Unemployment Insurance in Europe unemployment duration and subsequent

Unemployment Insurance in Europe: unemployment duration and subsequent employment stability EPUNet Conference 8 -9 May 2006 Barcelona Konstantinos Tatsiramos IZA, Bonn 1

Introduction • Unemployment is a major issue especially in Europe: – High unemployment rates – EU (9%), US (5. 8%) (1990 -2000) – High incidence of long-term unemployment – EU (40%), US (<10%) • Unemployment has economic and social consequences (loss of skills, loss of income, poverty) • EU “Lisbon Strategy” (2000): – achieve full employment by creating more and better jobs – attain higher geographic and occupational mobility – improve skills and information on job opportunities • Policy recommendations stress the importance of structural reforms: – To make the labour market more flexible – Work more attractive – Develop human resources 2

Unemployment Insurance (UI) • Supports the income of unemployed while searching for a new job • Negative features of UI system: – exert lower search effort - higher reservation wages – longer unemployment duration –> long-term unemployment • Positive features of UI system – increase effectiveness of search activities – afford search costs – increase efficiency of job matching – obtain the “right job” – “Benefits allow the unemployed to search for a job which matches their abilities, rather than being forced by financial hardship into accepting the first available job offer “ Source: OECD Employment Outlook (1996) 3

Motivation • The negative features have dominated the literature and reform proposals, while little is known on the positive features of UI • The empirical literature has focused on the effect of UI on the exit rate from unemployment emphasizing the disincentive effects of UI • Studies on the effect of UI on post-unemployment outcomes have focused on re-employment wages, while subsequent employment duration has been neglected • There is no empirical evidence for Europe 4

Empirical Question • Is there an effect of benefits on post-unemployment stability? • The focus is on employment duration as a measure of employment stability – Based on Jovanovic (1979) a good match lasts longer – Akerlof, Rose and Yellen (1989) provide evidence that nonpecuniary match aspects reduce the quit probability • Hypothesis: If UI has a post-unemployment beneficial effect then we expect higher employment stability for recipients 5

Preview of Results • Short run effect (on unemployment duration): Benefits reduce the exit rate from unemployment – Expected unemployment duration of recipients is higher than nonrecipients – This is in line with the conventional results both in theoretical and empirical literature • Long run effect (on employment duration): Benefits reduce the exit rate from subsequent employment – Recipients remain employed on average 4 months longer, which represents a 20% increase relative to the average employment duration. This partly offsets the short run effect of benefits – This effect is pronounced (a) in countries with relatively more generous benefits and (b) for recipients who have stayed for 6 months or more in unemployment 6

Plan of Talk • Theoretical Background • Empirical Literature • Data Description • Econometric Methodology • Empirical Results • Conclusions 7

Theoritical Framework (1) • Convention: Benefits create disincentives • Labor – Leisure model: Ø Lower opportunity cost of leisure (benefits ‘’subsidise’’ leisure) • Job Search models: Lower opportunity cost of unemployment Ø Lower search effort Ø Higher reservation wages Ø • Both models predict longer unemployment duration 8

Theoritical Framework (2) Other results: Benefits might improve post-unemployment outcomes • Higher reservation wages lead to higher re-employment wages • More efficient search leads to increased quality of job match Ø Benefits as a ‘’search subsidy’’ as they lower opportunity cost of search (Burdett, 1979) Ø Increased post-unemployment job stability as unemployed get the time to find the ‘’right job’’ reducing job mismatch (Marimon&Zilibotti, 1999) 9

Empirical Literature • UI leads to longer unemployment duration Ø Atkinson&Micklewright (1991), Devin&Kiefer (1991) • UI on post – unemployment wages Ehrenberg&Oaxaca (1976) – positive effect Ø Addison&Blackburn (2000) – weak effect Ø • UI on job stability Ø Belzil (2000) – positive but small effect (0. 5 -0. 9 days of expected job duration per additional week of benefits) • Future entitlement of UI on Employment Duration Baker&Rea (1998) – increase employment hazard Ø Jurajda (2002) – increase layoff hazard Ø 10

Data Description • European Community Household Panel (1994 – 2001) 60, 500 nationally representative households, 130, 000 individuals (16+) for the EU-15 • Countries considered in this study: Belgium, Denmark, France, Germany, Greece, Italy, Ireland, Portugal, Spain, and UK • Sample Strategy Ø Inflow sample of unemployed males, 18 -60 years old Ø Construct individual labour market histories using the calendar of activities (monthly information) Ø Multiple spells Ø Consider exits out of labour force as continued unemployment spells 11

Table 1. Transitions in the sample BEL DK FR GER GR # Un Spells 246 352 842 1150 971 Exit to Emp 170 (69. 11 %) 262 (74. 43 %) 569 (67. 58 %) 732 (63. 65 %) 759 (78. 17 %) Exit Back to Unem 62 (25. 20 %) 92 (26. 14 %) 223 (26. 48 %) 314 (27. 30 %) 483 (49. 74 %) IRE IT POR SP UK # Un Spells 441 1364 640 2324 537 Exit to Emp 333 (75. 51 %) 1012 (74. 19 %) 446 (69. 69 %) 1776 (76. 42%) 422 (78. 58 %) Exit to Unem 99 (22. 45%) 559 (40. 98 %) 172 (26. 88 %) 977 (42. 04%) 118 (21. 97 %) 12

• Information on benefits in ECHP 1) Question on receipt of benefits at interview 2 Question on amount of benefits received in previous year Use both these variables to construct a benefit indicator and benefits duration • Benefit Indicator : I(b≥t) equals 1 if b≥t and 0, otherwise • Benefit duration (B) Ø Observed only if B<T. Otherwise B=T (Cannot estimate the effect of benefits as benefit termination gets closer) 13

Figure 1. Type of Spells 14

Table 2. Summary Statistics France Germany Italy Portugal Spain UK B NB B NB Number of Spells 479 363 758 392 280 1084 171 469 973 1351 187 350 Higher Education . 163 . 235. 164 . 163 . 039 . 045 . 018. 026. 101 . 142 . 438 . 480 Secondary Education . 447 . 377. 580 . 574 . 257 . 310 . 122. 119. 141 . 192 . 109 Less than Sec. Ed. . 390 . 388. 256 . 263 . 704 . 645 . 860. 855. 758 . 666 . 460 . 411 Age 34. 6 32. 4 39. 0 36. 7 39. 4 33. 7 40. 2 33. 8 37. 5 32. 6 36. 5 34. 9 Married . 399 . 369. 626 . 559 . 704 . 433 . 719. 458. 659 . 391 . 481 . 460 Number of Kids . 852 . 758. 767 . 719 . 936 . 567 . 848. 857. 906 . 662 . 936 . 869 15

Table 3. Characteristics of the Unemployment Insurance System Employment/Contributions conditions Replacement Rate (% of previous gross earnings) Duration Belgium 312 days in 18 months to 624 days in last 3 years, depending on age 60 No limit Denmark 52 weeks in 3 years 90 1+3 years France 4 months in 8 months 75 4 -60 (A) Germany 12 months in 3 years 60 12 -64 (A+C) 125 in 14 months 40 5 -12 (C) Italy 52 weeks in 2 years 30 180 days Ireland 39 weeks in 1 year Flat Rate 390 days Portugal 540 days in 2 years 65 12 -30 (A) 12 months in 6 years 70 4 -24 (C) 2 years Flat Rate 182 days Greece Spain United Kingdom Source: European Commission Missoc 1994. (A) denotes that duration depends on age, (C) on contributions 16

17

18

19

Econometric Methodology • Two-state duration model – Hazard rate: the probability to exit a state at a certain point in time having being in that state up to that moment – States • Unemployment • Subsequent employment • Identification Assumption: Following (Arellano, Bover, Bentolila, 2002), we assume a two regime model Ø Effect of benefits on hazard is independent of remaining benefit duration • Econometric Issues: – Unobserved Heterogeneity – Endogeneity of Benefit Indicator 20

Multispell Mixed Proportional Hazard Model with flexible baseline hazard 21

22

23

24

• Estimations performed by country: – Unem. and Empl. separately without UH – Unem. and Empl. separately with UH – Unem. and Empl. jointly with correlated UH • Unemployment Specification (with Benefit Selection) – Dummy for receiving benefits • Employment specification: – Dummy for having received benefits during previous unemployment spell or – Dummy for having received benefits interacted with previous Unemployment Duration – Other controls: • education, age, married, spouse’s labor market status, number of kids, year dummies, regional unemployment rate 25

Results – Without Unobserved Heterogeneity BEL DEN FR GER GR IREL IT POR SP UK . 280 (. 079) -. 835 (. 126) -. 234 (. 052) -. 344 (. 106) . 115 (. 073) . 077 (. 195) . 017 (. 048) . 034 (. 166) Unemployment Duration Benefits -. 432 (. 185) -. 412 (. 143) -. 396 (. 088) -. 669 (. 083) . 050 (. 083) -. 510 (. 128) Employment Duration Benefits -. 329 (. 241) -. 054 (. 228) -. 039 (. 102) -. 048 (. 095) . 100 (. 072) -. 061 (. 193) Other Controls: education, age, married, spouse’s labor market status, number of kids, year dummies, regional unemployment rate 26

Results – With Unobserved Heterogeneity BEL DEN FR GER GR IREL IT POR SP UK -. 207* (. 125) -1. 204 (. 163) -. 460 (. 075) -. 345 (. 143) -. 036 (. 095) . 339 (. 228) -. 052 (. 062) -. 469 (. 175) Unemployment Duration Benefits -. 460 (. 225) -. 412 (. 143) -. 396 (. 088) -. 736 (. 113) -. 234* (. 141) -. 561 (. 199) Employment Duration Benefits -1. 832 -. 414* (. 310) (. 222) -. 232* (. 139) -. 193* (. 115) -. 028 (. 109) . 034 (. 194) Other Controls: education, age, married, spouse’s labor market status, number of kids, year dummies, regional unemployment rate 27

Results – With Unobserved Heterogeneity (JOINT) BEL DEN FR GER GR IREL IT POR SP UK -. 207* (. 125) -1. 031 (. 176) -. 358 (. 067) -. 344 (. 111) -. 404 (. 101) . 256 (. 215) -. 422 (. 064) -. 364 (. 177) Unemployment Duration Benefits -. 492 (. 193) -. 446 (. 147) -. 389 (. 091) -. 742 (. 095) -. 187 (. 142) -. 525 (. 199) Employment Duration Benefits -1. 539 -. 663 (. 284) (. 232) -. 352 (. 222) -. 376 (. 115) -. 148 (. 125) -. 065 (. 207) Other Controls: education, age, married, spouse’s labor market status, number of kids, year dummies, regional unemployment rate 28

Employment Duration: with UH and interactions (JOINT) BEL DEN FR GER GR IREL IT POR SP UK Benefits -2. 084 -. 284 (. 829) (. 720) -. 583 (. 252) -. 223 (. 312) -. 084 (. 289) -. 335 (. 446) . 289 (. 275) -. 003 (. 654) -. 441 (. 162) -. 427 (. 292) Benefits 1. 146 *Un Dur (. 814) (1 -6) -. 377 (. 767) . 168 (. 307) . 014 (. 321) -. 094 (. 276) . 221 (. 519) -. 811 (. 276) . 538 (. 664) -. 018 (. 173) . 326 (. 356) Benefits. 145 *Un Dur (. 880) (7 -12) -. 693 (1. 01) . 741* (. 429) -1. 232 (. 399) . 081 (. 336) 1. 235* (. 642) -. 514* (. 307) . 043 (. 748) . 122 (. 191) -. 573. 602 Other Controls: education, age, married, spouse’s labor market status, number of kids, year dummies, regional unemployment rate 29

Conclusions • We analyse empirically the effect of UI on unemployment and employment duration in a number of European countries • Results show that benefits: – Reduce the exit rate from unemployment – Reduce also the exit rate from subsequent employment • The effect is pronounced in countries with relatively more generous beneftis and • For recipients who have stayed unemployed for 6 months or more • Although the positive effect of benefits on employment duration does not compensate fully for the negative effect on unemployment duration, these results show that reforms proposals of the UI system should consider both effects 30