Training, quai André Citroën, PARIS Cedex 15, FRANCE

Transcription

1 Job vacancy statistics in France: a new approach since the end of Analysis of the response behaviour of surveyed firms after change in questionnaire Julien Loquet 1, Florian Lézec 1 1 Directorate for Research, Studies and Statistics of the French Ministry of Labour-Employment- Training, quai André Citroën, PARIS Cedex 15, FRANCE Disclaimer This report is released to inform interested parties of (ongoing) research and to encourage discussion (of work in progress). The views expressed on (statistical, methodological, technical, or operational) issues are those of the authors and not necessarily those of the Directorate for Research, Studies and Statistics of the French Ministry of Labour-Employment-Training. Abstract Since 2010, according to European regulations, member states of European Union have to send to Eurostat quarterly estimations of job vacancies 45 and 70 days after the end of the estimated quarter. After a change in questionnaire between the third quarter of 2010 and the fourth quarter of 2010, the French rate of job vacancies evolved from 0,4 % to 0,6 %, that is to say a sharp increase of 60 %. Draftings and publications of statistics of job vacancies are organized under the rules of regulations of the European Parliament and of the Council of April According to the Eurostat definition, job vacancies are: newly created paid posts; unoccupied paid posts; occupied and paid posts about to become vacant, for which employers intend to fill either immediately or within a specified period of time and are taking active steps and are prepared to take more steps to find a suitable candidate from outside the firm concerned. These posts could be temporary or permanent contracts, seasonal jobs, even short-term contracts. Posts open to internal candidates in firms are not considered vacant according to Eurostat. In France, the Directorate for Research, Studies and Statistics, the statistical service of the Ministry of Labour-Employment-Training, is providing these job vacancies estimators for firms of 10 or more employees of the private sectors. In this paper, we characterize and quantify the increase in the number of job vacancies in regard of the change of the questionnaire between the third and the fourth quarter of The analysis is based on the evolution of the response behavior of firms to job vacancy question before and after the change in questionnaire. In reply to this, several analysis are set up. First, the business cycle is separated from the change in questionnaire and quantified thanks to econometric modelling using related indicators to job vacancies (hires, recruitment difficulties, offers and demands on labour market). Then, the change in questionnaire is analysed by measuring intensive and extensive margin effect (some firms have declared more job vacancies, others have begun to declare job vacancies since the change in questionnaire). Each analysis is completed by characterization of firms in term of Nace sectors (European nomenclature of activities), size and types of subcategories of job vacancies declared. Key words: job vacancies, panel sample of firms, intensive and extensive margin effect

2 1. Characterisation and quantification of the increase in the number of job vacancies in regard of the change of the questionnaire Since the second quarter of 2003, France has been running stocks of job vacancies, occupied posts by sectors and firms sizes under the quarterly survey named Acemo (as Activity and employment conditions of labour). Acemo has conducted since It provides current indicators about the evolutions of wages, working time and employment. Acemo survey answers European and government concerns (for example, minimum wage in France is indexed with hourly base wage for labour force estimated from Acemo). The field includes firms of 10 or more employees in competitive sectors from Metropolitan France, excluding temporary workers and trainees. The survey excludes workers of agriculture and public sectors. Every unit above 250 employees is surveyed every quarter. Units under 250 employees are selected using a stratified survey approach: the elementary stratum is defined by crossing the economic sector (under French nomenclature of activities in 88 positions) and the size of the firm (on 6 positions). For units under 250 employees, sample rates are produced with a constraint Neyman based on sizes of sample stratums so that the weak accuracy of estimators (including job vacancies) in published stratums has to be below a level of 5 %. This constraint in publication stratums fixes a minimal number of units in sample stratums. The sample is valid for 12 months. As a part of the stratum is exhaustive for firms of more than 250 employees, the sample is never completely renewed. Indicators produced by the survey are of two kinds: level and variation so that the survey requires two consecutive answers from a firm. The stratums with less than 250 employees rotate only for 25 % each year and are completely renewed over four years. Acemo survey is sent by mail every quarter to units representing 12 millions of the 23 millions of all employees in the total French economy. Each quarter, the global response rate is stable at around 80 %. For the question of number of job vacancies, it is always above 70 % every quarter. The questionnaire Acemo has included a question about job vacancies since the second quarter of 2003 in that form: Figure 1: Previous translation of job vacancy question Since the fourth quarter of 2010 in order to better match the definition of job vacancies recommended by Eurostat, the question has been changed. A think tank has been set up to clarify the concept in order to better answer European obligations. Indeed, measurement of job vacancies is very different between European countries. In France, the think tank advised to use the periphrasis "posts for which you are taking active steps to find a suitable candidate" to replace the terms job vacancies or unfilled jobs. Additionally, a new question is adjoined on the breakdown of these posts depending if they are newly created, unoccupied or about to be vacant. On the other hand the question about recruitment difficulties has been removed.

3 Figure 2: Current translation of job vacancy question This new question has appeared on questionnaires sent since the beginning of 2011 ; the first results released relate to the fourth quarter of As a result of the new questionnaire, estimation of job vacancies sharply increased (+60%) between the third quarter of 2010 and the fourth quarter of While all other related short-term indicators were stable at that period (or even decreased), job vacancies indicator increased by 30% the quarter after (2011Q1), and has stabilized ever since. Given the impact of this change, France specially demanded Eurostat not to publish stocks (only job vacancy rates are published by Eurostat) while waiting for stabilization of the results. Figure 3: Total number of job vacancies (2008Q4-2011Q3, base 100 = 2009Q3) This increase does not seem to reflect economic short-term reality but appears to be related to another way firms understand the question and answer it. More particularly, the new breakdown by three types of posts considered as vacant could explain this change. The fact that respondent units didn't declare a certain type of posts before the change in questionnaire is an assumption to make. On initial examination, we can suppose that firms that answered job vacancies question on previous questionnaire and did not take time to read the explanatory paragraph below, considered that job vacancies meant existent unfilled posts (posts that are hard to be filled). After the change in questionnaire, these firms better understood the concept and began to declare other job vacancies:

4 newly created posts and posts about to be vacant. Indeed, these two new types of job vacancies represent 60% of declared job vacancies. In detail, by Nace sectors, the number of job vacancies sharply increases almost in all sectors. Specially in the sector named other service activities (recreational activities, R sector), where a lot of seasonal workers are employed, which can explain why firms of this sector often need workers for fixed-term period posts (counted as vacant), and in financial and insurance activities (K sector) which is a sector with high turnover of the labour force. Figure 4: Evolution of the number of job vacancies by Nace sectors The number of job vacancies in the other sectors of industry (mining and quarrying, water supply, electricity, gas, etc.) did not increase. These sectors have recruitment difficulties (more than 75% of the firms are concerned at 2010Q3). In the sectors, we can suppose that firms declared only hard to be filled job vacancies (whose number remains constant quarter after quarter). By firm size, the distribution of job vacancies was unchanged (except for firms of 500 or more employees). Figure 5: Evolution of the number of job vacancies by firm size

5 After the change in questionnaire, on the field of firms of 50 employees and more, the bigger the firm size the lower the rate of job vacancies was. A high rate of job vacancies in small firms confirms a result demonstrated in other firm surveys: the turnover is high in small firms. However, answers from firms of 250 or more employees contribute for more than 30% of the number of job vacancies increase. In fact, because they are bigger these firms should need a larger volume of job vacancies. That's why they contribute more sharply to this increase. By sub categories of job vacancies introduced with the new questionnaire, firms have declared that approximately 40% of their job vacancies were unoccupied posts, the same share were newly created posts, 20% were posts about to be vacant. This distribution has been constant since it was introduced in questionnaire, which is showing the structural characteristic of firms's answers., in % Figure 6: Proportion of the sub categories of job vacancies If some firms know how to count each of the three sub categories of job vacancies, other units can just declare the total of job vacancies and not the sub categories. These firms answer generally more jobs vacancies than others, this phenomenon is specially significant in firms of 100 and more employees. Table 1: Answers of units to job vacancy question 2010Q4 2011Q1 2011Q2 2011Q3 Average number of job vacancies per unit 6,0 6,8 6,7 6,5 Rate of response without detail 6,30% 2,60% 2,30% 1,40% 2. Analysis of response behaviour of firms to job vacancy question After these first hypothesis which could explain and measure the sharp increase in job vacancies between the third and the fourth quarter of 2010, two different goals are achieved in the paper: - first, we isolate the effect due to the change in response behaviour of surveyed firms from the business cycle ;

6 - then we explain and measure the sharp increase of job vacancies (+60%) observed between the third and the fourth quarter of 2010 (intensive and extensive margin effects). Thanks to a panel sample of several thousand surveyed firms, we carry out three different analysis. The panel sample expands from the fourth quarter of 2008 to the third quarter of 2011 (12 consecutive quarters). In the first analysis, we isolate the specific effects of change in the question from business cycle by estimating the level of job vacancies at the fourth quarter of 2010 we should have observed without change in the question on job vacancies. In the second analysis, after highlighting the business cycle effect, we measure the intensive margin effect due to change in the question on job vacancies and characterize it. The intensive margin effect represents the firms which already declared job vacancies before the change in questionnaire and whose number of job vacancies has increased since the change in questionnaire. We also analyse what type of job vacancies (unoccupied posts, newly created posts and posts about to be vacant), these firms declare. In the third and final analysis, we measure the extensive margin effect due to change in questionnaire and analyse the characteristics of these new respondent firms. The extensive margin effect represents the new respondent firms to the question on job vacancies which were already surveyed but which did not respond to job vacancies question. We measure the increase in the number of job vacancies due to extensive margin effect. We also point out the characteristics of these new respondent firms: the distribution of job vacancies between the three types of posts, the sizes of the firms and their sectors. 2.1 Micro econometric modelling The first analysis separates the effects of change in the questionnaire and business cycle (micro econometric modelling). A panel sample of 2053 firms surveyed and respondent to Acemo survey for 12 consecutive quarters (the fourth quarter of 2008 to the third quarter of 2011) was used. All of these firms answered or not the job vacancies question. We estimate the number of job vacancies at the fourth quarter of 2010 for these firms in the event of a lack of change in questionnaire. We isolate the impact of the business cycle. The econometric modelling uses different variables related to job vacancies (available in other statistics data) for each firms: - the change in the number of employees - the change in the number of hires The equation: Vacancies where V e i, T = 0 + α1 ( employeesi, T employeesi, T 1) + α 2 *( hiresi, T hiresi, T 1 ) Vacancies, is the number of job vacancies of the firm i at the quarter T i T employees, is the number of employees of the firm at the quarter T i T α * + ε hires i, T is the number of hires of the firm i at the quarter T ε i,t is the residual, unexplained by the equation i, T α 1,α2 are between-time-period estimators (based on data from the fourth quarter of 2008 to the third quarter of 2010) In order to obtain a significant modelling of the evolution of job vacancies, we use between-timeperiod estimators. Different modelling were tested: - between-time-period estimators - between-group estimators - one and two way fixed-effects model

7 The between time periods were the only consistent estimators, it means that some factors affect all of the firms of the panel (identically) and these factors are due to time period. The modelling is based on quarterly data from the fourth quarter of 2008 to the third quarter of The fourth quarter of 2010 is predicted. So for the fourth quarter of 2010, we have: Vacancies = 16,72+ 0,04*( employeesi,2010q4 employeesi,2010q3) + 0,02*( hiresi,2010q4 hiresi,2010 3) i, 2010Q4 Q The result is that in the event of a lack of change in questionnaire, the number of job vacancies would have decreased by 6% at 2010Q4. With separating the effects of change in the questionnaire and business cycle (with micro econometric modelling), we conclude that business cycle does not have any impact on the increase of job vacancies between the third quarter of 2010 and the fourth quarter of Macro econometric modelling In order to confirm the result obtained with micro econometric modelling, we also check what would have been the evolution of job vacancies without change in questionnaire, based on other sources. A macro econometric model is estimated on a longer period of time (the second quarter of 2003 to the third quarter of 2010). The total number of job vacancies is estimated from different related indicators (hires, job offers, intentions to hire, recruitment difficulties, ). Several of these variables are not available for each of our firms of the panel sample that is why we do not use them for micro econometric modelling. An autoregressive model is used to perform forecasting of the fourth quarter of The outcome is almost similar to the micro econometric result. According to the macro econometric modelling, the number of job vacancies would have stabilized between the third quarter of 2010 and the fourth quarter of 2010 without change in questionnaire Evolution of the total number of job vacancies and forecasts in the event of a lack of change in questionnaire Observed job vacancies Predicted job vacancies 2003Q2 2003Q4 2004Q2 2004Q4 2005Q2 2005Q4 2006Q2 2006Q4 2007Q2 2007Q4 2008Q2 2008Q4 2009Q2 2009Q4 2010Q2 2010Q4 2011Q2 Figure 7: Macro econometric modelling of job vacancies based on related economic indicators

8 Once again, we can conclude that business cycle does not have any impact on the increase of job vacancies observed between 2010Q3 and 2010Q Intensive margin effect In order to measure the effect of firms which have declared more job vacancies than before the change in questionnaire (intensive margin effect), we use a panel sample of firms that at least declared job vacancies at both the third quarter of 2010 and the fourth quarter of The panel sample is composed of firms surveyed until at least the third quarter of 2010 (the oldest firms in the panel have been surveyed since the fourth quarter of 2008). The result: 17% in the increase of job vacancies is explained by the intensive margin effect. Firms concerned by the intensive margin effect declared that 51 % of their job vacancies are unoccupied posts, 33 % are newly created posts and 16 % are posts about to be vacant. More than a half of these firms had recruitment difficulties (64%). This suggests that before the change in questionnaire 1, firms only declared unoccupied posts for vacancies (the most important part of declared job vacancies), which means posts for which the firms have difficulties to find an applicant (=recruitment difficulties). Since the change in questionnaire, these firms have declared other vacancies that are part of the usual turnover of a firm and increase the total declared. 2.4 Extensive margin effect An other part of the job vacancies increase could be due to firms that declare job vacancies after the change in questionnaire, that is to say the extensive margin effect. In order to measure this effect, we use a panel sample of firms that at least answered «no» to job vacancies question at the third quarter of 2010 and «yes» at the fourth quarter of The result: 68% of the increase of job vacancies is explained by the extensive margin effect. Firms concerned by the extensive margin effect declared that 53 % of their job vacancies are newly created posts, 20 % are unoccupied posts and 27 % are posts about to be vacant. More than 75% of these firms have less than 250 employees and these firms are mainly in industry, wholesale and retail trade and professional, scientific and technical activities, administrative and support service activities. All of them created jobs at the fourth quarter of 2010, we supposed that newly created job vacancies became new filled posts. The important thing is: why didn t the firms declare job vacancies before the change? As we said before, we assume that the question was clearer after the change. Indeed, before the change in questionnaire, newly created posts appeared in a small paragraph in italic characters, while firms have now to declare it in a specific box (see the change in the question, supra for more details). We also analyse the extensive margin effect in another phenomenon. In Acemo survey, the sample is partially renewed once a year (25 % of the sample) at the fourth quarter of each year. So, in order to measure the extensive margin effect on the new surveyed firms (of the renewal of the sample), we compare the proportion of job vacancies declared by new respondent firms between the renewal of the sample at the fourth quarter of 2010 and at the fourth quarter of 2009 or at the fourth quarter of At the fourth quarter of 2010, the new respondent declared 12 % more job vacancies than at the fourth quarter of 2009 or at the fourth quarter of And again, these firms declared that more than 50 % of their job vacancies were newly created posts. 1 Before the change in questionnaire, the three sub categories of job vacancies were not available.

9 In conclusion, the increase observed in the number of job vacancies between the third quarter of 2010 and the fourth quarter of 2010 (+60%) can be explained by: - intensive margin effect (17% of the increase); - extensive margin effect of firms already in the sample (68% of the increase); - extensive margin effect of new surveyed firms (12% of the increase); - residual (3% of the increase). Increase of job vacancies = intensive effect (17%) + extensive effect (80%) + residual (3%) 3. Conclusion and opening: Job vacancies in France, the difficulty to measure two different concepts In France, administration decided to translate the English Eurostat definition into a French definition which is a bit simple. The Eurostat definition of job vacancies is broad and general and it covers two types of job vacancies: - Job offers by firm that are quickly filled (in a period of time under 3 months for example). These offers cover different types of contracts like permanent or fixed-term contracts It represents the usual turnover of employees in firms. - Unfilled jobs which are unfilled for a long time (more than 6 months for example) because the firm can not find a suitable applicant. It represents the recruitment difficulties of the firm. In Acemo survey, hard to be filled job offers seem to be correctly measured. The employer does not have difficulties to count them normally. They are not so numerous by firm, contrary to job offers which are quickly filled. For firms with a high turnover, which hire almost only short-term contracts (with a period that is under a month for example), it may be difficult for the employer to count all of them. The important point is: what is being measured? To job offers relating to turnover, we should oppose short term unemployment (and measure the two concepts on a same basis: administrative data, indeed these two concepts seem to be non-measurable with a survey). To job offers hard to be filled which are synonym of recruitment difficulties for a firm, we should oppose long-term unemployment and again measure the two concepts on a same basis.