what does it take to be (counted as) unemployed? the case...

25
What does it take to be (counted as) unemployed? The case of Spain Luis Garrido (UNED and CESC*) Luis Toharia (Universidad de Alcalá and CESC*) February 2003 Submitted to the 2003 EALE Conference, Sevilla, September FIRST DRAFT (*) Centro de Estructuras Sociales Comparadas, Universidad Nacional de Educación a Distancia

Upload: others

Post on 10-Jul-2020

6 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

What does it take to be (counted as)

unemployed? The case of Spain

Luis Garrido (UNED and CESC*)

Luis Toharia (Universidad de Alcalá and CESC*)

February 2003

Submitted to the 2003 EALE Conference, Sevilla, September

FIRST DRAFT

(*) Centro de Estructuras Sociales Comparadas, Universidad Nacional de Educación a Distancia

Page 2: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

ABSTRACT

This paper analyzes the effects of the new European Commission regulation

1897/2000 which establishes a new definition of unemployment for the purposes of

Labour Force Surveys. The paper first examines the conditions that unemployed people

have to meet in order to be excluded by the new notion, i.e. being “passive job seekers”,

to turn then to an application to the case of Spain, a country where the new regulation

was expected to have strong implications. Various characteristics of workers are related

to the process of exclusion, but the most significant result is that there is a wide regional

dispersion of exclusion rates, which also seems to vary significantly over time. The

conclusion is that determining who are the “true” unemployed may be more difficult

than the regulation hoped for.

The second issue, however, is to what extent the very nature of the regulation is

warranted, i.e. whether passive job seekers deserve to be excluded from unemployment

because their labour market behaviour is closer to that of people outside the labour

force. The conclusion here is mixed: in terms of their probability of finding a job within

one quarter, passive job seekers are in between active job seekers (the “true”

unemployed) and non-seekers (the inactive). Why then should they be counted in either

group? Given the difficulties to isolate them clearly in labour force surveys, and given

that passive job-seeking seems to be quite unstable over time, it might be wiser to leave

them within unemployment. Defining them as inactive, as the new EC regulation does,

does not appear to be justified in any way.

Page 3: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

1

1 Introduction Being unemployed is a situation which, in principle, is easy to understand. An

unemployed person is someone who is willing to work but is unable to find a job. Of

course, this general statement may be qualified in various ways, for example, by

imposing some condition about being willing to work “at the going wage”, as

economists usually do, or about the way in which job search is carried out.

Labour market statisticians have long established a standard definition of

unemployment (adopted at ILO Conferences) whereby an unemployed person has to

meet three conditions to be considered, and hence being counted as, unemployed:

- not having worked in the past, “reference”, week, meaning not even one hour

and no matter whether for pay or as an independent worker (including unpaid

work in a family business; voluntary unpaid activities are not considered as

work);

- being “available” for work, i.e. being in a position to be able to start a possible

job offered within a fortnight;

- having performed “active” job search, usually meaning that the person can

mention at least one way in which s/he has looked for a job in the previous four

weeks, from a list of methods defined as “active” (precise steps taken; as

opposed to “passive”, e.g. waiting to be called).

Many countries around the world, including, to be sure, the United States,

Canada and the European Union, have adopted this definition as their guideline to

measure unemployment. Most of them, as well, use sample household surveys to

determine such figures. However, the precise way in which the concepts above are

translated into precise questions aimed at extracting the correct information from

individuals varies from country to country. The most difficult question refers to job

search. The usual recommendation (for example, by EU regulations) is that interviewers

read a list of methods at least until three of those considered “active” have been read.

But the practice seems to be varied and no unique procedure is followed.

In September 2000, the European Union passed a new regulation concerning the

operational definition of unemployment. In simple terms, this regulation defined active

search in a more stringent way, especially as regards contact of the unemployed with the

public employment services. Although no official statement of the results of

Page 4: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

2

implementing this new regulation exists, there is at least one country, Spain, where its

effects on the unemployment figures have been devastating1. As a matter of fact, many

experts and commentators tend to believe that the Spanish unemployment figures were

directly aimed at by the new regulation2. In the first quarter of 2001, the Spanish

Statistical Office amended the LFS questionnaire so that the new definition of

unemployment could be implemented. Starting in the first quarter of 2002, together with

two other methodological changes (new population figures and a new weighting

scheme, both of which increase employment and unemployment), the new definition

has officially been adopted.

Against this background, the purpose of this paper is to investigate to what

extent the changes in the definition of unemployment suggested in the new regulation

make sense from an economic and sociological viewpoint. The Spanish case is used as

an illustration. Thus, after presenting the new definition of unemployment in more

detail in Section 2, showing that the new notion of unemployment is a subset of the

concept used before, Section 3 turns to quantifying the reduction of unemployment

implied by it in the Spanish case, for which information is available for the eight

quarters of 2001 and 2002. The reduction is presented not only in terms of the general

evolution but also in terms of the characteristics of the unemployed.

Section 4 then turns to the issue of whether the elimination of a significant group

of workers from the rosters of the unemployed is meaningful. The analysis here is made

on a double front: first, a conceptual discussion is presented on the role played by active

job search and the recourse to public employment services as a job search method;

secondly, an analysis of the employment behaviour of those excluded from the

unemployment definition is presented, in terms of their probability of leaving

unemployment and getting a job within one quarter. The main conclusion is that, on

both counts, one may argue that these people are different from the unemployed as

newly defined; however, they are also different from people counted as inactive or “not

1 At a Seminar which took place in October 2000 at the Universitat Pompeu Fabra of Barcelona, one of the authors made an early estimation of the potential impacts of the new definition. See Toharia (2000). For the point of view of an international observatory on the debate surrounding the new figures, see EIRO (2002).

2 This is of course impossible to document and is based on casual observation by various people involved in statistical and policy-related matters, including the authors. For example, at one point, Eurostat tried to change the definition of unemployment used in the ECHP leaving “unclassified” half a million people in Spain. This change was dropped later on, however.

Page 5: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

3

in the labour force”. On the whole, therefore, excluding them from unemployment is not

fully grounded on economic and sociological analysis. As discussed in the concluding

section, unemployment is a fuzzy category in which one can find people fully

committed to labour market activities and people less so committed.

2 The new definition of unemployment Regulation 1897/2000 of the European Commission defines unemployment in a

pretty standard way, following the classical ILO conventions. Thus, an unemployed

person is one without work during the reference week, currently available for work and

actively seeking work during the four week period ending with the reference week. The

new twist of the definition is the “specific steps” which are considered proofs of active

job search. Among them, and most significantly, the precise wording of the relationship

between the jobseeker and the public employment offices3. While theretofore

registration with one such office would be considered sufficient proof of active job

search, from now on only a contact “to find work” will be considered a specific step of

active job seeking, i.e. a visit to the employment office with the purpose of merely

renewing administrative registration will no longer be interpreted as an active job search

method.

From a purely abstract point of view, there is nothing to question in this stance

adopted by the European Commission. It is true that a person who merely visits a public

employment office for the purpose of renewing an administrative registration which

may provide other benefits in terms of access to various services or courses cannot be

taken to be actively seeking work. On the other hand, one may wonder whether this

person can reasonably be equated to other who do not want to work, and openly declare

it to be so. In this respect, it should be mentioned that all of the labour force survey

questionnares used ask whether the person is seeking work or not before asking how

this search is being done. So what the new regulation is doing is sending to the pool of

the “inactive” population all those who, despite declaring initially that they are seeking

work, cannot prove it because their contact with the public employment offices is

“passive”. We shall come back to this question in Section 4.

3 The new regulation includes other minor changes which we omit here for reasons of space.

Their influence in the unemployment figures is negligible.

Page 6: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

4

What is to be stressed here is that the new unemployed group is a subset of the

old group. This is very convenient, as it allows for an analysis in terms of “exclusions”

from the old notion to reach the new one, as shall be presented in Section 3. Starting

with the old definition of unemployment, it is worth understanding how the unemployed

pool is precisely determined. Following the EC regulations4, questions on employment

must precede any question on job search. So LFS questionnaires first determine whether

the interviewee may be classified as employed. If not, then the basic job search question

follows. Those providing a positive answer are then asked about search methods and

availability (the order here being irrelevant). In general, a large list of possible search

methods is read to respondents and no restriction applies as to the number of positive

answers given5. Those able to mention at least one search method and declaring that

they are available for work are then classified as unemployed. The rest is inactive

population. In general, initial jobseekers excluded from unemployment represent less

than 5 percent.

As explained above, the new regulation distinguishes between active and passive

job search methods. The procedure to determine unemployment is exactly the same as

described in the preceding paragraph, with the only caveat that now only those

mentioning at least one active method are considered unmeployed. It should be obvious,

however, that the probability of being excluded heavily depends on the number of

methods mentioned in the first place. Thus, if one person declares two methods, s/he is

counted as unemployed under the old definition. If one of them turns out to be

considered passive, s/he is still counted as unemployed under the new definition.

However, if one person mentions only one method, s/he is counted as undemployed

under the old definition but should this method turn out to be passive, s/he should be

excluded under the new one.

As already mentioned, out of the methods which may turn out to be considered

passive under the new regulation, the most significant one is the registration with a

public employment office. For simplicity6, let us consider this method as the only one

which might turn out to become passive (should the conditions regarding active contact

4 This point is again mentioned in Regulation 1897/2000. 5 As a matter of fact, Regulation 1897/2000 states that “job search methods are enumerated until

at least three active methods have been mentioned”. 6 And as a matter of empirical realism, as the other methods considered passive under the new

regulation are of negligible significance in practice.

Page 7: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

5

not be met). This implies that the only people counted as unemployed under the old

definition who can be excluded are those who only mention as search method

registration in a public employment office.

Under these assumptions, an exclusion rate X may be defined as follows:

X = α . β + ε [1]

Where X is the number of those excluded under the new definition as a

proportion of old unemployment, α is the proportion of old unemployed who only

declare registration at the PES as search method, β is the proportion of the latter who are

excluded because their contact with the employment office is considered passive, and ε

is the residual of other exclusions related to other changes included in the regulation (as

a proportion of olde unemployment). In the next section we present the results of

applying this framework to the Spanish case.

3 The consequences of the new definition: “old” versus “new” unemployment

In Section 2, we have argued that the implementation of the new unemployment

figures to be calculated under EC Regulation 1897/2000 may be understood as a process

of exclusion from the unemployment figures defined in the old way. Furthermore, we

have argued that the “exclusion rate” may be decomposed in two main factors plus a

residual. In this section, we apply this analytical framework to the Spanish data. We use

the eight quarters (from 2001.Q1 to 2002.Q4) for which information can be obtained

using both definitions, as the Spanish Statistical Office modified its questionnaire in the

first quarter of 2001 so that the new unemployment definition could be used. This

section is divided into three subsections. First, we present the general data on exclusions

and on the factors behind it; we observe a clear drop in the exclusion rate in 2002, when

the new definition was officially used, so that two general periods may de distinguished

for further analysis, each one corresponding to one full year. Next, we present

disaggregated information on the exclusion rate, trying to see whether there are specific

characteristics which make one person more likely to ble excludable, i.e. to behave in a

a less active job search manner. Specific attention is made to the regional dimension

which turns out to be so important that s full subsection is devoted to it. In the final

subsection, we merge all the preceding information in a series of logistic regressions of

the probability of being excluded.

Page 8: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

6

3.1 General exclusions

Figure 1 plots the number of unemployed under the old definition in Spain since

the first quarter of 2000; starting from 2001.Q1, the new unemployment figure is also

provided as is also the “exclusion rate” (right scale, bars)7. Old unemployment and new

unemployment followed a similar path during 2001, as the proportion of exclusions

remained more or less stable at around 20%. Then, in the first quarter of 2002, the rate

of exclusion fell under 15%, so that unemployment growth was exaggerated by the new

definition. Both series evolved in parallel until the last quarter when the proportion of

those excluded rose again to well over 16%, so that new unemployment understated the

actual growth in the numbers of those excluded.

Figure 1. Unemployment in Spain (thousands) under the old and the new definitions, and “exclusion rate” (Source: LFS microdata files made available by INE)

1300

1400

1500

1600

1700

1800

1900

2000

2100

2200

2300

2400

2500

2600

2700

2000Q1 2000Q2 2000Q3 2000Q4 2001Q1 2001Q2 2001Q3 2001Q4 2002Q1 2002Q2 2002Q3 2002Q4

(thou

sand

s)

0

10

20

30

40

50

60

70

(per

cent

ages

of o

ld u

nem

ploy

ed)

% excludedUnemployed-OldUnemployed-New

Figure 2 presents the breakdown of the exclusion rate by its three components as

identified in expression [1]. First, the proportion of the old unemployed who only

mention registration at a public employment office as job search method (α in

7 The 2000 data are provided to show that the old unemployment figures fits perfectly with the

figures for earlier quarters. Also, it should be mentioned that the unemployment figures provided have been calculated using the new weighting scheme which uses newer, higher population figures and also corrects, at the regional level, for the unbalanced age structure of population stemming from the LFS sample. Both of these changes tend to increase the employment as well as the unemployment numbers.

Page 9: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

7

expression [1]) remained at a level of around 30% during 2001 and then dropped to a

figure around 25% for the following year. This was the main factor behind the drop of

the esclusion rate in the first quarter of 2001. Simple arithmetics shows that some 80%

of the total variation of the exclusion rate is explained by this drop.

Figure 2. The three components of the exclusion rate, 2001-2002 (Source: LFS microdata files made available by INE)

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

34

2001.Q1 2001.Q2 2001.Q3 2001.Q4 2002.Q1 2002.Q2 2002.Q3 2002.Q4

perc

enta

ges

of o

ld u

nem

ploy

ed

0

4

8

12

16

20

24

28

32

36

40

44

48

52

56

60

64

68

perc

enta

ge o

f job

see

kers

onl

y th

roug

h PE

S

Job seekers only through PESOther exclusionsTotal exclusion rateProportion of non-active contacts

As for the proportion of jobseekers declaring registration at the PES as their only

search method who were excluded because of their contact with the offices was not to

find work (β in expression [1]), it follows a smoothly declining trend, dropping from

68% in the first quarter of 2001 to 53% in the third quarter of 2002, and the jumps back

to 63% in the last quarter of 2002. In terms of the decline of the total exclusion rate

observed between 2001.Q4 and 2002.Q1, the decrease of β explains more or less the

remaining 20% of the decline in the total exclusion rate8. However, its surge in the last

quarter of 2002 is the only factor under the recovery of the exclusion rate, given that α

remained stable (even declining somewhat).

These evolutions might be taken to suggest that the process of obtaining the new

unemployment figures from the old ones does not provide a stable exclusion rate, as

8 There is an interaction term implicit in the decomposition, as well as the effect of the small

variation of the third term ( ε in expression [1]).

Page 10: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

8

could be expected from the conceptual definitions. If the purpose of the new regulation

was to calculate more finely the number of unemployed workers by “cleaning” the

figures from the contaminating activities of people who should not be considered active

job seekers, onw should expect a relatively constant exclusion rate. If this is not the

case, it implies that the behaviour of would-be unemployed people is erratic over time,

in the sense that at some point in time they meet the requirements of active search and at

some other points they do not. Should this be accepted as an explanation, it would imply

that the grounds to exclude them from unemployment are rather weak.

However, Figures 1 and 2 tend to suggest that the process of exclusion is not

erratic. Actually, there are two clearly different periods in the figures. One corresponds

to 2001, when the new definition was used on an experimental basis, and the other one

corresponds to 2002, when the new definition was formally implemented. In addition,

there is another anomalous observation in the last quarter of 2002, when the proportion

of those excluded because of a lack of active contact with the public employment

offices increased.

In order to determine the reasons behind these peculiar evolutions, in the

following two subsections, information shall be presented broken down by various

characteristics of the unemployed as well as by their region of residence, a very

important variable. In all of these analyses, the comparisons will then be made by taking

three points of observation: the average of 2001, the average of the first three quarters of

2002 and the gourth quarter of 2002.

3.2 Exclusions by characteristics of the unemployed

The first analysis to be presented refers to the gender and age of the

unemployed. Figure 3 presents the corresponding breakdowns. A clear pattern emerges

in terms of exclusions: they are lower, and rather stable for males under 55 and higher

for older men; in the case of females, they tend to rise with age. These patterns remain

relatively stable over the three periods considered, with a changing general level,

dropping from 2001 to the first three quarters of 2002 and increasing in the last quarter

of 2002. On the whole, exclusion rates are higher for females than for males, the

difference being around 4 percentage points in all three observation moments.

Page 11: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

9

Figure 3. Exclusion rates from old unemployment by gender and age, Spain, 2001-2002 (Source: INE, LFS microdata files)

The second variable that can be analyzed is the level of education. Figure 4 plots

the exclusion rates by educational level and gender in the three periods considered. The

patterns emerging from this figure are interesting. Starting with males, the somewhat

erratic curve depicted for 2001 becomes much smoother and U-shaped in the first three

quarters of 2002, with a minimum at lower secondary education and a maximum at the

higher education levels. The last quarter seems to maintain a similarly more or less U-

shaped form but with some erratic variations, although not coincident with those

observed in 2002. In the case of females, the curve is similar for 2001 (more or less U-

shaped but with a central peak for upper secondary education), then becoming more

clearly U-shaped in 2002, although with less steep than for males. In general, the

extreme levels of education (primary and university) seem to be more prone to be

excluded from unemployment than central levels, be it academic or vocational (with the

exception of upper secondary education at some points).

8

10

12

14

16

18

20

22

24

26

28

3016

-24

25-2

9

30-3

4

35-3

9

40-4

4

45-4

9

50-5

4

55-5

9

60-6

4

ALL

MA

LES

16-2

4

25-2

9

30-3

4

35-3

9

40-4

4

45-4

9

50-5

4

55-5

9

60-6

4

ALL

FE

MA

LES

MALES FEMALES

perc

enta

ge o

f old

une

mpl

oyed

20012002-Q1:Q32002-Q4

Page 12: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

10

Figure 4. Exclusion rates from old unemployment by gender and level of education, Spain, 2001-2002 (Source: INE, LFS microdata files)

In one sense, the results so far are not very surprising. It is well-known that

people differ in terms of their search behaviour depending on their gender, age and level

of education, and this is what the exclusion process is all about: eliminating from the

unemployment count people who do not look for work actively enough. The direction

of these variations, however, makes more sense in the case of the age variable (older

people search less intensively than younger ones) than in the case of the education

variable (one would expect highly qualified people to search more intensely).

An interesting variable here is whether the unemployed person receives

unemployment benefits. Economists tend to believe generally that receiving

unemployment benefits inhibits job search9. Is there a relationship between

unemployment benefits and exclusion? Figure 5 plots the data, again broken down by

gender. Interestingly enough, the rate of exclusion is higher for those who, although

registered at the employment offices, do currently not receive unemployment benefits,

9 For a somewhat dissenting view, see García and Toharia (2000). Using both LFS data and

information from a specific survey to registered unemployed workers, they argue that benefits does indeed inhibit job search to the point that such people is counted as inactive. However, for those remaining as unemployed, the effect is in most cases not statistically significant.

8

1012

14

16

1820

22

24

2628

30

Illit.,

no e

duca

tion

Prim

ary

educ

atio

n

Low

er s

econ

dary

edu

catio

n

Upp

er s

econ

dary

edu

catio

n

Low

er v

ocat

iona

l tra

inin

g

Upp

er v

ocat

iona

l tra

inin

g

Low

er u

nive

rsity

edu

catio

n

Upp

er u

nive

rsity

edu

catio

n

Illit.,

no e

duca

tion

Prim

ary

educ

atio

n

Low

er s

econ

dary

edu

catio

n

Upp

er s

econ

dary

edu

catio

n

Low

er v

ocat

iona

l tra

inin

g

Upp

er v

ocat

iona

l tra

inin

g

Low

er u

nive

rsity

edu

catio

n

Upp

er u

nive

rsity

edu

catio

n

M A LES FEM ALES

perc

enta

ge o

f old

une

mpl

oyed

2001

2002-Q1:Q3

2002-Q4

Page 13: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

11

this being the case in all periods of observation and for males as well as females. Of

course, other effects may be at play, such as the duration of unemployment (because

benefits are limited in time). Figure 6 shows that, truly enough, higher search durations

are associated with higher exclusion rates. The observed effect of higher exclusion rates

for those registered but not receiving unemployment benefits might probably be due to

their presumably longer search duration. We shall return to this point in the following

sub-section.

Figure 5. Exclusion rates from old unemployment by gender and unemployment benefit situation, Spain, 2001-2002 (Source: INE, LFS microdata files)

4

6

8

10

12

14

16

18

20

22

24

26

28

Receives UB Registered but noUB

Not registered Receives UB Registered but noUB

Not registered

MALES FEMALES

perc

enta

ge o

f old

une

mpl

oyed

20012002-Q1:Q32002-Q4

Page 14: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

12

Figure 6. Exclusion rates from old unemployment by gender and job search duration, Spain, 2001-2002 (Source: INE, LFS microdata files)

The final variable which is worth being discussed as a breakdown variable for

exclusion rates is the region of residence. Figure 7 presents the exclusion rates for the

17 Spanish regions 10; the regions have been ranked from highest to lowest in terms of

their exclusion rate in 2001. This figure is very striking. It shows big differences among

Spanish regions regarding the proportion of unemployed excluded by the new

definition. Thus, while in Catalonia the percentages remain under 5%, other regions

such as Asturias or Extremadura show figures above 35% and even close to or over

50% in 2001. Another interesting feature of this figure is the reduction in this high

divergence between 2001 and 2002: standard deviation dropped from 11 to 6 points for

males, and from 13 to 8 for females11.

10 Data for the North African towns of Ceuta and Melilla have been excluded from this analysis,

due to their abnormally high exclusion rates, over 80%. 11 This decrease is less clear in relative terms, due the general decrease of exclusions.

4

6

8

10

12

14

16

18

20

22

24

26

28

Less than 3months

3-5 months 6-11 months 12-23months

24 months &over

Less than 3months

3-5 months 6-11 months 12-23months

24 months &over

MALES FEMALES

perc

enta

ge o

f old

une

mpl

oyed

20012002-Q1:Q32002-Q4

Page 15: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

13

Figure 7. Exclusion rates from old unemployment by gender and region of residence, Spain, 2001-2002 (Source: INE, LFS microdata files)

These results are difficult to interpret. Of course, the main reason behind these

differences is the varying proportion of job seekers who only mention registration of

employment offices as search method12. And significant differences, both across regions

and over time, as they are observed in Figure 7, are difficult to understand. One can

argue that the problem resides with the way in which the survey is carried out. Should

this be the case, as we believe it to be to a large extent, the conclusion should not be that

the Statistical Office is to be blamed. Rather, the problem is that the notion of active job

search is difficult to investigate using the methods recommended by the European

Commission. A large list of possible methods (most of which are supposed to be read

when deliveting the interview), the inherent looseness of the methods included in the

list (e.g. “asking friends or relatives” is a valid active job search method) and the fact,

sometimes overlooked, that the respondent may be proxied by someone else in the

12 Not presented here for reasons of space.

0

5

10

15

20

25

30

35

40

45

50

55

Astu

rias

Extre

mad

ura

Arag

ón

La R

ioja

Can

tabr

ia

Mad

rid

Cas

tilla

-La

Man

cha

Gal

icia

Nav

arra

Cas

tilla

y L

eón

Can

aria

s

Mur

cia

Anda

lucí

a

Com

.Val

enci

ana

País

Vas

co

Bale

ares

Cat

aluñ

a

La R

ioja

Astu

rias

Extre

mad

ura

Arag

ón

Can

tabr

ia

Nav

arra

Mad

rid

Cas

tilla

-La

Man

cha

Gal

icia

Cas

tilla

y L

eón

Anda

lucí

a

Mur

cia

Can

aria

s

Com

.Val

enci

ana

Bale

ares

País

Vas

co

Cat

aluñ

a

MALES FEMALES

perc

enta

ge o

f old

une

mpl

oyed

20012002-Q1:Q32002-Q4

Page 16: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

14

household, impart a significant difficulty to the notion of active search and to the way in

which it is asked in labour force survey questionnaires.

3.3 The probability of being excluded, 2001 and 2002

To complete the preceding graphical analyses, we now present a multivariate

logistic regression analysis of the probability of being excluded. The regressions have

been run separately for males and females and for the three periods considered in the

preceding sub-section: 2001, the first three quarters of 2002 and the fourth quarter of

2002.

The results, presented in Table 1, suggest that the simple graphical analysis

carried out in the preceding pages stands up when all the possible cross effects are taken

into account. Age produces a U-shaped effect in the case of males and an increasing

effect for females. The influence of education is mixed, without a well-defined pattern.

It is surprising that males with upper university education tend to have (in 2002) the

highest probability of being excluded, an effect which is not so obvious in the case of

females. The negative effect of unemployment benefits on the probability of being

excluded whithers away, probably due to the inclusion of search duration regressors,

which show the expected positive relationship with the probability of exclusion

(although concentrated in the extremes, as the coefficients tend not to be statistically

different for the groups 6-11 months and 12-23 months). The region of residence is

clearly the most decisive variable, repeating the results already shown in Figure 7.

Page 17: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

15

Table 1. Logistic regressions of the probability of being excluded from unemployment under the new definition, conditional on being considered unemployed under the old definition, Spain, 2001-2002 (Source: estimated from LFS microdata provided by INE)

16-24 0,18 ** 0,28 ** 0,27 * 0,22 ** -0,41 ** -0,37 ** -0,41 ** -0,39 **25-29 0,09 0,21 ** 0,19 0,15 ** -0,36 ** -0,29 ** -0,33 ** -0,32 **30-34 -0,04 -0,06 0,00 -0,02 -0,20 ** -0,37 ** -0,23 ** -0,25 **35-39 0,01 0,12 0,06 0,06 -0,25 ** -0,04 -0,01 -0,14 **40-44 (&)45-49 0,04 0,04 0,22 0,06 -0,03 0,01 0,01 -0,0150-54 -0,16 * -0,11 0,31 * -0,09 0,04 0,02 0,35 ** 0,0755-59 0,28 ** 0,49 ** -0,02 0,31 ** 0,15 * 0,20 * 0,42 ** 0,19 **60-64 0,30 ** 0,80 ** 0,78 ** 0,55 ** 0,08 0,07 0,34 0,10

Illit.,no education 0,03 0,22 * 0,13 0,10 * -0,15 * 0,06 0,10 -0,05Primary education -0,09 0,12 0,01 0,00 -0,11 ** -0,02 0,06 -0,05Lower secondary education -0,34 ** -0,21 ** -0,32 ** -0,31 ** -0,24 ** 0,02 0,06 -0,12 **Upper secondary education (&)Lower vocational training -0,30 ** 0,09 0,25 -0,11 * -0,38 ** -0,08 0,04 -0,23 **Upper vocational training -0,30 ** 0,05 -0,07 -0,17 ** -0,22 ** -0,04 -0,04 -0,15 **Lower university education -0,19 ** 0,14 0,00 -0,05 -0,11 * 0,04 0,28 ** -0,01Upper university education -0,12 0,56 ** 0,61 ** 0,22 ** -0,03 0,06 0,19 0,03

Receives UB (&)Registered but no UB 0,05 0,03 0,05 0,03 0,08 * 0,01 0,07 0,05 *Not registered -0,76 ** -0,62 ** -0,54 ** -0,69 ** -0,85 ** -0,74 ** -0,84 ** -0,82 **

Less than 3 months -0,67 ** -0,61 ** -0,75 ** -0,65 ** -0,32 ** -0,50 ** -0,21 ** -0,37 **3-5 months -0,23 ** -0,14 * -0,39 ** -0,22 ** -0,01 -0,16 ** -0,13 -0,08 **6-11 months 0,04 0,15 * -0,31 ** 0,04 0,09 * -0,08 0,13 0,0412-23 months (&)24 months & over 0,14 ** 0,32 ** 0,07 0,18 ** 0,18 ** 0,12 ** 0,22 ** 0,16 **

Andalucía (&)Aragón 1,33 ** 1,36 ** 0,24 1,22 ** 1,32 ** 1,59 ** 0,77 ** 1,34 **Asturias 1,81 ** 0,90 ** 0,91 ** 1,42 ** 1,36 ** 0,83 ** 0,45 ** 1,10 **Baleares -0,24 0,41 * -0,27 0,00 -0,19 0,32 * 0,19 0,05Canarias 0,26 ** 0,69 ** 0,37 ** 0,41 ** -0,03 -0,01 -0,02 -0,03Cantabria 1,08 ** 1,01 ** 0,18 0,95 ** 0,81 ** 0,18 0,24 0,58 **Castilla y León 0,34 ** 0,71 ** 0,39 ** 0,46 ** 0,22 ** 0,45 ** 0,32 ** 0,31 **Castilla-La Mancha 0,61 ** 0,99 ** 0,70 ** 0,74 ** 0,57 ** 0,84 ** 0,38 ** 0,63 **Cataluña -1,30 ** -1,52 ** -2,39 ** -1,46 ** -1,52 ** -1,77 ** -1,62 ** -1,61 **Com.Valenciana -0,19 ** 0,30 ** -0,68 ** -0,07 -0,14 ** 0,42 ** -0,34 ** 0,04Extremadura 1,27 ** 0,64 ** 0,70 ** 1,01 ** 1,23 ** 0,97 ** 0,77 ** 1,07 **Galicia 0,50 ** 0,39 ** 0,33 ** 0,45 ** 0,21 ** 0,25 ** -0,04 0,19 **Madrid 0,80 ** 1,22 ** 1,20 ** 1,00 ** 0,57 ** 0,97 ** 0,87 ** 0,74 **Murcia 0,28 ** -0,08 -0,56 * 0,07 0,20 ** 0,21 * -0,38 * 0,14 **Navarra 0,64 ** 0,35 -0,55 0,46 ** 0,69 ** -0,30 0,02 0,33 **País Vasco -0,37 ** -0,28 * -0,24 -0,32 ** -0,50 ** -0,08 -0,26 * -0,34 **La Rioja 1,14 ** 0,31 0,90 * 0,86 ** 1,49 ** 0,81 ** 0,78 ** 1,13 **Ceuta y Melilla 3,00 ** 3,32 ** 3,28 ** 3,14 ** 3,04 ** 3,12 ** 3,09 ** 3,06 **

2001 (&)2002 Q1-Q3 --- --- --- -0,46 ** --- --- --- -0,42 **2002 Q4 --- --- --- -0,24 ** --- --- --- -0,26 **Constant -1,49 ** -2,45 ** -1,77 ** -1,64 ** -0,99 ** -1,58 ** -1,50 ** -1,07 **Sample size(&) indicates reference category included in the constant(**) significant at the 99% level; (*) significant the the 95% level

2002Q4 ALLMALES FEMALES

AGE

LEVEL OF EDUCATION

2001 2002Q1-Q3 2002Q4 ALL2001 2002Q1-Q3

UNEMPLOYMENT BENEFIT SITUATION

SEARCH DURATION

REGION OF RESIDENCE

YEAR OF OBSERVATION

16317 12188 4383 32888 23428 18021 6226 47675

Page 18: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

16

4 Is the change meaningful? Longitudinal evidence The results presented in Section 3 suggest that the separation from

unemployment of “passive job seekers” has encountered practical difficulties, at least in

the case of Spain, probably the country where such a group was expected to represent

the most significant share of total unemployment. In this section, we ask ourselves a

different question: does the proposed change make economic sense? That is, no matter

how complicated it may be to grasp the real nature of active job search, it is important

to do so because the group concerned cannot be sensibly be equated to the other job

seekers. As a matter of fact, the implicit argument in the 1897/2000 regulation is that

these people ought to be counted within the inactive population, where they really

belong.

In order to test this hypothesis, in this section we study the process of exit of the

various groups of jobless people and their probability of getting a job. The three main

groups considered are the unemployed according to the new definition, those excluded

from unemployment and the inactive population. The analysis is restricted to those

under 65, to avoid the biases that would be present if older people were included in the

analysis. The idea is to see whether the labour market behaviour, in terms of their

employment dynamics, of the excluded can be equated to that of the inactive population

or to that of the active job seekers. First, a descriptive analysis of the gross probability

of entering employment within one quarter is presented. A more complete regression

analysis follows. The data used comes from the longitudinal version of the Spanish

LFS, made available to researchers on a general basis by the Spanish Statistical Office

(INE).

4.1 Exit rates towards employment

Figure 8 presents the rates of exit from joblessness towards employment within

one quarter. Six categories of jobless workers are considered: unemployed, excluded

from unemployment and inactive, distinguishing in all three cases betwen those with

past job experience and those wihout such experience. Seven observations are available,

starting in the first quarter of 2001 and ending in the fourth of 2002.

Page 19: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

17

Figure 8. Exclusion rates from old unemployment by gender and region of residence, Spain, 2001-2002 (Source: INE, LFS microdata files)

The results are very interesting. Both for males and females, there exist a clear

modulation of the gross probability of entering employment within one quarter: the

unemployed are the group showing the largest probabilities, followed by those excluded

from unemployment, leaving way behind the inactives. This hierarchy may be observed

both for those with and withour past job experience. The conclusion that stems from this

figures is clear: in terms of their probability of finding a job, people excluded from

unemployment by the European Commission regulation do behave differently from

those who are left inside; however, despite this disadvantage vis-à-vis the active hob

seekers, the passive job seekers show a clearly higher propensity to enter employment

than those outside the labour force. They thus represent a middle group between the

“pure unemployed” and the “pure inactive”. Should they be included in one group or the

other? The evidence is obviously inconclusive, but what can be said is that this people

are not totally detached from the labour market. They may have higher difficulties to

find jobs, probably related to their less active job search attitude, but they cannot in any

way be equated to inactive people who do not search work for whatever reason.

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

32

34

36

2001-I 2001-II 2001-III 2001-IV 2002-I 2002-II 2002-III 2001-I 2001-II 2001-III 2001-IV 2002-I 2002-II 2002-III

MALES Initial period FEMALES

Perc

enta

ge o

f peo

ple

in in

itial

per

iod

Unemployed with past job experienceUnemployed with exp./ExcludedUnemployed without past job experienceUnemployed w/o exp./ExcludedInactives with past job experienceInactives without past job experience

Page 20: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

18

4.2 The probability of exiting joblessness

In order to underpin the results presented in the preceding subsection, it is

necessary to take into account the influence of other variables in the probability of

finding a job, which might be influencing the gross probabilities examined in Figure 8.

To do so, logistic regressions of the probability of finding a job within one quarter have

been estimated, for all those without a job in the initial period of observation (active and

passive jobseekers as well as inactive). Regressors include gender (although separate

regressions have also been run for males and females), age, education, labour market

status, past job experience, situation with respect to unemployment benefits (payable to

both unemployed and inactive persons as defined by the LFS), all of them observed at

the initial moment. Controls for the region of residence and the period of observation

have also been included. Pooled regressions for the seven transitions observed have

been run13.

Table 2 presents the results of these regressions. The most interesting results

from the point of view of this article relate to the coefficients of the labour market status

variable. As can be seen, being an active job seeker as defined by the new European

Commission regulation provides a clear advantage in terms of finding a job, this being a

significant result for males and females alike. The distance, it should be added, is larger

for females. In addition, being a passive job seeker also provides a clear, even larger,

advantage over non-seekers in terms of finding a job. In this case, the distance is

somewhat smaller for females.

Other results of these regressions are quite standard: age is negatively correlated

to the probability of finding a job, as is past job experience. The level of education

shows a U-shaped influence in the case of males, probably due to the well-known

divergent skill structure of Spanish employment14. In the case of females, however, only

tertiary education exerts a clearly positive influence. Finally, the situation with respect

to unemployment benefits shows an unexpected result, similar to that found earlier in

this article (recall Figure 5): receiving unemployment benefits is associated with a

higher probability of finding a job, this being the case for males and females.

13 Regressions estimated for each of the transitions individually gave substantially similar

results; pooling the regressions has the advantage of increasing sample size. 14 On this point, see, for example, Fina et. al. (2000)

Page 21: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

19

Table 2. Logistic regressions of the probability of finding a job within one quarter, jobless persons, Spain, 2001-2002 (Source: estimated with the matched files of the LFS provided by the INE)

The latter result deserves further investigation. One problem with the regressions

presented in Table 2 is that they include a large number of observations from inactive

persons, for whom other informations, such as the duration of search is lacking. So,

once we have established the result that passive job seekers are clearly different from

non-seekers (inactives), it may be worthwhile to rerun the regressions just with the two

groups of jobseekers. The results are shown in Table 3. The main difference with the

regressions in Table 2 is that a search duration variable has been included.

The results provide very similar differences in terms of the advantage of active

job search for the probability of finding a job. There are two main differences with

respect to the results presented above. First the influence of the education variables

Unemployed (new definition) 0,334 ** 0,261 ** 0,383 **Excluded from unemployment(&)Inactive -0,783 ** -0,874 ** -0,690 **

Yes (&)No -0,198 ** -0,213 ** -0,189 **

Males (&)Females -0,088 **

16-24 0,650 ** 0,496 ** 0,735 **25-29 0,535 ** 0,480 ** 0,554 **30-34 0,186 ** 0,212 ** 0,158 **35-39 0,138 ** 0,100 0,156 **40-44 (&)45-49 -0,268 ** -0,242 ** -0,294 **50-54 -0,524 ** -0,637 ** -0,457 **55-59 -1,074 ** -1,183 ** -1,022 **60-64 -1,758 ** -2,093 ** -1,458 **

Primary education or less 0,067 ** 0,144 ** -0,004Lower secondary education 0,043 0,143 ** -0,027Upper secondary education (&)Tertiary education 0,290 ** 0,204 ** 0,337 **

Receiving benefits(&)Registered, no benefits -0,304 ** -0,267 ** -0,329 **Not registered -0,844 ** -0,659 ** -0,964 **Constant -0,911 ** -0,813 ** -1,600 **SAMPLE SIZENote: Controls also included for region of residence and initial period of observation(&) indicates category included in reference term (constant)(**) significant at the 99% level; (*) significant the the 95% level

UNEMPLOYMENT BENEFIT SITUATION

285310 92118 193192

PAST JOB EXPERIENCE

GENDER

AGE

LEVEL OF EDUCATION

ALL MALES FEMALESLABOUR MARKET STATUS

Page 22: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

20

vanishes except for the more educated who still derive a positive differential. Secondly,

the positive effect of receiving unemployment benefits also disappears for males and

loses strength and statistical signification in the case of females. On this latter variable,

a result which should be stressed is the much lower probability of finding a job for those

who declare being not registered in an employment office. This is interesting

particularly in the light of the very low probability of being excluded that this group

shows. Because they do not seek work though an official emnployment office, the

channels they use are probably more informal, but belonging to the “active” methods

category. If the probability of finding a job test should be applied to them, however,

they would deserve being excluded from unemployment on much more solid grounds

than so-called “passive” jobseekers do.

Page 23: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

21

Table 3. Logistic regressions of the probability of finding a job within one quarter, jobseekers only, Spain, 2001-2002 (Source: estimated with the matched files of the LFS provided by the INE)

Excluded from unemployment (&)Unemployed (new definition) 0,311 ** 0,221 ** 0,371 **

Yes (&)No -0,115 ** -0,115 ** -0,114 **

Males (&)Females -0,076 **

16-24 0,352 ** 0,198 ** 0,488 **25-29 0,325 ** 0,200 ** 0,431 **30-34 0,107 * 0,076 0,136 *35-39 0,107 * 0,070 0,136 *40-44 (&)45-49 -0,033 -0,002 -0,07750-54 -0,199 ** -0,256 ** -0,15555-59 -0,506 ** -0,567 ** -0,462 **60-64 -0,859 ** -1,095 ** -0,445 *

Primary education or less 0,025 0,067 -0,042Lower secondary education 0,056 0,087 0,030Upper secondary education(&)Tertiary education 0,196 ** 0,108 * 0,238 **

Receiving benefits (&)Registered, no benefits -0,076 ** -0,049 -0,087 *Not registered -0,138 ** -0,026 -0,207 **

Less than 3 months 0,751 ** 0,772 ** 0,710 **3-5 months 0,599 ** 0,572 ** 0,622 **6-11 months 0,381 ** 0,368 ** 0,389 **12-23 months (&)24 months & over -0,283 ** -0,350 ** -0,230 **CONSTANT -1,646 ** -1,445 ** -2,361 **SAMPLE SIZE 54326 22416 31820Note: Controls also included for region of residence and initial period of observation(&) indicates category included in reference term (constant)(**) significant at the 99% level; (*) significant the the 95% level

UNEMPLOYMENT BENEFIT SITUATION

SEARCH DURATION

PAST JOB EXPERIENCE

GENDER

AGE

LEVEL OF EDUCATION

ALL MALES FEMALESLABOUR MARKET STATUS

Page 24: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

22

5 Final remarks This paper has explored the consequences of the new definition of

unemployment introduced by the European Commission in its 1897/200 regulation. Our

first point has been to make it clear that the new definition implied a process of

exclusion of some workers considered to behave in a “passive job search” way, enough

to let them bc considered “inactive” people, similar to those who do not even declare

that they are seeking work. We have next analyzed this process of exclusion for the case

of Spain, a country particularly interesting in this context, as it is one of the

paradigmatic cases of extensive passive job search through employment offices. The

first result we have come up with has been that the main element behind exclusions, i.e.

behind passive jobseeking is the number of search methods mentioned, in particular, the

fact that a substantial proportion of job seekers only declare registration in a public

employment office as search method.

We have further analyzed the specific characteristics of people which increase

their propensity to act as passive job seekers and hence to be excluded. Alongside more

standard variables, such as age, gender, education and job search duration, influencing

job search in the expected way, we have found that unemployment benefits do not seem

to clearly influence active job search. More importantly, we have found that there have

been significant differences across regions, which furthermore have tended to change

over time. This has suggested a first conclusion: even accepting that the difference

between active and passive job seekers may be meaningful, there are clear difficulties to

clearly define in practice these two groups through labour force surveys. The European

Commission regulation does not appear to have established clear instruments to do so.

Next we have questioned the initial distinction which provides the starting point

for the EC regulation: are there differences between active and passive jobseekers in

terms of their economic behaviour? Our analysis has centered on the probability of

finding a job within one quarter. Passive jobseekers have been found to be an

intermediate group in between active jobseekers and nonseekers, although the distance

with the latter appears to be larger.

The final question is whether passive jobseekers should be excluded from the

unemployment count. On a purely conceptual basis, the answer may be positive. On a

practical, LFS-oriented, account, the answer is less clear, given the difficulties to clearly

Page 25: What does it take to be (counted as) unemployed? The case ...campus.usal.es/~ehe/Papers/Toharia2.pdf · counted as unemployed under the old definition. If one of them turns out to

23

identify, on a stable basis, the group of passive jobseekers. On a labour market

perspective, finally, the answer is mixed, as passive jobseekers are an intermediate

group. On the whole, then, the mere exclusion of passive jobseekers from

unemployment may be justified, but their inclusion with the inactive is not. The

advantages (conceptual, and political?) of reducing unemployment through a statistical

artifact may be outweighed by the practical and analytical disadvantages stemming from

the higher volatility of the new unemployment concept and the unwarranted inflation of

the inactive population group.

References EIRO (European Industrial Relations Observatory (2002), “Controversy over

new definition for measuring unemployment”, available on the internet at the address

http://www.eiro.eurofound.ie/print/2002/02/feature/ES0202214F.html.

European Commission (2000), Regulation no. 1897/2000, Official Journal of the

European Communities, L.228, 8th September, pp. 18-21.

Fina, Ll., Toharia, L., García Serrano C. and Mañé, F. (2000), “Cambio

ocupacional y necesidades educativas de la economía española”, in F. Sáez (ed.),

Formación y empleo, Madrid, Fundación Argentaria, 2000, pp. 47-154 .

García, I. and Toharia, L. (2000), “Prestaciones por desempleo y búsqueda de

empleo”, Revista de economía aplicada, 23, vol. VIII, autumn 2000, pp. 5-33.

Toharia, L. (2000), “La nueva definición de desempleo”, in Round table on

Regulation and Labour Market Institutions, Workshop to present the new Labour

Sciences Degree Curriculum, Universitat Pompeu Fabra, Barcelona, october, mimeo.