Mechanisms (making sense of the findings) Labor demand story? ◮ Employers exploit confusion of the old regime and capture part of the transfer → Result driven by new hires rather than incumbents → Result driven by small and incorporated firms → Anecdotal evidence on perceptions of transfers → Systematic violation of union’s CBA? ( in progress ) Labor supply story? ◮ Confused employees bargain more aggressively after the event (pay equity/fairness concerns) → Ruled out by immediate effect at t=0 and new hires → Also effect broken by firm exposure is not U-shaped 21 / 31
All workers vs incumbents j = − 6 γ j · d j G ¯ f , t = � 5 w f , t + µ f + µ t + ǫ f , t 15 All workers Constant pesos (base = Jan 2004) 10 5 0 Incumbents -5 -10 -5 -4 -3 -2 -1 0 1 2 3 4 Months relative to treatment Note: incumbents: workers present -7/+7 months around the event. 22 / 31
Mechanisms (making sense of the findings) Labor demand story? ◮ Employers exploit confusion of the old regime and capture part of the transfer → Result driven by new hires rather than incumbents → Result driven by small and incorporated firms → Anecdotal evidence on perceptions of transfers → Systematic violation of union’s CBA? ( in progress ) Labor supply story? ◮ Confused employees bargain more aggressively after the event (pay equity/fairness concerns) → Ruled out by immediate effect at t=0 and new hires → Also effect broken by firm exposure is not U-shaped
Small vs non-small firms j = − 13 γ j · d j D w f , t = � 24 f , t + µ f + µ t + ǫ f , t 20 Small [<10] 15 Constant pesos (base = Jan 2004) 10 5 0 Large [10+] -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: firm size is the average number of employees from t-12 to t-1. 23 / 31
Incorporated vs Unincorporated firms j = − 13 γ j · d j D w f , t = � 12 f , t + µ f + µ t + ǫ f , t 20 Incorporated 15 Constant pesos (base = Jan 2004) 10 5 0 Non-incorporated -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: first two digits of tax id determine if firm is incorporated or not. 24 / 31
Incorporated firms: small vs non-small j = − 13 γ j · d j D w f , t = � 12 f , t + µ f + µ t + ǫ f , t 40 Small [<10] Constant pesos (base = Jan 2004) 30 20 10 0 Large [10+] -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: small firms have less than ten employees [ < 10] while large more [10+]. 25 / 31
Heterogeneities Incorporated Small Large Non Small Large [ < 10] [10+] Incorpo Incorpo [ < 10] [10+] (1) (2) (3) (4) (5) (6) Reduced form ∆ monthly wage 9.86*** 4.94*** 0.81 11.37*** 19.27*** 5.73*** (in pesos) (1.96) (1.55) (1.74) (1.72) (3.30) (1.81) First stage ∆ transfer -97.02*** -82.62*** -96.86*** -87.52** -94.32*** 81.97*** (in pesos) (0.54) (1.55) (0.65) (0.41) (0.75) (0.40) 2sls ∆ wage -0.10*** -0.06*** -0.01 -0.13*** -0.20*** -0.07*** ∆ transfer ( τ e ) (0.02) (0.02) (0.02) (0.02) (0.04) (0.02) Number of firms 20,253 15,534 13,029 22,758 9,843 12,915 Observations 1,694,509 1,367,361 1,080,767 1,981,103 833,347 1,080,767 Note: Standard errors clustered at firm level in parenthesis. G w f , t = β 1 Window f , t + β 2 · Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + µ f + µ t + ǫ f , t , where Window is an indicator for the event window . 26 / 31
Mechanisms (making sense of the findings) Labor demand story? ◮ Employers exploit confusion of the old regime and capture part of the transfer → Result driven by new hires rather than incumbents → Result driven by small and incorporated firms → Anecdotal evidence on perceptions of transfers → Systematic violation of union’s CBA? ( in progress ) Labor supply story? ◮ Confused employees bargain more aggressively after the event (pay equity/fairness concerns) → Ruled out by immediate effect at t=0 and new hires → Also effect broken by firm exposure is not U-shaped
Anecdotal evidence on recipient’s perception (1) 1. Quote from a book on social security: “... the old system (SFC) blurred the image of the State as responsible for it . (...) The roles are confused. People consider that these benefits integrate their salary and that employers are responsible for them . They even ignore that it is the State that pays for them ...” Note: “ Pol´ ıticas de Protecci´ on familiar, R´ egimen de Asignaciones Familiares y principales planes sociales en la Rep´ ublica Argentina ”, CIESS (2007). 27 / 31
Anecdotal evidence on recipient’s perception (2) 2. Survey evidence (phone 2018 - SSA) Who is the responsible of paying the transfer (FA)? Answer type A. Government 35.4% B. Employer 8.6% C. Other 4.0% D. Don’t know 52.0% Source: based on a SSA report (Cruces, 2019). 28 / 31
Mechanisms (making sense of the findings) Labor demand story? ◮ Employers exploit confusion of the old regime and capture part of the transfer → Result driven by new hires rather than incumbents → Result driven by small and incorporated firms → Anecdotal evidence on perceptions of transfers → Systematic violation of union’s CBA? ( in progress ) Labor supply story? ◮ Confused employees bargain more aggressively after the event (pay equity/fairness concerns) → Ruled out by immediate effect at t=0 and new hires → Also effect broken by firm exposure is not U-shaped
Horizontal equity 0.05 0.00 2sls coefficient -0.05 -0.10 -0.15 Exposure 0.35 0.43 0.48 0.52 0.55 0.60 0.65 0.71 -0.20 0-30% 10-40% 20-50% 30-60% 40-70% 50-80% 60-90% 70-100% Lowest Highest share share Treated workers Note: each dot refers to a different regression of the type: G w f , t = β 1 Window f , t + β 2 · Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + µ f + µ t + ǫ f , t . Exposure Density 29 / 31
Misc Long-run wage effects : 1. Mean wage 2. p 25 and p 75 3. Pure event (levels) Other real outcomes : 4. Composition 5. Employment 6. Credit score (debt) 30 / 31
Conclusions ◮ ∆ in the remittance system (from employer to govt): ◮ Wages ↑ ∼ 2% after firms switch to the new system Pass-through: under old payment system employers capture ∼ 10 / 20% of the transfer by paying lower wages ◮ Welfare improving reform from worker’s point of view ◮ The way transfers are disbursed matters and affects the final incidence (not captured dollar for dollar by workers) ◮ These results raise questions about the use of employers as intermediaries to disburse benefits; less salient schemes may lead to capture by employers 31 / 31
Thanks!
Persistent inflation throughout the period 4,000 400 CPI RIPTE (current Argentinean pesos) 350 3,000 CPI (index) 300 250 RIPTE 2,000 200 150 1,000 Jan03 Apr04 Jul05 Oct06 Jan08 Apr09 Jul10 Date Note: CPI denotes consumer price index while RIPTE to the average salary of Back FA registered workers (in current pesos).
Evolution of FA brackets and minimum wage 5,000 Top bracket 4,000 2nd Current Argentinian pesos 3,000 3rd 2,000 1,000 Minimum Wage 0 Dec02 Mar04 Jun05 Sep06 Dec07 Mar09 Jun10 Date Back FA Note: brackets are updated roughly once per year, but no in fix intervals.
Distribution of monthly wage Back FA 40,000 80 Density (left) Transfer (right) Notch 1 Notch 2 Notch 3 30,000 60 Child benefit (in pesos) Wage Earners 20,000 40 10,000 20 0 0 0 500 1,000 1,500 2,000 2,500 Gross Monthly Wage (pesos) Note: figure corresponds to May’04; employees w/ kids working for 12 months. Notch 1 is located at p40, Notch 2 is located at p70, Notch 3 is located at p80.
1st stage (within firm T-C): beneficiaries Back FA 0 Share receiving transfers (p.p.) -.05 -.1 -.15 Point estimate 95% C.I. -.2 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to event (turn 18) Note: event study when a kid turns 18 and workers lose eligibility.
1st and 2nd stage (within firm T-C) Back FA 40 30 20 Constant Pesos 2004 10 0 -10 -20 -30 Child transfer Wage earnings -40 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to event (turn 18) Note: event study when a kid turns 18 and workers lose eligibility.
Collective bargaining agreements (CBA) Jun03 Jul10 300 Number of agreements 200 100 0 Jan03 Apr04 Jul05 Oct06 Jan08 Apr09 Jul10 Date Note: collective bargaining agreements occur every month. Two-thirds of them are Back FA firm level agreements.
Average # workers by switching date Back 140 Aug'08 120 100 Avg. N of employees 80 60 40 20 0 2003m7 2005m1 2006m7 2008m1 2009m7 2011m1 Switching date
Explanation of no-results for switchers after July 2008 ◮ Drop in economic activity (figure EMAE/GDP growth) ◮ Stabilization of employment growth (figure wage earners) ◮ Interesting if the effect comes from new hires. ◮ Lower ATR? (figure ratio transfer/min W) ◮ This doesn’t seem to be the main raison. Back FA Back Freq
Large drop in economic activity (EMAE) 140 Monthly economic activity estimator 130 (base =2004) 120 110 100 90 Jan04 Jan05 Jan06 Jan07 Jan08 Jan09 Jan10 Jan11 Date Note: large drop in economic activity from August 2008 onwards. Back FA Back Freq Back RB
Stabilization of private employment growth 7.00 Private employees (in millions) 6.00 5.00 4.00 3.00 2003Q1 2004Q3 2006Q1 2007Q3 2009Q1 2010Q3 Date Note: stabilization of employment in the third quarter of 2008. Back FA Back Freq Back RB
Evolution of the ATR ( ∼ ratio transfer/min wage) 0.20 Ratio transfer/minimum wage 0.15 0.10 0.05 1 st 2 nd 3 rd bracket 0.00 Jul03 Jul04 Jul05 Jul06 Jul07 Jul08 Jul09 Jul10 Date Note: ATR remains roughly constant during the period of analysis. Back FA Back Freq
Transfer saliency in payslip xxxxPaid by employers (SFC)xxxxxxxxxxxPaid by govt (SUAF) Go back
Reform’s motivation According to the Law ◮ Make sure beneficiaries receive the transfer ◮ Transparency and efficiency reasons ◮ Financial relief for firms Government was pushing for this to happen. − → Identified some cases of fraud. Didn’t find anecdotal evidence on Unions asking for this reform. Go back
Macro roll-out (official budget information) 0.80 0.60 Share paid through SFC 0.40 0.20 0.00 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Year Go back Note: gradual decline in the share of FA paid through the old system (SFC).
Total spent and aggregates comparison (SFC+SUAF) 4,000 FA spending (million pesos from 2003) 3,500 3,000 2,500 2,000 1,500 Source of data: Macro aggregates Microdata 1,000 2003 2004 2005 2006 2007 2008 2009 2010 Year Note: (a) increase in FA payments as time passes, (b) replicate macro aggregates Go back using micro-data.
Beneficiaries (number of children) Go back 3.5m 3m Number of children (in millions) 2.5m 2m 1.5m 1m 0.5m 0 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Year Note: N children receiving the transfer increases (economy booming + formalization).
The reform (3) Table: Key dimensions under the two payment systems SFC SUAF (1) (2) Legal liability Employee ( ¯ τ ) Employee ( ¯ τ ) Employer ( τ e ) Government ( τ g ) Remittance responsibility Information reporting Form 931 Form 931 Salience to employers High = Employees perception ( q ) Low ↑ Source of funding Contributory Contributory Employer SSC Employer SSC Note: no change in timing, transfer’s amount.
Incorporation schedule: memo (1) Go back (a) Memo (body text)
Incorporation schedule: memo (1) Go back (b) Memo annex (with employer identifiers)
The schedule ◮ We digitized 50+ schedule plans recovering approximately 60K firms with its corresponding “internal deadline”. ◮ Note: not all firms appeared in the schedules (or at least, in the publicly available ones). ◮ Less than 0 . 1% of firms appeared in more than one schedule. ◮ Compare deadline with the effective incorporation date. ◮ ∼ 90% of firms were incorporated before deadline.
Scheduled vs effective incorporation date (micro-data) 1.0 0.8 Deadline to be incorporated (ANSES' schedule) 0.6 CDF 0.4 0.2 0.0 -20 -15 -10 -5 0 5 10 15 20 Months before/after deadline Go back Note: ∼ 90% of firms were incorporated before government’s deadline.
Formal approval: memo (2) Go back (a) Memo (body text)
Formal approval: memo (2) Go back (b) Memo annex (with employer identifiers)
Formal approval: memo (2) ◮ Difficult to track the universe of approval memos ◮ However, it was possible to do a public query so as to check the formal incorporation date ◮ Random sample of 300 firms ◮ Compare formal incorporation vs effective incorporation date ◮ ∼ 80% firms incorporated in the same date as formal approval ◮ No incentives to delay: ◮ Cannot compensate the money (+ inflation) ◮ Workers’ notification: Within ten days after the switch, firms should inform their workers about the new payment mechanism and the overall scheme of the FA system
Roll out query Go back
Formal inclusion and observed incorporation (micro-data) 1.00 0.80 Cum share 0.60 0.40 0.20 0.00 -20 -16 -12 -8 -4 0 4 8 12 16 20 Months Before/after formal incorporation Note: ∼ 80% of firms were incorporated the same date as the formal approval. Go back
Notification to employees (affidavit) Go back Versión 1.3 Notificación del Régimen de Form. Asignaciones Familiares Sistema PS.2.61 Único de Asignaciones Familiares Frente 1 Este Formulario reviste carácter de Declaración Jurada y se debe completar en letra de imprenta, sin tachaduras ni enmiendas RUBRO I – DATOS DEL TRABAJADOR (a completar por todos los trabajadores con o sin cargas de familia) Apellido y Nombre Completo Fecha de Nacimiento Nacionalidad Cuil Tipo y Nº Doc /CUIL Sexo Estado Civil Domicilio - Calle - Nuemero Piso Depto. Código Postal Localidad Provincia Teléfono Dirección de Correo Electrónico RUBRO I I – DATOS DEL EMPLEADOR Razón Social CUIT Domicilio - Calle - Nuemero Piso Depto. Código Postal Localidad Provincia Teléfono Dirección de Correo Electrónico Dejo constancia, por medio de la presente, que en el día de la fecha, me he notificado de las normas básicas y principales derechos que me asisten con relación al Régimen de Asignaciones Familiares y que surgen del cuadro existente al dorso de la presente, recibiendo copia, en este acto, de la Ley Nº 24.714, sus normas reglamentarias y de la Resolución ANSES Nº 292/08 y sus modificatorias. Asimismo, me notifico que los trámites para solicitar la liquidación y pago de las Asignaciones Familiares que me correspondan deberé realizarlos personalmente o a través de un “Representante” designado por mí para tal fin, dentro de los plazos que surgen del cuadro existente al dorso de la presente, en cualquiera de las Unidades de Atención de ANSES, presentando -cuando corresponda-, debidamente confeccionados, los Formularios respectivos y la documentación que en cada caso se detalla, además de la que adicionalmente me pudiera ser requerida. Tomo conocimiento, además, que cualquier reclamo deberé formularlo personalmente ante ANSES dentro de los plazos de caducidad establecidos por la normativa vigente, presentando el Formulario PS.2.72 “Reclamos Generales para los Sistemas SUAF y UVHI”, debidamente cumplimentado. Dejo constancia también, que asumo el compromiso de notificar a mi empleador toda novedad/modificación que se produzca con relación a mis cargas y relaciones de familia, acompañando la documentación que las acredite, a efectos de que éste las informe a ANSES a través del Programa de Simplificación Registral. Me comprometo a informar a ANSES el medio de pago a través del cual deseo percibir las Asignaciones Familiares. Finalmente me notifico que todos los datos que aporte a ANSES personalmente, a través de un “Representante” o de mi Empleador, para la percepción de las Asignaciones Familiares, tendrán carácter de Declaración Jurada, reconociendo el derecho de ANSES a reclamarme su restitución o compensar automáticamente los importes con otras asignaciones en caso de percepción indebida de mi parte, sin necesidad de notificación previa por parte del citado Organismo. Localidad, .................... de .............................……. de ..........…… Firma/Aclaración de Firma Firma/Aclaración de Firma y del Trabajador Sello del Empleador
Sample construction ◮ Panel dimension 1. Identify firms belonging to the short panel (unbalanced) i.e., exist 12 months around the event (6 before and 6 after). 2. Identify firms belonging to the full panel i.e., 96 months (period 2003-2010). ◮ Event identification 1. Identify latest FA payment during the period (2003-2010). 2. Look at what happened 6 months before the event and identify those that paid during all these months. Repeat the analysis with 4, 5, 7 and 8 months. ◮ Checks 1. Balanced panel � 2. Sensitivity FA payments � Go back
First difference model Back Outcome variable: monthly wage used as base for employers’ SSC. We get the mean wage for each firm-group-month and take the difference across groups G ¯ w w T w C f , t = ¯ f , t − ¯ f , t Then, for each firm we have a time series of first differences → run regular event study specification (f: firms, t: month-year) − 12 � G ¯ w γ j · d j f , t = α + f , t + ǫ f , t (1) j = − 13 ...
First difference model Back Outcome variable: monthly wage used as base for employers’ SSC. We get the mean wage for each firm-group-month and take the difference across groups G ¯ w w T w C f , t = ¯ f , t − ¯ f , t Then, for each firm we have a time series of first differences → run regular event study specification (f: firms, t: month-year) − 12 � G ¯ w γ j · d j f , t = α + f , t + ǫ f , t (1) j = − 13 ... 12 � w mean γ j · d j = β T i , f , t + f , t · T i , f , t + µ f , t + ǫ i , f , t (2) i , f , t j = − 13
Empirical approach: Reduced form Back In levels 5 � γ j · d j w mean = β T i , f , t + f , t · T i , f , t + µ f , t + ǫ i , f , t (3) i , f , t j = − 6 Pooled γ s: Alt (1) w mean = β 1 T i , f , t + β 1 T i , f , t · Window f , t + β 2 T i , f , t · Window f , t · Post f , t + i , f , t (4) β 3 T i , f , t · (1 − Window f , t ) · Post f , t + µ f , t + ǫ i , f , t Pooled γ s: Alt (2) w mean = β T i , f , t + β − 6 d − 6 + β 5 d 5 f , t · T i , f , t f , t · T i , f , t (5) i , f , t � �� � � �� � d n 6 d 5 β after d [0;4] · T i , f , t + µ f , t + ǫ i , f , t f , t � �� � d after Note: in (3) j = − 1 omitted category; in (5) months [-5;-1] are omitted. Window = 1 if months [-5;4].
Empirical approach: Reduced form Back First difference 5 � γ j · d j G w f , t = α + f , t + ǫ f , t (6) j = − 6 γ ’s should be numerically the same as those estimated in eq.(3) Pooled γ s: Average G w before = G w before = ( γ − 5 + γ − 4 + γ − 3 + γ − 2 + 0) / 5 Average G w after = G w after = ( γ 0 + γ 1 + γ 2 + γ 3 + γ 4 ) / 5 Average effect = G w average = G w after − G w before Note: in (6) j = − 1 omitted category.
Empirical approach: Reduced form Back First difference 5 � γ j · d j G w f , t = α + f , t + ǫ f , t (7) j = − 6 γ ’s should be numerically the same as those estimated in eq.(3) Pooled γ s: Alt (1) G w f , t = α + β 1 Window f , t + β 2 Window f , t · Post f , t (8) + β 3 (1 − Window f , t ) · Post f , t + ǫ f , t Pooled γ s: Alt (2) + β after d [0;4] f , t = α + β − 6 d − 6 G w + β 5 · d 5 + ǫ f , t (9) f , t f , t f , t ���� ���� � �� � d 5 d n 6 d after Note: in (6) j = − 1 omitted category; in (8) months [-5;-1] are omitted. Window = 1 if months [-5;4].
Empirical approach: Reduced form Back Expected values: Alt (1) 1. E ( G w / Win = 0 , post = 0) = α 2. E ( G w / Win = 1 , post = 0) = α + β 1 3. E ( G w / Win = 1 , post = 1) = α + β 1 + β 2 4. E ( G w / Win = 0 , post = 1) = α + β 3 Expected values: Alt (2) 1. E ( G w / Win = 0 , post = 0) = α + β − 6 2. E ( G w / Win = 1 , post = 0) = α 3. E ( G w / Win = 1 , post = 1) = α + β after 4. E ( G w / Win = 0 , post = 1) = α + β 5
First difference model Back Switch G w Window = 0 Window = 1 Window = 1 Window = 0 post = 0 post = 0 post = 1 post = 1 -5 -1 0 4 Distance Event window
Empirical approach: 2sls Back First stage G Transfer = α + δ 1 Window f , t + δ 2 Window f , t · Post f , t f , t + δ 3 (1 − Window f , t ) · Post f , t + ǫ f , t Reduced form G w f , t = α + β 1 Window f , t + β 2 Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + ǫ f , t 2sls (Wald estimator) → Θ = β 2 then we should have − δ 2 Note: if I drop the binned points I get numerically the same number.
Empirical approach Back First difference with firm and time FE 5 � γ j · d j G w f , t = f , t + µ f + µ t + ǫ f , t (10) j = − 6 Pooled γ s: Alt (1) ... closer but not exact G w f , t = β 1 Window f , t + β 2 · Window f , t · Post f , t (11) + β 3 (1 − Window f , t ) · Post f , t + µ f + µ t + ǫ f , t Pooled γ s: Alt (2) ... closer but not exact + β after d [0;4] G w f , t = β − 6 d − 6 + β 5 · d 5 + µ f + µ t + ǫ f , t (12) f , t f , t f , t ���� ���� � �� � d 5 d n 6 d after Note: in (10) j = − 1 is the omitted category; in (12) months [-5;-1] are the omitted ones. Window = 1 if months[-5;4].
Empirical approach First difference 2sls with firm and time FE Reduced form (z = instrument) + γ after d [0;4] G w f , t = γ − 6 d − 6 + γ 5 · d 5 + µ f + µ t + ǫ f , t f , t f , t f , t ���� ���� � �� � d n 6 d 5 d after First stage: + γ after d [0;4] G transfer = γ − 6 d − 6 + γ 5 · d 5 + µ f + µ t + ǫ f , t f , t f , t f , t f , t ���� ���� � �� � d 5 d n 6 d after Note: the ratio reduced form/first stage is close to the Wald estimator where, again, the difference is due to the controls.
Empirical approach: Heterogeneities Reduced form G w f , t = α + β 1 Window f , t + β 2 Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + + β 4 High f + β 5 Window f , t · High f + β 6 Window f , t · Post f , t · High f + β 7 (1 − Window f , t ) · Post f , t · High f + ǫ f , t If we were to include firm and time FE ( µ f and µ t ), then we need to add also the interaction between µ t and High f . β 4 disappears as it is absorbed by µ f . Specification 2sls with High/low:
Robustness checks Back ◮ Sample sensitivity ◮ Balanced panel (96 months) ◮ Sensitivity FA payments ◮ Results not driven by modeling choices: ◮ � = specs: Simple mean, firm & time FE, firm linear trends ◮ Event window size: 10, 12, 14,... 24 months, etc. ◮ Event by event ◮ Sensitivity to the treatment group definition ◮ Fully treated vs never/partially treated (life events). ◮ Treatment status based on year of birth.
Balanced panel j = − 13 γ j · d j D w f , t = � 12 f , t + µ f + µ t + ǫ f , t 15 Constant pesos (base = Jan 2004) 10 5 0 -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: results remain unchanged when considering the balanced panel. Back
Sensitivity FA payments (2sls) 0.06 0.03 0.00 2sls coefficient -0.03 -0.06 -0.09 -0.12 -0.15 4 5 6 7 8 Sensitivity FA payments Note: each dot refers to a different regression of the type: G w f , t = β 1 Window f , t + β 2 · Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + µ f + µ t + ǫ f , t , where we vary the sample according to FA payments restriction. Back
Regression specification (mean) (1) (2) (3) Reduced Form ∆ monthly wage 6.94*** 7.71*** 5.40*** (in pesos) (0.90) (1.25) (1.27) 2sls ∆ wage -0.08*** -0.08*** -0.06*** ∆ transfer ( τ e ) (0.01) (0.01) (0.01) Simple mean difference � Firm and time FE � � Firm linear trend � Observations 3,061,870 3,061,870 3,061,870 Note: Standard errors clustered at firm level in parenthesis. Go back
Pure event for workers w/ and wo/ kids (mean) j = − 13 γ j · d j w f , t = � 12 ¯ f , t + µ f + µ t + ǫ f , t 20 Treat 15 Constant pesos (base = Jan 2004) 10 5 0 Control -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: increase in wage is driven by treated workers. Go back
Window’s dynamic (mean) 15 Two years Constant pesos (base = Jan 2004) 10 5 One year around event 0 -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: results remain unchanged for a time window of 6 months before/after. Back
Effect’s dynamics – 2sls -0.07 -0.08 2sls coefficient -0.09 -0.10 -0.11 -0.12 All post Last-half Only last periods post periods period [0:11] [6:11] [11] Post periods included Note: each dot refers to a different regression of the type: G w f , t = β 1 Window f , t + β 2 · Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + µ f + µ t + ǫ f , t , where we vary the post periods included in the event window ( W ).
Event by event – 2sls 0.050 0.000 2sls coefficient -0.050 -0.100 -0.150 Jul03 May04 Mar05 Jan06 Nov06 Sep07 Jul08 Excluded month Note: each dot refers to a different regression of the type: G w f , t = β 1 Window f , t + β 2 · Window f , t · Post f , t + β 3 (1 − Window f , t ) · Post f , t + µ f + µ t + ǫ f , t , Go back where we exclude switchers in a given month.
Rolling window of events (whole roll-out) 0.04 40 0.00 30 # of firms (in K) -0.04 2sls 20 -0.08 10 -0.12 -0.16 0 Jul03- May04- Mar05- Jan06- Nov06- Sep07- Jul08- Jun05 Apr06 Feb07 Dec07 Oct08 Aug09 Jun10 Rolling window of events (2-year window) Note: each dot refers to a different regression with a rolling window of events. Go back Go macro context
Alternative treatment group definition (mean) Workers that have at least one child born in [1992-2002] . This means that these workers are fully treated during the period 2003-2010 . January 2003 December 2010 Born in (age): Born in (age): 1992 (11) 1992 (18) 2002 (1) 2002 (8) The rest of the workers belong to the control group ( never treated and/or partially treated ).
Alternative treatment group definition (mean) 15 Constant pesos (base = Jan 2004) 10 5 0 -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: results hold when using an alternative definition of the treatment group. Go back
Alternative T group definition (mean) using year of birth 15 Constant pesos (base = Jan 2004) 10 5 0 -5 -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: results hold when using an alternative definition of the treatment group. Go back
Whole period Before crisis 15 Constant pesos (base = Jan 2004) 10 5 0 -5 After crisis -10 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: before crisis refers to firms switching into the new system before August 2008; Go back after between August 2008 and July 2010.
Other potential behavioral responses 1. Worker composition 2. Employment 3. Full vs part-time workers 4. Bunching at notches
[1.] Worker composition j = − 13 γ j · d j f , t = � 12 G N f , t + µ f + µ t + ǫ f , t 0.75 0.50 Gap in number of workers (Treatment - Control) 0.25 0.00 -0.25 -0.50 -0.75 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: no (short-run) effects on firm’s composition i.e., number of treated minus control workers.
[2.] Employment N f , t = � 12 j = − 13 γ j · d j f , t + µ f + µ t + ǫ f , t 2.0 1.5 1.0 0.5 Firm size 0.0 -0.5 -1.0 -1.5 -2.0 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: no (short-run) effects on employment i.e., firm size.
[3.] Full vs part-time workers G S FT j = − 13 γ j · d j = � 12 f , t + µ f + µ t + ǫ f , t f , t 0.010 0.007 Gap in the share of full-time workers 0.005 (Treatment - Control) 0.003 0.000 -0.003 -0.005 -0.007 -0.010 -12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 Months relative to treatment Note: no (short-run) effects on full-time job.
[4.]Collusion and bunching at notches ◮ Hypothesis: bunch not to lose transfer. ◮ Result: no visible bunching at notches Bunch ⇒ some spikes, but not � = pattern for workers w/ and w/o children Kids ◮ No-bunching is not due to poor enforcement. Sharp discontinuity in transfer’s amount above notches. ◮ Explanation: (i) difficult E-E coordination; (ii) not large incentives to bunch (change in ATR ∼ 2)
Bunching at AAFF notches Go back 30,000 MW Notch 1 Notch 2 Notch 3 30 $ 450 $ 725 $ 1225 $ 2025 25,000 25 20,000 Transfer/ Earnings (%) 20 Salaried workers 15,000 Freq. (left) 15 ATR 2 kids (right) 10,000 10 5,000 5 0 0 0 500 1,000 1,500 2,000 2,500 Gross Monthly Earnings (pesos)
By number of children Go back .03 No kids 1 kid 2+ kids .025 .02 Density .015 .01 .005 0 0 500 1,000 1,500 2,000 2,500 Gross Monthly Earnings (pesos)
Across months Go back Notch 1 .04 $ 725 Salaries 04-2005 April 2005 .03 Density .02 .01 0 0 500 1,000 1,500 2,000 2,500 Gross Monthly Earnings (pesos)
Across months Go back Notch 1 .04 $ 725 Salaries 07-2005 April 2005 July 2005 .03 Density .02 .01 0 0 500 1,000 1,500 2,000 2,500 Gross Monthly Earnings (pesos)
Recommend
More recommend
