closeness of the 'statisticians' of the organization to the actual registration process.

Monitor the Data Collection over Time

In view of the importance of ensuring consistency in time series, it s of course not enough to establish the particulars of the data collection process at one point in time. The way this process develops over time must be monitored. Such monitoring will have to be partly through information generated internally to the process itself, and partly through the calibration process mentioned below. The need to monitor the process may result in a need for information which is additional to that directly used to produce the statistics - for example: when de-registering "unemployed persons" it will be important to know the dates when a person reported that s/he had been given a job or should no longer be considered available for job-offers, came to the employment service to register this and the date when this registration was recorded in the file used for to produce statistics, to be able to monitor how long it takes for all, or a given proportion of, departures from the register of active job-seekers in a defined period to be reflected in the statistics produced. This information is of vital importance for the decision on when to generate estimates of the number of unemployed persons with reference to a certain date and the number of quitters during a reference period, and thus for the timeliness of the statistics.

Help Improve Data Quality

The statistical service in the PES has an obvious obligation to try to help the administrative units to improve the quality of the data which they collect, by bringing both individual errors and more general weaknesses to the attention of those responsible for the registrations and by making suggestions for improvements. However, even within the same agency different units may resent "outside" interference on their turf. Even if suggested changes would be seen to result in improved data quality also from the perspective of those using the information for day-to-day operations, they may not be interested in giving such improvements sufficient priority to actually make the necessary investments and/or changes in rules and/or procedures.

Calibrate the Results

The natural sciences have a long tradition of calibrating a measurement instrument against other instruments capable of measuring the same phenomenon, to ensure that the performance qualities of the instruments are both know and satisfactory. Most social sciences, on which official statistics are primarily based, do not have the same tradition. This is a serious shortcoming which is important to rectify also when PES registrations are used as basis for statistics. Statistical sampling theory and the accumulated knowledge and experience about statistical data collection instruments can form a basis for the necessary development of methodologies for such calibration on a regular basis. The calibration can take two forms, depending on the main objective: (i) Re-visit on a sample basis those job-seekers and employers who are registered to establish whether the information registered is (still) correct; and (ii) draw a general sample of the working age population and employers to investigate how well the job seekers and vacancies registered with PES reflect the total unemployment and vacancy situation, structurally as well as numerically. While the results from both forms of calibration surveys will enhance the users' understanding of the quality of the PES registered-based statistics and labour market information, the primary objective of (i) obviously will be to provide a basis for improving the quality of the PES registration procedures; while the primary objective of (ii) will be to understand how this information reflects (parts of) the labour market.

Use the Registrations as Basis for Estimates

When official statistics are based on statistical sample surveys it is elementary that the numbers presented represent estimates of the population parameters. The basis for the producing the estimates is statistical theory, as well as knowledge about the underlying reality and the way the observations have been generated. The same can be said for the estimated parameters of econometric and similar models about the labour market. Similarly it is recognized that, e.g., the number of LFS-unemployed persons should be estimated from responses to questions concerning job search and availability for work. The concepts of 'employment' and 'unemployment' are such that respondents should not be asked directly whether or not they are "unemployed". It should similarly be recognized that in much of the use of statistics derived from PES administrative registrations they should be used as observations from which one can estimate the parameters which are needed for analysis and description. The estimators to be used need to be based on detailed understanding of the underlying reality which one tries to describe and of the processes which have generated the observations - for example the way and extent to which the rules for qualifying for unemployment benefits influence who will register as well as how they will behave and what information they will give. The results from the calibration studies will be essential information for such estimates. It is therefore also essential that they are repeated regularly as changes in procedure sand behaviours will necessitate re-calibration of the estimated relationships between PES-registrations and 'realities'. Combining PES and LFS based observations may lead to results which would not have been possible otherwise.

Explain the Results.

Those producing a set of statistics have an obvious responsibility to explain the proper understanding the results, and warn against unwarranted use of them. The producers are, after all, be better placed than almost all users to understand the strength and weaknesses both of the underlying data and of the procedures and methods used to arrive at the released statistics. It is not sufficient to point at weaknesses, such as changes in regulations which have caused a break in time-series, and then say that "this should be borne in mind when using the data". This isa cop-out equivalent to saying that the results from a sample survey are subject to imprecision due to sampling (i.e. the so called 'sampling errors'), without saying anything about their size or how they may influence the evaluation of differences between groups or over time. This is, of course, one reason why many releases of statistics on PES-unemployment will include references to LFS-unemployment estimates for the same periods.

Issues directly related to statistics on PES-unemployment

References have been made already to the registration process for persons who may be counted as PES-unemployed. In principle such registrations may refer to the following contacts between the job-seeker (benefit claimant): (1) An initial visit/contact to register as job-seeker (and/or benefit claimant); (2) regular follow-up visits/contacts to check (on the self-service bulletin boards) for possible job offers; (3) information from the job-seeker to the PES about search activities and income earned; (4) information from the PES to the job-seeker about possible job-offerings and referrals to employers; (5) report from job-seeker and/or employer about outcome of job interviews; and (6) report from the job-seeker that (s)he no longer as available for employment, because as (s)he has found a job, or for other reasons such as death, military service, serving a prisons sentence or migration to another region or country.

Descriptions of registration procedures in actual use in national PESs which we have received in the ILO, e.g. for chapter 4 of ILO, 1999, shows that the circumstances of contacts (2)-(6) and how they are registered vary significantly between national PESs. However, these contacts, and in particular contacts of type (6), are crucial in determining whether once registered a particular job-seeker should still be included in the stock of job-seekers and therefore also be included among the PES-unemployed, provided the other conditions for inclusion are satisfied. It seems clear that some PESs do not impose rules for removing registered job-seekers from being counted as 'unemployed' after non-contact for a specified period. This may be one explanation why some PES-unemployment estimates are higher than the corresponding LFS estimates, and why a significant proportion of those registered with the PES may not be considered as LFS-unemployed. That clear and reasonable rules, based on national circumstances, are needed on the 'removal' of registered job-seekers from the count for PES-unemployed is one of the most basic rules for the use of these registrations as basis for statistics which can indicate the level of 'unemployment' in a particular labour market.

Issues directly related to statistics on vacancies

The above concern with the removal of no-longer available job-seekers from the count of PES-unemployed has its obvious parallel in relation to statistics on vacancies based on the registrations in the PES, whether these are limited to those vacancies reported directly to the PES or also include vacancies found advertised in newspapers, journals or on web-sites: What are the mechanisms which will remove no longer available jobs from being included in the statistics? It seems reasonable to assume that most of these reported or advertised vacancies will be closed without the PES being notified. Thus the use of time-related rules for the removal of 'no longer assumed to be open' vacancies from those to be included in the statistics on vacancies, will be even more important than the corresponding rules for PES-unemployment statistics.

In Farm, 2000 it is pointed out that many 'vacancies' are reported to the PES well ahead of the date at which the jobs are to be filled, and that in order for the statistics on 'vacancies' to give a correct picture of 'unfilled job openings' at a particular point in time it will be necessary to register both the date when a notification about a vacancy is received, the first possible date of work for the successful job applicant(s) and the date at which this applicant started to work in the new job. In particular the last information is difficult to obtain in practice, as there is no incentive for the employer or the worker to report this, nor will it be of practical help to the PES placement officer.

While the total PES-unemployment estimates in countries with well developed PES generally are of the same order of magnitude as the LFS-unemployment estimates, it is common to indicate that the 'vacancies' reported to the PES in these countries frequently represent only 15-30 percent of the total number of actual 'vacancies' at any point in time. One of the unexplored issues if this is the case is the consequent validity of studies and econometric models which use the 'balance' between registered 'vacancies' and 'unemployment', or changes to this 'balance', as indicator of (changes to) pressures in the respective labour market. ⁽⁸⁾

Statistics on changes and 'flows'

The demand for 'consistency' in statistics as well as the need for statistics which are timely and have an appropriate frequency is based on the need for reliable statistics on changes, often because only such estimates can provide information on whether policies have the desired impact or whether a situation is improving or deteriorating, whether or not there are any policies at stake ⁽⁹⁾. Relatively speaking the quality of statistics about changes and 'flows' based on PES administrative registrations are often considered a 'better product' relatively to those on the corresponding levels and stocks, in particular when the concern is with short-term changes, e.g. over the business cycle or a budgetary period. There are several reasons for this: (1) LFS-unemployment estimates are particularly imprecise for changes, even with large samples designed to support such estimates; (2) well designed PES statistical programmes can have a high frequency and good timeliness, even for small groups and areas, compared with survey based statistics; and (3) direct estimates of gross flows into PES-unemployment and vacancies can be made, and this means that one is not limited to estimates of net changes as the difference between the stocks at two different reference dates, as one often is with survey based statistics. It is thus a bit surprising that not more attention is given to statistics on the inflow into PES-unemployment, given that such statistics (i) score better on the various quality dimensions than most other statistics based on PES registrations; and (ii) that such statistics should be of obvious relevance as a short-term labour market business cycle indicator for many sectors in an economy.

It is frequently assumed that over the short term most sectors of the economy will tend to move in parallel, and that PES-unemployment and vacancy statistics may provide be good indicators on changes which takes place in the total economy. This is obviously an assumption which deserve careful consideration, where both the weight of the PES clients in the respective labour markets and their behaviour relative to other markets need to be carefully analyzed. The above call for calibration studies is directly relevant in this connection.

PESs which have been able to institute the use of client identifiers, may be able to provide statistics on 'unemployment careers' of persons, not just a statistics on 'cases of unemployment'. This also opens up possibilities for studying the selection to and impact of individual labour market and employment schemes. However, complete impact studies based on administrative registrations will require (i) the use of common 'personal identification numbers (PIN)' by the PES as well as by other administrative bodies and the statistical services, in particular educational authorities and tax or social security administrations; and (ii) that the records of these agencies can be combined for the production of statistics and related research. These requirements are satisfied only in few countries outside the Nordic ones ⁽¹⁰⁾, and is not a realistic option for those in other countries considering how to best use PES-registers as basis for statistics.

Concluding remarks

Statistics and other summary information based on the administrative registrations of the PES can play a very important role in the overall labour market statistics and labour market information of a country. The better developed the PES and the better it is able to integrate in the registrations supporting its primary role concerns with the quality issues of these statistics, the better basis will the resulting statistics give for formulating, implementing, understanding and evaluating labour market, economic and social policies. It is important to note, however, that good quality statistics based on PES administrative registrations are not necessarily 'inexpensive', if all costs in their production are included in the calculations. Thus it is important to carefully evaluate their role and costs both in the context of the role and operations of the PES itself, and in the context of the overall statistical system of a country. In this context the introduction of a wider range of agencies providing services to the labour market and their participants as well as new channels for advertising and searching for vacant posts and candidates will present those responsible for overall labour market statistics and information systems with new challenges: (i) How to obtain reliable and relevant statistics on the operation and impact of the agencies in a cost effective way and avoid that the quality of the statistics based on the PES administrative registrations is undermined without having introduced satisfactory replacements. (ii) How can one effectively use the information acquired through posting of vacancies and availability on the web, and the interactive communication which takes place in connection with the web-sites of the PES and other agencies. In many ways these may be the major challenges for those responsible for planning tomorrow's statistical system for the labour market. It seems likely that some of the quality issues outlined in this paper may become less significant, e.g. those of timeliness and frequency, while others will represent even more difficult challenges to those responsible for planning tomorrow's information system and statistics for the labour market, e.g. those of consistency over time and space. The issue of how to define the appropriate geographic labour markets may take whole new forms for those jobs which can rely on electronic interaction with colleagues, superiors and clients. It seems likely that as sources for labour market statistics the respective roles of administrative registrations, including those reflecting web-site postings and visits, and statistical surveys which solicit information from individuals and organisations, will become more distinct - in the sense that they will provide the basis for very different forms of statistics: the former are likely to serve as basis for indicators of short term changes and 'flows', while the latter will be used to describe the resulting labour market structures and distributions. Those responsible for either form of statistics will face important challenges.

References

EASMAT (1997): Labour statistics based on administrative records: Guidelines on compilation and presentation. ILO Regional Office for Asia and the Pacific. Bangkok.

Farm, A. (2000): Defining and Measuring Vacancies. Swedish Institute for Social research and Statistics Sweden. Stockholm, 2000 (Mimeographed)

Hoffmann, E. (1995): "We must use administrative data for official statistics - but how should we use them?" Statistical Journal of the United Nations ECE, 12, pp. 41-48.

Hoffmann, E. (1996): "Requirements and possible sources for statistics on the dynamics of employment: a producer perspective". Statistical Journal of the United Nations ECE, 13, pp. 335-337.

Hoffmann, E. (1999a): "Collecting statistics on imbalances in the demand for labour". Statistical Journal of the United Nations ECE, 16, pp. 105-121.

Hoffmann, E. (1999b): "Developing labour account estimates: Issues and approaches", Household Accounting: Experiences in the use of concepts and their compilation, Vol. 2: Household satellite extensions. Department for economic and social affairs, Statistics Division. United Nations, New York. 1999.

ILO (1999): Sources and Methods: Labour Statistics. Vol. 9: Transition Countries. International Labour Office, Geneva, 1999.

ILO (2000): Sources and Methods: Labour Statistics. Vol. 4: Employment, Unemployment and Income from Employment: Administrative Records. Second edition. International Labour Office, Geneva, 2000.

Phan, T., E. Hansen & D. Price (2000): Public Employment Service in a Changing Labour Market. International Labour Office, Geneva, 2000.

1. Invited paper for the for the workshop "Labour Market Information as a key function of Public Employment Services" organized by World Association of Public Employment Services (WAPES), 22-24 March 2000, Budapest, Hungary. The views expressed are those of the author and do not necessarily reflect those of WAPES, the other participants at the Workshop, nor of the ILO or its Bureau of Statistics. The author apologizes for all errors and omissions, and would welcome comments and suggestions for improvements and correction. Address: CH-1211 GENEVE 22, Switzerland; e-mail: hoffmann@ilo.org

2. Note that this paper throughout will refer to the Public Employment Service (PES). However, most of the points made with respect to the operation of the PES and the use of its records as basis for statistics apply also to the operations and records of private employment agencies.

3. Among such costs are: possible loss of self-respect and prestige, travel costs and time lost from other activities when (re-)registering with the PES.

4. Not considered here is the situation where those producing statistics at the central office can access directly individual records as soon as they are registered locally.

5. One relevant example is information about 'occupation' in schemes for the administration of migrant workers, e.g. where the rules specify minimum wages by occupation, or restrict the recruitment to certain occupations. Thus the Philippine restrictions on the recruitment of "domestic helpers" is reputed to have resulted in a sharp increase in the recruitment of "governesses" and female "gardeners" and "drivers".

6. The severity of this problem may of course differ between local offices and therefore also be a source of spatial inconsistency in the reported statistics.

7. The wish to ensure the confidentiality of information about e.g. individual employers may in certain cases lead to more aggregate tabulations than some users would prefer.

8. The formulation "if this is the case" is motivated by the fact that there exist very few examples of statistics on 'vacancies' which are not based on PES records. See Hoffmann,1999a and Farm, 2000, as well as references given there, for further discussion and examples.

9. Note, however, that also those concerned with changes need information on the level and actual situation, in order to evaluate whether observed changes are significant or not.

10. It should be noted, however, that the experience from the Nordic countries clearly demonstrate that linking information about individuals from the registers of different administrative agencies and from the statistical office with the use of PINs is complicated and very resource demanding. Thus statistics from the resulting 'integrated files' are still better suited for research-projects than for the production of current statistics.