Inebase / Economically Active Population Survey / Re-estimate of the 1976-2000 unemployment series according to the 2002 APS definition (extract from the working document)

As established by the Agreement passed by the Executive Committee of the Government for Economic Issues on the improvement of transparency in the economic and statistical information provided by the Government, the National Statistics Institute will undertake the historical combination of all the series whose methodological base has changed, in order to preserve historical information.

The combination presented in this document refers to the main unemployment series obtained from the Active Population Survey. The definition of this variable was modified in 2001-2002. It encompasses data from the third quarter of 1976 to the fourth quarter of 2000 and considers the main aggregates (unemployment by sex and persons under 25 years old and aged 25 years old and over) on a national level and by autonomous communities.

The operative definition of unemployment used until then in the APS was modified in 2002 so as to adapt it to European regulations.

In favour of a greater harmonisation of employment and unemployment figures in the EU, Commission Regulation (EC) 1897/2000 of September 7th 2000 establishes (alongside other requirements related to the order of the questions, treatment of special groups, etc.) practical rules for the specific implementation of the conditions that govern when a person is considered unemployed in the European Union. These conditions differ slightly from those considered up until the year 2002. Specifically, according to the terms envisaged in Annex I point 1 of the Regulation, the only methods established for seeking employment are the following:

As regards Spain, the condition establishing that contact with the job centre must be specifically to find employment, without considering inscriptions due to strictly administrative reasons as an active method for seeking employment, has had a considerable impact on the figures and has had noticeable consequences in terms of the unemployment level resulting from the APS.

The questionnaire was modified throughout 2001 and information was collected after performing field work ensuring it was possible to calculate unemployment figures considering the operative definition envisaged in Reg. 1897/2000 and previous. Data for this year were used as the link between the old and the new definition.

The questionnaires used before 2001 did not include the notion of whether contact with the job centre was carried out to find employment or due to other reasons, and therefore it is not possible to determine the unemployment figures that would have resulted in the APS if this new definition were to have been used for periods prior to 2001 using the variables of the survey. On the contrary, it is possible to obtain the unemployment figure for 2001-2004 both considering the old and the new (official) definition. The information referring to 2001-2004 has been considered as the base to recalculate the main unemployment series for 1976-2000, using the mathematical procedure described below.

Persons who were considered unemployed according to the old definition but not according to the new one fall into the category of inactive persons, since they do not fulfil the operative condition established as 'actively seeking' employment. Consequently, this change of definition reduces the number of unemployed persons, the rate of unemployment and the rate of activity.

After testing several procedures, a regression with binary response variable was selected implementing a lineal trend in the reduction of the effect towards the past (see the full report for a more detailed reference to the different alternatives contemplated to recompile retrospective series using the new unemployment definition).

Available base information

he data used to estimate the probability of a person remaining in the classification of unemployed persons according to the old and new definition were the inscriptions of unemployed persons according to the old definition regarding the population aged 16 years old and over for the period between the first quarter of 2001 and the fourth quarter of 2004 (period for which there is information for both definitions).

After obtaining these probabilities, they were applied to the historical unemployment series in order to determine an estimate of the number of unemployed persons according to the definition of unemployment implemented in the year 2002. Since both the methodology and the compilation of the APS and different factors that influence the Spanish labour market and other social aspects and economic situations have changed over time, it is considered that the effect of this new definition will disappear gradually after 2001.

Description of the procedure

The study has been performed breaking down data by sex and autonomous community. The national total has been obtained after aggregating all figures. Nevertheless, it is important to note that the results of the estimate for the series of unemployed persons obtained by aggregating Autonomous Communities and those obtained by applying the method directly to the national total provide similar data.

After obtaining the transition probability for a person to be considered unemployed according to the old definition or according to the new one, via a probit regression, the next step was the adjustment of a lineal trend, so that for the first observation (second quarter of 1976) this probability amounted to one and for the last (fourth quarter of 2002) it equalled the probit obtained.

Another option that was considered was to take into account the possible seasonal variations and the analysis of the adjustment of the trend in terms of the quarter of the year the observations refer to, thus obtaining a different probability for each of the four quarters. Nevertheless, if the data is different for each quarter, the sample of the APS that provides basic data would be very small in communities that had the lowest populations, and this would affect the estimates of the parameters. On the other hand, no major differences would appear in large communities, compared to the data obtained omitting seasonal variations. Therefore, it was established that a single common probit would be calculated for all four quarters.

In summary, the procedure implemented consists of the following stages:

The calculation of the transition probability for a person to be considered unemployed according to the old definition or according to the new one using a probit regression model, considering a binary variable that measures whether the person is unemployed according to both definitions as the endogenous variate, and the sex, age group (under 25 or 25 years old and older) and interaction between both factors, weighted by the raising factor, as the exogenous variates.
After obtaining these transition probabilities, a lineal trend is applied. Therefore, in the third quarter of year 1976 the effect of this change of definition has disappeared totally. This provides adjusted probabilities.
Implementation of adjusted probabilities to the historical unemployment series.

The different models calculated using the aforementioned procedures, the probability tables for permanence in an unemployed situation according to the 2002 definition and total unemployment results by sex and distinction between persons under 25 years old or aged 25 years old or over, by autonomous communities, can be viewed in the complete working document.