Patterns of integration Low educated people and their jobs in - PowerPoint Presentation

Slide1

Patterns of integrationLow educated people and their jobs in Norway, Italy and Hungary

János KöllőInstitute of Economics, Budapest & IZA

IZA DP 7632 and BWP 2013/15Slide2

I compare people with 0-10 years of school-based education in three countries, which provide their unskilled population with work highly successfuly (Norway), more-or-less successfully (Italy) and completely unsuccessfully (Hungary)Hungary’s failure is not unique. The CEEs fail as a group, which calls for explanations other than those referring to national institutions and policies (MW, benefits, ALMP, etc.)This paper looks at the role of skills and skill requirements, supposing that integration is easier to achieve if:

the supply of simple jobs is abundantor low educated adults accumulate sufficient cognitive and non-cognitive skills to attend complex jobsor firms are able/willing to bridge the gap between skills and skill requirements

The paperSlide3

1) How the composition of jobs by complexity and firms’ willingness to hire low educated labor (given

complexity) contribute to unskilled employment?2) To what extent are the low educated engaged in various forms of post-school skills formation?3) What is the role of small-firms and self-employment in the unskilled labor market?**Asked under the assumption that small businesses can deal with skill deficiencies more successfully than do large formal organizations

Related questionsSlide4

Hungary, Norway and Italy participated in two international skill surveys: Adult Literacy and Life Skills Survey (ALL, 2003, 2008). International Adult Literacy Survey (IALS 1998)The analysis is predominantly based on ALL, which contains data on 4493 Norwegians, 5830 Italians and 4875 Hungarians (prime age adults, students excluded), of wich 21, 46 and 25 per cent are low educated.

The key variables are education, workplace duties requiring literacy, numeracy and communication skills, participation in skill enhancing activities and firm sizeThe literacy test scores are used marginally because of endogeneity (test score  employment, test score  job complexity)DataSlide5

Literature5

The bulk of the related literature on CEEs analyze national institutions (welfare, ALMP, stc.) or the transition process (OST models)The questions of this paper have more to do with the SBTC and job polarization literature discussing the linkages between job content and relative demand. However, the SBTC literature typically deals with the college/no college or production/non-production division.ALL is new, no academic research as yet. Papers working with IALS data predominantly compare test results or look at the linkages between test scores and wages.Micklewright & Brown (

2004), Micklewright & Schnepf (2004), Devroye

& Freeman (2000), Blau & Kahn (2000)

, Denny et al. (2004), Carbonaro (2002

)

McIntosh

& Vignoles (

2000

)Slide6

The low-educated population6

0) Basic descriptive statisticsSlide7

Table 1: Share of the unskilled

Norway

Italy

Hungary

0-10 years in school

a

IALS

19.0

52.4

28.6

ALL

20.5

46.1

24.6

ISCED 0-2

IALS

11.7

55.2

28.3ALL12.648.724.7The data relate to the population aged 15-64 excluding students and persons, who are older than 35 and never worked.a) Completed 0-10 years in school. Based on the question on the number of completed schoolyears not counting repeated years (a1) in both surveys, except for Norway in the IALS (a8no). Sampling errors t.b.a.

Norway

adopted

the international methodology of defining ISCED codes only in 2006. The reform brought up the ISCED 0-2 share from 10 to 20%

Size of the low educated population

7Slide8

Skills of the low educated population

8Absolute terms: Norway >> Hungary > Italy

Relative terms: Norway > Hungary  ItalyTreat the absolute values with cautionSlide9

Employment of the low educated population

Absolute terms: Norway  Italy >> HungaryRelative terms: Norway  Italy >> HungaryNote that the data above exclude persons older than 35, who never worked before (mostly Italian

housewives). Discussion in the paper.Slide10

101) Job complexity and unskilled employment Slide11

Measuring job complexity

11Slide12

Measures of job complexity: Number of tasks (0-17)

Number of Type 1 tasks (0-11) Number of tasks weighted with their effect on wages in the pooled sample 17 task dummies as controls, occasionally

Measuring job complexity

12Slide13

Share effect

How the share of the unskilled in j-type jobs relates to their population share?

Size effect

How many j-type jobs are at the disposal of the entire working age population?

Total contribution

Contribution of j-type jobs to the employment to population ratio of the low educated population

D

ecomposition

of the unskilled employment to population ratio

13Slide14

The distribution of jobs by complexity14

Norway: most jobs require 6-16 tasks, with a mode at R=10

Italy: many simple jobs, a mode of R=0 Hungary: a distribution closer to Italy’s but far less simple jobs

95% confidence intervals based on jacknife survey standard errorsSlide15

The unskilled share by job complexity15

95% confidence intervals based on jacknife survey standard errors. The population shares are indicated by the horizontal lines.

Norway: steeply rising shares as we move toward simple jobs. Shares above the population share if R<9

Italy:very high shares in simple jobs, above the population share if R<6.Hungary: no steep rise as we move to simple jobs. Shares exceeding the population share if R<5. (R=0 if only Type 1 tasks are considered). The shares are lower than in Italy in middling and highly complex jobsSlide16

Total contributions to the unskilled employment to population ratios

Norway: the bulk of unskilled employment comes from jobs requiring 6-12 tasks

Italy: simple (R=0) jobs have far the largest contributions

Hungary: similar to Italy but the contributions are smaller in simple and middling jobs 

95% confidence intervals based on jacknife survey standard errorsSlide17

172) Post-school skill formationHow can low educated people attend complex jobs?Slide18

Post-school skills formation Adult training

Informal learning activities Acquiring skills at homeCivil integrationEndogeneity: work as a source of literacy18Slide19

Post-school skill formation19

Norway performs slightly better in absolute termsThe ranking in relative terms is ambiguousNote that the Norwegian school system does not perform very well according to the PISA surveysSlide20

Post-school skill formation20

Participation in formal adult training is about 6 times as frequent in Norway as in Italy and HungaryAbout 2 -2.5 times more frequent in relative termsSlide21

Post-school skill formation21

Substantial difference between Norway and ItalySmaller difference between Italy and Hungary, especially in relative termsSlide22

Post-school skill formation22

Substantial difference between Norway and Italy/Hungary except for reading newspapers and magazinesSlide23

Post-school skill formation23

Huge difference between Norway and ItalySmaller between Italy and Hungary, especially in relative terms.Participation rates are very low in absolute terms in Italy and especially in HungarySlide24

Post-school skill formation24

Huge difference between Norway and Italy/HungaryPractically no difference, and very low absolute levels in Italy and HungarySlide25

Post-school skill formation25

Practically all Norwegians take part in at least one activity compared to only ½ of the Italians and 1/3 of the HungariansOver 40 per cent of the Norwegians take part in 6 or more activities compared to only 7 and 2 per cent of the HungariansSlide26

Post-school skill formation26

Index= employment probability * average number of literacy tasks to be performed at work = prob*degree of exposureIntended to capture the exposure of the low educated population to literacy-intensive tasks through workHuge difference between Norway and ItalySmaller but significant difference between Italy and HungarySlide27

27

3) Small businesses How can low educated people attend complex jobs?Slide28

Decomposition of the unskilled employment ratio by firms size

Norway: the bulk of unskilled employment comes from large firmsItaly: the bulk of unskilled employment comes from small firms, including self-employmentHungary: roughly equal contributionsSlide29

29Small firms in Italy

The Italian small business sector employs a much higher share of low educated workers in complex jobs. Is it about a selection effect (from within a large low educated population)?Slide30

Efficient selection and/or learning by doing?In generalItaly: large uneducated population

 wider scope for selection by actual skills.Indeed: test scores rise more steeply as we move toward complex jobs in ItalyMore efficient learning by doing might also produce the same correlation but such an assumption seems quite arbitrary Slide31

31Efficient selection and/or learning by doing?By firm size

If not a selection effect?Selection on non-cognitive skills?Intense inter-personal interactions

help in overcoming skill deficiencies?The skills gap is not managed

successfuly i.e. ‘solved’ at the cost of inefficiencies?Slide32

More simple jobs than in Norway but much less than in ItalyThe unskilled share is low everywhere including simple jobs (substitution with vocationally trained workers?)***Far from Norway in terms of post-school skill formation and civil integration

Far from Italy in terms of integration through a ‘permissive’ small business sector***We basically (still) observe the implications of the damage that state socialism had made to the traditional private economy, on the one hand, and civil society, on the other***Heading for Italy or Norway?

Conclusions for HungarySlide33

33Thank youSlide34

Extremely low rate of unskilled employment in both absolute and relative terms. The CEEs fail as a groupUnskilled E rate = CEE + u (country level OLS, EU LFS data) explains 61 per cent of the cross-country variation

Prob(E| unskilled) = CEE (individual level, pooled EU LFS) correctly classifies 73 per cent of the males, 55 per cent of the women and 59 per cent of both genders. No evidence of job polarization*: continuing failure, risk of social fragmentation, erosion of institutions and slower growth (Easterly at al. 2006)*) See Autor, Levy & Murnane 2003, Acemoglu & Autor 2012, Levy & Murnane 2004, Manning 2004

Motivation

34Slide35

35Slide36

36Slide37

Are unskilled wages too high in Hungary?It seems they are not. On the contrary.

37Slide38

Poor state of health?Several proxies of health status. Hungarian are worse off relative to Italians but not to Norwegians.The relative health status of the low educated is not worse in HungaryThe interaction effect of poor health and low education is positive for most cases in Hungary. Following Norton, Wang and Ai (2004):

38Slide39

Informal work?The employment data are self-reported, fall very close to the ILO-OECD figures.ILO-OECD employment comprises a large part of unregistered employment in Hungary: it is higher by 15-20 per cent than registered employment (Benedek, Elek & Köllő 2013)Indirect methods of estimating black work (unobserved even in the LFS) conclude that low educated/unemployed people earn less in the black economy on average than do more educated/employed people (Benedek et al 2012, Molnár & Kapitány 2012)

39Slide40

A Roma problem?The problem is similar in other CEEs with no or small Roma populationThe employment rate of the low-educated non-Roma (60% for prime-age unskilled males) is still much lower than in either Norway (85%) or Italy (80%)One coud probably single out a minority in Italy (perhaps less so in Norway) with a social standing and employment rate similar to those of Hungarian Roma.

40Slide41

41Slide42

Italy: low educated workers are over-represented in small firms in the domain of relatively complex jobs

42Slide43

43Slide44

Let

yij denote the expected productivity yield of j

-educated workers (j=1,2,…,J) when employed in job type

i (i=

1,2,…,I), and the wj

-s their reservation wages, assumed to vary with educational attainment but not with the type of job

.

Assuming that wages are set as a weighted average of reservation wages and the productivity yield of a given match – with 0



1 standing for the relative bargaining power of employers in a country or region – the firm solves:

Suppose that job types can be characterized with a continuous or ordinal measure of complexity (

R

) so that

R

1

< R

2

<…<

RI, and that the productivity yields from employing a j-educated worker in a job of R-level complexity can be approximated with the linear projection yij= jRi.. Equation (1) can be re-written as:When employers decide on hiring an individual their choices are based on wages and expected productivity that they predict on the basis of the applicant’s education and further proxies of his/her skills. These may be observed by the employer but not by the researcher and are therefore summarized in a residual term  satisfying E()=0, cov(,w)=0 and cov(, R)=0. For an applicant of j-level education expected profit is:For an applicant for the same job with education J:Subtracting 3b from 3a we have: and the probability that J is chosen for job type i is:

This is a McFadden model with a need to observe/estimate reservation wages, which furthermore have to vary within educational levels (by region or country). Given the quality of the wage data and our wish to estimate country-by-country, a more parsimonious model is to be chosen

 binary or multinomial logit/probit

44Slide45

Table x: Job characteristics and the probability of employing a low-educated worker

Average partial effects after logit in ALL (per cent)Dependent variable: the worker employed in the job has primary education attainment

Weighted samples, robust standard errors

Unweighted samples, bootstrap standard errors

Norway

Italy

Hungary

Norway

Italy

Hungary

Small firm

-1.01

5.58

***

-2.54

***

1.39

6.62

***-2.46***(0.67)(4.43)(3.51)

(1.06)

(7.83)

(5.07)

Firm size unknown

..

-3.94

*

0.87

..

-2.49

*

-0.14

(1.88)

(0.81)

(1.82)

(0.17)

Age of the match

0.82

***

0.65

***

0.13

0.80

***

0.60

***

0.11

***

(15.38)

(12.82)

(2.73)

(18.36)

(16.22)

(3.02)

Type 1 literacy tasks

-3.49

***

-5.21

***

-2.71

***

-4.67

***

-5.16

***

-2.12

***

(12.59)

(26.56)

(17.33)

(27.98)

(33.62)

(15.24)

Type 2 literacy tasks

1.15

***

1.28

***

0.45

*

1.46

***

1.29

***

0.30

(2.64)

(3.35)

(1.68)

(4.65)

(4.42)

(1.26)

Part-time job

-1.36

-5.74

***

-2.29

**

0.81

-6.04

***

-1.22

*

(1.03)

(4.84)

(2.47)

(0.78)

(7.19)

(1.68)

Managerial job

-7.92

***

-11.5

***

-14.2

***

-8.41

***

-7.70

***

-11.53

***

(2.87)

(3.77)

(12.76)

(4.12)

(4.12)

(12.90)

Other skilled job

-8.58

***

-9.48

***

-6.54

***

-8.07

***

-9.04

***

-5.37

***

(6.33)

(7.98)

(12.97)

(8.71)

(11.17)

(13.70)

Semi-skilled job

1.09

8.86

***

-0.43

9.70

***

8.93

***

-1.88

*

(0.37)

(3.24)

(0.28)

(4.11)

(4.56)

(1.84)

Observations

3618

3179

2607

3618

3179

2607

Pseudo-R

2

0.1122

0.2200

0.1997

0.1627

0.2327

0.1824

Observed share (per cent)Z-values based on robust and bootstrapped standard errors (with 100 replications) in parentheses. Significant at the *) 0.1 **) 0.05 and ***) 0.01 level. Reference categories: large firm (>20 workers), full-time job (36 or more hours a week), elementary occupations.

45Slide46

Size effects and share effects ( and )Jobs distinguished by the number of Type 1 tasks

46Slide47

Size effects and share effects ( and )Jobs distinguished by the number of Type 1 tasks

47Slide48

Size effects and share effects ( and ) Jobs distinguished by the number of Type 1 tasks

48Slide49

Size effects and share effects ( and ) Jobs distinguished by the number of Type 2 tasks

49Slide50

Size effects and share effects ( and ) Jobs distinguished by the number of Type 2 tasks

50Slide51

Size effects and share effects ( and )Jobs distinguished by the number of Type 2 tasks

51Slide52

Size effects and share effects combined () Jobs distinguished by the number of Type 1 tasks

52Slide53

Size effects and share effects combined () Jobs distinguished by the number of Type 2 tasks

53Slide54

Do Italian small firms tolerate skills mismatch (many literacy tasks  low education) more than do their larger counterparts?If yes, we would expect a positive interaction effect of literacy tasks (R) and small size. As R

increases, the L share should drop less in small than in large firms.Warning: the interaction effect of a continuous (x1) and a dummy variable (x2) in logit is not equal to 12 of the interaction term (Ai & Norton 2003). The cross-derivative can be calculated as below. The effects and significance differ from case to case depending on Xi

Re-estimating the equation with an interaction term yields



54Slide55

Simple jobs

55Slide56

Simple jobs

56Slide57

Simple jobs

57Slide58

Test performance

Low-educated

High-educated=1

(per cent, year)

(ratio)

NO

IT

HU

NO

IT

HU

Mean score

Prose

258

204

240

0.87

0.81

0.86

Document

259

201

235

0.86

0.81

0.85

Numeracy

254

210

238

0.87

0.83

0.84

Problem solving

248

199

..

0.85

0.81

..

Fractio

n

of

low-performers

c

Prose

21.4

65.7

35.8

4.86

2.07

3.09

Document

25.4

67.1

41.3

5.18

1.91

3.15

Numeracy

25.1

62.3

36.4

3.51

2.32

4.61

Problem solving

51.3

84.9

64.2

2.97

1.54

1.84

a) Employment rate times the mean number of Type 1 (Type 2) job-related literacy tasks performed by those in employment

b) Mean of the five plausibe values.

Sampling and imputation errors t.b.a.

c) Level 1-2 out of 5

Norway: test scores are much higher in absolute but not in relative terms

58Slide59

The large difference is in  rather than . But the abundance of very simple jobs largely contribute to unksilled employment in Italy

HH

II

H

I

59Slide60

 by particular tasks

60Slide61

Share of the low-educated by particular tasks

fp7_egyenkent.dta61Slide62

 by particular tasksfp7_egyenkent.dta

62Slide63

63Slide64

Type 1 and Type 2 requirements in the pooled sample

64