1 Financial work incentives and the long-term unemployed. 2 The effect of tax-benefit policy changes on the trilemma of welfare reform

Transcription

1 1 Financial work incentives and the long-term unemployed 2 The effect of tax-benefit policy changes on the trilemma of welfare reform Diego Collado May, 2018

2 1 Financial work incentives and the longterm unemployed

3 Introduction Illustration: budget constraint Net income (1 t ) $ Current financial gain of working 0 20h * wage Earnings (h * wage) 3

4 Introduction Increasing the often inadequate out-of-work incomes for long-term unemployed (LTU, >12 months) might decrease their likelihood of transitioning to work à I study effect of changes in financial work incentives on this likelihood If there was an effect, increasing in-work benefits (or other policies) would be necessary to not lower this likelihood. Since this is costly, the targeting of those benefits could be increased As this would affect marginal tax rates, I also study changes in hours worked by part-timers 4

5 Methodology and data Regression analysis: Extensive margin: Participation Tax Rate (PTR) = % of earnings taken away in tax paid and withdrawn benefits when moving to E! " #$%& ( #$ = Λ(,!./ #$ + 1 $ + 2 #$ #$ ) Intensive margin: h89:; #$ =, <=_(?./ #$ + 1 $ + 2 #$ #$ Data: 7 stacked 2-year transitions Creating EUROMOD longitudinal input files based on EU- SILC and simulating long-term UBs 5

6 Methodology: measuring incentives with participation tax rates (PTRs) PTR = proportion of household earnings taken in tax and withdrawn benefits when moving to employment E.g.: in year 0 gross earnings 2000 (100%), taxes 500 (=25%) and UB 1000 (=50%): PTR = = 75% = 1 = 1 25% (50%) 800 (40%) EMTRs measure the same when working +5% hours./0 = (hh. 456 hh );< + (hh hh. 456) 85A9D9@C 2000 (100%) 500 (25%) 2000 (100%) 500 (25%) PTR=75% PTR=65% ΔPTR= 10pp 6

7 Methodology: policy changes affecting incentives SC rebate (EUR 2012/month) SC rebate expansions Full-time equivalent monthly earnings (EUR 2012/month)

8 Methodology: policy changes affecting incentives Change in UB parameters in relation to changes in wage index (pp) Difference between the parameters of the 1st and 2nd year of a spell Max earnings to declare single Max earnings to declare head & cohabitating Replacement rate single Replacement rate cohabitating Max UB single Max UB head Max UB cohabitating Parameters of UBs that are the same since the 2nd year of a spell Max earnings to declare head & cohabitating Replacement rate single Replacement rate cohabitating Max UB single Max UB head and cohabitating

9 Methodology ΔPTR and ΔEMTR by groups that defined main policy changes LTU Mean ΔPTR single LTU<= Mean ΔPTR head LTU<= Mean ΔPTR cohabitee LTU<= Mean ΔPTR single LTU> Mean ΔPTR head LTU> Mean ΔPTR cohabitee LTU> Mean ΔPTR non-elig SC rebate Mean ΔPTR elig SC rebate Part-timers Mean ΔEMTR non-elig SC rebate Mean ΔEMTR elig SC rebate

10 Methodology Illustration of budget constraint change of hypothetical person Belgium : BE_2005_bl BE_2006_bl Single individual [Wage: /h] income (1 PTR) (1 PTR) (1 EMTR) net income benefits SCs gross earnings (monthly) 10

11 *** p<0.01 ** p<0.05 * p< Results (statistically significant and selected) MODEL (2) (2) (2) (3) DV: U->E DV: U->E DV: U->E DV: U->E logit logit logit ivprobit VARIABLES odds AME (contrasts) AME (levels) AME Change in PTR (10 pp) 0.136*** ** [ ] [ ] [ ] Change in PTR * U months *** [ ] [ ] Change in PTR * U > 24 months ** [ ] [0.013,0.109] [ ] Change in PTR * Female = *** [ ] [ ] Change in PTR * Female = Omitted [ ] [-0.000,0.081] [ ] Change in PTR * 1st eq. hh inc. tercile = *** [ ] [ ] Change in PTR * 1st eq. hh inc. tercile = [ ] [-0.017,0.083] [ ] Change in IV2 log eq. hh income ( 2012) Baseline probability 9% [ ] Pseudo-R N_sub

12 12 AME of PTR by categories ( interactions )

13 2 The effect of tax-benefit policy changes on the trilemma of welfare reform

14 2 The effect of tax-benefit policy changes on the trilemma of welfare reform Trilemma among redistribution, work incentives/employment and net revenue (Blundell, 2002; Adam, Brewer, & Shephard, 2006) What have been the direct (mechanical) and LS effects of tax-benefit policy changes on poverty and net revenue? Or what has been the Antipoverty Marginal Cost of Public Funds ( P z)/( R z)? Low road Tax-benefit High road Tax-benefit Trap road Tax-benefit LS LS Pgap Pgap Low/High road Tax-benefit LS Pgap Highway Tax-benefit 14 LS Pgap LS Pgap

15 Methodology: counterfactual decomposition Isolating the effect of policy change by putting no population change counterfactual distributions between the original cross-sectional ones (similar to crossed terms in Oaxaca) I = aggregate index such as poverty or net revenue d = tax-ben function p = tax-ben parameters y = gross income distribution α = inflation index (benchmark for uprating: average wage or CPI) Eg poverty:! " #$ % #$, ' #$! " )* % )*, ' )* = +{![" #$ (% #$, ' #$ )]![" #$ (% #$, 1' )* )]} other +{![" #$ (% #$, 1' )* )]! " )* % )*, ' )* } policy Superscripts refer to year and subscripts to LS behavior under that year tax-ben {![" #$ (% #$, ' #$ )]! " )* % )*, ' )* = +{![" #$ (% #$, ' #$ )]![" #$ (% #$, 1' #$ )]} other +{![" #$ (% #$, 1' #$ )]![" #$ (% #$, 1' )* )]} LS (2 nd order policy) +{![" #$ (% #$, 1' )* )]![" )* (% )*, ' )* )]} policy 15

16 Methodology ΔTax-benefit Policy effect LS model Policy effect ΔE LS effect LS effect ΔPgap Δ ΔOther (minimum wage, inequality, childcare, ALMP, wage subsidies, demography ) 16

17 Results: Decomposition of policy effects by taxbenefit component For JL per AROP person: 2005= Note: JL or W based on status at the moment of the interview. Anchored poverty line

18 Results Note: JL or W based on status at the moment of the interview. Anchored poverty line. 18

19 Results: full decomposition w/o LS effects Note: JL or W based on status at the moment of the interview. Anchored poverty line. 19

20 20 Methodology Cross-sectional model in levels:!"($ %& = 1) = * +,- /!01 %& + 3 & We can then predict the prob. of E for the pop. of 2005 with the PTRs of 2014 (only short-term UBs simulated). 1 st year UB parameters in relation to changes in wage index (pp) Year-to-year wage index growth Replacement rate single Replacement rate head Replacement rate cohabitating Max UB single Min UB single Max UB head Min UB head Max UB cohabitating Min UB cohabitating

21 Results: full decomposition with LS effects PTR (data 2005): 77% à 81% 1 st order: -1 /person pov costs 10 of revenue +2 nd order: -1 /person pov cost 25 of revenue Note: Probabilistic approach based on expected income 21

22 Main limitations and possible next steps Limitations LS model only simulates short-term UBs. In addition, in 2014 replacement rates for short-term U but for LTU No simulation of benefits in-kind nor special cases of UBs Possible next steps Improve probabilistic approach Simulating next year s UBs and calculating present value PTR (still might not capture changes in timing) Simulating long-term UBs assuming people observed U=12m à in flat zone of UBs (with longitudinal data, this was the case for large majority). 22

23 Recommendations (both papers) Stablishing a close link between changes in out-of-work benefits and in-work compensations. This particularly for groups that seem to be more sensitive to changes in work incentives. If out-of-work benefits were raised, to offset surges in expenditure, the targeting of current in-work compensations could be increased. This might have less of an effect in hours worked. Combine tax-benefit microsimulation and registry data (work in progress at the FPB) and simulating tax-benefits longitudinally. This to have retrospective data (e.g. E status, SCs, etc.) and future tax-benefits. 23

24 Thank you Questions, comments and suggestions? 24

25 B. Methodology: Regression analysis # $ %"&' ) %" = Λ(- #/0 %" +! " + 2 %" %" ) Control variables: Transition fixed effects! " controls for common changes (eg demand) Changes in: Region-age-education-specific employment rates Eq. hh incomes (income effects) 1st year (including interactions with ΔPTRs) U whole previous year or more Gender Cohabitation (observed) Income tercile Age Region (2nd year)

26 Results: descriptive statistics Extensive (LTU) Intensive (PT) Observations Weighted observations 1,321,174 4,356,634 Mean SD Mean U_E=1 Mean SD Transition U->E / ΔWeekly hours ΔPTR/IV1 EMTR (pp) T1 PTR/IV1 PTR (pp) U whole previous year or more Female Cohabitating T1 tercile ΔEmp. reg-age-edu (pp) IV2 ΔEq.inc. (monthly 2012) T1 weekly hours % with UB as out-of-work income 26

27 Ex-post LS model To isolate the effect of z, besides mechanical effect, one needs e.g. the counterfactual probability of employment for the population of 2005 with MTRs of Regression with group-average simulated IV (=keeping the composition groups as in e.g and travel through several policy years recalculating the averages). Uses z (and other zs between more years) for identification. 1 st stage: ()* +, =. /0_ , + : < 2 nd stage: (@(B +, = 1) =. 234 ()* D +, + E, + E 8 + : < Probabilistic simulation approach: B F GH_GH_IJK +, = L 1 GH +M F GH +, + L 0 GH OP, GH F +, B F GH_QR_STG +, = L 1 GH +M F QR +, + L 0 GH OP, QR F +, B F GH_QR_STQ +, = L 1 QR +M F QR +, + L 0 QR OP, QR F +, Policy effect LS effect 27

28 Ex-post LS model IV_PTR based on fam_type, wage_quant and sex ivprobit E (PTR_10pp = IV_PTR_10pp) i.year i.fam_type i.wage_quant i.sex age First-stage regression summary statistics Adjusted Partial Robust Variable R-sq. R-sq. R-sq. F(1,26162) Prob > F PTR_10pp Delta-method dy/dx Std. Err. z P> z [95% Conf. Interval] PTR_10pp Wald test of exogeneity (/athrho = 0): chi2(1) = 3.50 Prob > chi2 =