Santiago Garriga Dario Tortaroloeconomics.dtortarolo.com.ar/Garriga-Tortarolo-slides-apr2020.pdf · 1. Employer-employee (SICOSS) (2003-2010) I E-E panel available since 1995 month

Wage effects of employer-mediated

transfers

Santiago Garriga Dario TortaroloParis School of Economics UC Berkeley

Labor Lunch, UC Berkeley

April 28, 2020

1 / 31

Motivation

I Governments often rely on firms as intermediaries in thetax-benefit system (e.g., employer-based health insurance; payroll/incometax withholding; family transfers, etc.)

I Sensitive information is revealed in the process and rentopportunities arise (wage effects)

I Yet, little evidence on the economic incidence/wage effectsof means-tested transfers

I We focus on employer-mediated transfers, which are morewidespread than publicly known

1 / 31

Motivation

Employer-mediated family allowances around the globe:

I Latin American countries

I Brazil (Salário Faḿılia)I Chile (Asignación Familiar)I Paraguay (Asignación Familiar)I Perú (Asignación Familiar)

I Developed countries

I USA (Advanced Earned Income Tax Credit) 1979-2010I UK (Working Family Tax Credit) 1999-2003I Italy (Bonus Renzi 80 Euro)I Switzerland (Familienzulagen)

2 / 31

This project

Does the way means-tested transfers are paid matter?

Do employers capture part of the transfer bylowering wages when being the remitter?

I Exploit a change in the payment system in Argentinagradually rolled out in 2003-10:

I Before: employers (intermediaries)I After: social security administration (direct deposit)

I Event study by switching date comparing employees w/ andwo/ children within firms

I Rich population-wide admin data from Arg (2003-2010)

3 / 31

Contribution to the literature

1. Literature on incidence:

I Classical PE literature: relative supply and demand elasticitiesI SSC: Saez et al. QJE’12 and Saez et al. AER’19I Saliency: taxes [Chetty et al. AER’09]; SSC [Bozio et al. ’19]I Remittance and compliance costs: taxes [Slemrod, NTJ’08;

Kopczuk et al., AEJ-EP’16]

I In-work subsidies: EITC [Rothstein AER’10; Leigh ’10; Hoynes-PatelJHR’18] and WFTC [Azmat, QE’18]

−→ We focus on transfers; change payment system holdingconstant ATR & MTR

2. Design of welfare programs and social protection:

−→ Unintended consequences of decentralizing key duties

4 / 31

Preview of findings

(1) The way transfers are disbursed matters (affects thefinal incidence):

I Monthly wages ↑ ∼ 2% when the govt’ becomes the remitterI Pass-through: employers were capturing ∼ 10/20% of the

transfer by paying lower wages

(2) Mechanism (preliminary): labor demandLikely, transfer understood as wages under the old system

I Driven by new hires rather than incumbents; small firms

I Exercises that rule out pay equity concerns (bargaining)

5 / 31

Outline

1. Setting: FA scheme, reform and roll out

2. Data and empirical strategy

3. Results and robustness checks

4. Mechanisms (labor demand vs supply)

Family allowance program (FA)

I Child benefit for wage earnersI Monthly payment that varies by:

(i) Number of kids < 18 years old;(ii) Monthly wage (3 brackets)

I Individually-based but only one spouse is entitled

I Funding: contributory system based on employer SSCI Specific component devoted to FA funding 7.5%

I Adjusted annually due to inflation (period 2003/10)I More on setting

6 / 31

Child benefits in Argentina

Notch 1 Notch 2 Notch 3

$ 40per child

$ 30per child

$ 20per child

0

5%

10%

15%

20%

25%

30%

35%

40%Tr

ansf

er /

Mon

thly

wag

e (%

)

0 500 1,000 1,500 2,000Gross Monthly Wage (pesos)

1 kid

2 kids

3 kids

Note: schedule in place from 1996 to 2004. Then updated ≈every year.

7 / 31

The reform (1): a change in the payment system

Old system: Sistema de Fondo Compensador (SFC)

I Employers paid child benefits together with the monthly wageI Must appear in pay-slip by law

I Deduct benefit amount from employer SSC

New system: Sistema Unico de Asignaciones Familiares (SUAF)

I Eliminated the intermediary role played by employers

I SSA deposits the benefit directly into workers’ bank account

Motivation

8 / 31

The reform (2)

Old system (SFC)

Employees Employers Government

SSC − Transfer(τ e)Wage + Transfer(τ e)

New system (SUAF)


SSCWage

Transfer(τ g )

9 / 31

The reform (2)

Old system (SFC)


SSC−Transfer(τ e)Wage+Transfer(τ e)

New system (SUAF)


SSCWage

Transfer(τ g )

9 / 31

Reform’s roll-out

I Gradual roll-out: btw Jun-2003 and Jun-2010 (8 years)

I Limited capacity to incorporate millions of beneficiaries at once

I Important: # beneficiaries and FA spending don’t ↓

I Incorporation: switching date set by the SSA rather than firms

SSA Memo (1)

Incorporationschedule/plan

(about 1-6 months)

docs presented and revised

firms contacted by SSA

SSA Memo (2)

FormalIncorporation

Timeline

(within 10 days)

form PS.2.61

employers notify workers

10 / 31

Micro roll-out (using E-E microdata)

Jun03 Jul10

Firms

Workers

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70Sh

are

unde

r old

sys

tem

Dec02 Mar04 Jun05 Sep06 Dec07 Mar09 Jun10Switching date

Note: gradual transition of firms and workers into the new system.

11 / 31

Roll-out and firm size

[

Administrative Data

1. Employer-employee (SICOSS) (2003-2010)

I E-E panel available since 1995 month by monthI Main vars: monthly pre-tax wage, monthly transferI Other: 4d sector, health insurer code, zip codes

2. Family relationships databaseI Links family members (spouse, children); also DOB

3. Other: (a) Monthly financial situation of employers(Apr’03-Nov’04); (b) Union’s CBA dataset

13 / 31

Empirical strategy: event study

Switch

Before (SFC)

Transfer paid by employers

Gap in mean wage btw T & C

G w̄f ,t = w̄Tf ,t − w̄Cf ,t

After (SUAF)

Transfer paid by govt

Gap in mean wage btw T & C


-1-2-3-4-5 0 1 2 3 4

Event window

14 / 31

Empirical strategy: event study

I Sample: unbalanced panel of firms for which we observe anevent (old system −→ new system) Details

→ paying FA for at least 6 consecutive months before event→ present at least −6/+ 6 months around the event→ with eligible (T) and non-eligible (C) workers in the window

T: employees w/ children ages [0-17]C: employees wo/ children ages [0-17]

→ collapse data at the firm-month-year level (f,t)

I Run a regular event study specification

G w̄f ,t =12∑

j=−13γj · d jf ,t + µf + µt + �f ,t (1)

15 / 31

Event frequency

Aug'08

Jun'09

1k

2k

3k

4k

5k

6k

7kN

umbe

r of f

irms

2003m7 2005m1 2006m7 2008m1 2009m7 2011m1Switching date

Note: massive incorporation in Aug’08 (Great Recession), Jun’09, Mar-Jul’10.Macro context Average size

16 / 31

Event frequency

Aug'08

Jun'09

Jun'10

2k

4k

6k

8k

10k

12k

14k

16kN

umbe

r of f

irms


Note: massive incorporation in Aug’08 (Great Recession), Jun’09, Mar-Jul’10.Macro context Average size

16 / 31

First stage (∆ transfer paid by employers)

0

35

70

105C

onst

ant p

esos

(bas

e =

Jan

2004

)

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12Months relative to treatment

Note: on average, treated workers receive ∼ 90 pesos more in transfer, paid byemployers, than the control group (simple mean difference).

17 / 31

Reduced form (∆ in mean wage)G w̄f ,t =

∑12j=−13 γj · d

jf ,t + µf + µt + �f ,t

-10

-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: mean wage of workers w/ kids increased by ∼ 9 pesos, relative to workers wo/kids, after firms switched to new system (pre Aug’08). Levels T vs C 18 / 31

Reduced form (∆ p25 and p75)Gwf ,t =

∑12j=−13 γj · d

jf ,t + µf + µt + �f ,t

p25

p75-10

-5

0

5

10

15

20

25

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: increase in wage is larger for those workers located at the bottom of thedistribution (p25); likely more treated due to the transfer scheme. 19 / 31

Pass-through

All post periods Last 6 periods Last period[0;11] [6;11] [11]

(1) (2) (3)Reduced form∆ monthly wage 7.71*** 8.74*** 9.23***

(in pesos) (1.25) (1.55) (1.87)

First stage∆ transfer (τ e) -90.98*** -92.17*** -91.44***

(in pesos) (0.35) (0.37) (0.37)

2sls∆wage

∆transfer(τ e) -0.08*** -0.09*** -0.10***

(0.01) (0.02) (0.02)

Number of firms 35,787 35,787 35,787Observations 3,061,870 2,847,148 2,670,757Avg wage at t-1 868 868 868

Note: Standard errors clustered at the firm level in parentheses.

Gwf ,t = β1Windowf ,t+β2 ·Windowf ,t ·Postf ,t+β3(1−Windowf ,t)·Postf ,t+µf +µt+�f ,t ,where Window is an indicator for the event window. Robustness Checks

20 / 31

Mechanisms (making sense of the findings)

Labor demand story?

I Employers exploit confusion of the old regime and capturepart 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?

I 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 incumbentsG w̄f ,t =

∑5j=−6 γj · d

jf ,t + µf + µt + �f ,t

All workers

Incumbents

-10

-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)

-5 -4 -3 -2 -1 0 1 2 3 4Months relative to treatment

Note: incumbents: workers present -7/+7 months around the event.22 / 31


Labor demand story?



Labor supply story?



Small vs non-small firmsDwf ,t =

∑24j=−13 γj · d

jf ,t + µf + µt + �f ,t

Small [

Incorporated vs Unincorporated firmsDwf ,t =

∑12j=−13 γj · d

jf ,t + µf + µt + �f ,t

Incorporated

Non-incorporated

-10

-5

0

5

10

15

20

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: first two digits of tax id determine if firm is incorporated or not.24 / 31

Incorporated firms: small vs non-smallDwf ,t =

∑12j=−13 γj · d

jf ,t + µf + µt + �f ,t

Small [

Heterogeneities

IncorporatedSmall 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

∆transfer(τ e) -0.10*** -0.06*** -0.01 -0.13*** -0.20*** -0.07***

(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 12,915Observations 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.

Gwf ,t = β1Windowf ,t+β2 ·Windowf ,t ·Postf ,t+β3(1−Windowf ,t)·Postf ,t+µf +µt+�f ,t ,where Window is an indicator for the event window.

26 / 31


Labor demand story?



Labor supply story?



Anecdotal evidence on recipient’s perception (1)

1. Quote from a book on social security:

“... the old system (SFC) blurred the image of the Stateas responsible for it. (...) The roles are confused. Peopleconsider that these benefits integrate their salary andthat employers are responsible for them. They even ignorethat it is the State that pays for them...”

Note: “Poĺıticas de Protección familiar, Régimen de Asignaciones Familiares yprincipales planes sociales en la República 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 typeA. 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


Labor demand story?



Labor supply story?



Horizontal equity

Exposure0.35 0.43 0.48 0.52 0.55 0.60 0.65 0.71

-0.20

-0.15

-0.10

-0.05

0.00

0.052s

ls c

oeffi

cien

t

0-30%Lowestshare

10-40% 20-50% 30-60% 40-70% 50-80% 60-90% 70-100%Highestshare

Treated workers

Note: each dot refers to a different regression of the type:Gwf ,t = β1Windowf ,t+β2 ·Windowf ,t ·Postf ,t+β3(1−Windowf ,t)·Postf ,t+µf +µt+�f ,t .

Exposure Density29 / 31

Misc

Long-run wage effects:

1. Mean wage

2. p25 and p75

3. Pure event (levels)

Other real outcomes:

4. Composition

5. Employment

6. Credit score (debt)

30 / 31

Conclusions

I ∆ in the remittance system (from employer to govt):

I Wages ↑ ∼ 2% after firms switch to the new systemPass-through: under old payment system employers capture∼ 10/20% of the transfer by paying lower wages

I Welfare improving reform from worker’s point of view

I The way transfers are disbursed matters and affects the finalincidence (not captured dollar for dollar by workers)

I These results raise questions about the use of employers asintermediaries to disburse benefits;less salient schemes may lead to capture by employers

31 / 31

Thanks!

Persistent inflation throughout the period

CPI

RIPTE

1,000

2,000

3,000

4,000

RIPT

E (c

urre

nt A

rgen

tinea

n pe

sos)

150

200

250

300

350

400CP

I (in

dex)

Jan03 Apr04 Jul05 Oct06 Jan08 Apr09 Jul10Date

Note: CPI denotes consumer price index while RIPTE to the average salary of

registered workers (in current pesos). Back FA

Evolution of FA brackets and minimum wage

Top bracket

2nd

3rd

MinimumWage

0

1,000

2,000

3,000

4,000

5,000C

urre

nt A

rgen

tinia

n pe

sos

Dec02 Mar04 Jun05 Sep06 Dec07 Mar09 Jun10Date

Note: brackets are updated roughly once per year, but no in fix intervals. Back FA

Distribution of monthly wage Back FA


020

4060

80C

hild

ben

efit

(in p

esos

)

010

,000

20,0

0030

,000

40,0

00W

age

Earn

ers

0 500 1,000 1,500 2,000 2,500Gross Monthly Wage (pesos)

Density (left)

Transfer (right)

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

-.2-.1

5-.1

-.05

0Sh

are

rece

ivin

g tra

nsfe

rs (p

.p.)

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12Months relative to event (turn 18)

Point estimate

95% C.I.

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

-10

010

2030

40C

onst

ant P

esos

200

4

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12Months relative to event (turn 18)

Child transfer

Wage earnings

Note: event study when a kid turns 18 and workers lose eligibility.

Collective bargaining agreements (CBA)

Jun03 Jul10

0

100

200

300Nu

mbe

r of a

gree

men

ts

Jan03 Apr04 Jul05 Oct06 Jan08 Apr09 Jul10Date

Note: collective bargaining agreements occur every month. Two-thirds of them are

firm level agreements. Back FA

Average # workers by switching date Back

Aug'080

2040

6080

100

120

140

Avg.

N o

f em

ploy

ees


Explanation of no-results for switchers after July 2008

I Drop in economic activity (figure EMAE/GDP growth)

I Stabilization of employment growth (figure wage earners)I Interesting if the effect comes from new hires.

I Lower ATR? (figure ratio transfer/min W)I This doesn’t seem to be the main raison.

Back FA Back Freq

Large drop in economic activity (EMAE)

90

100

110

120

130

140M

onth

ly e

cono

mic

act

ivity

est

imat

or(b

ase

=200

4)

Jan04 Jan05 Jan06 Jan07 Jan08 Jan09 Jan10 Jan11Date

Note: large drop in economic activity from August 2008 onwards. Back FA

Back Freq Back RB

Stabilization of private employment growth

3.00

4.00

5.00

6.00

7.00Pr

ivat

e em

ploy

ees

(in m

illion

s)

2003Q1 2004Q3 2006Q1 2007Q3 2009Q1 2010Q3Date

Note: stabilization of employment in the third quarter of 2008. Back FA Back Freq

Back RB

Evolution of the ATR (∼ ratio transfer/min wage)

0.00

0.05

0.10

0.15

0.20R

atio

tran

sfer

/min

imum

wag

e

Jul03 Jul04 Jul05 Jul06 Jul07 Jul08 Jul09 Jul10Date

1st 2nd 3rd bracket

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

I Make sure beneficiaries receive the transfer

I Transparency and efficiency reasons

I 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)

0.00

0.20

0.40

0.60

0.80Sh

are

paid

thro

ugh

SFC

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011Year

Note: gradual decline in the share of FA paid through the old system (SFC). Go back

Total spent and aggregates comparison (SFC+SUAF)

1,000

1,500

2,000

2,500

3,000

3,500

4,000FA

spe

ndin

g (m

illion

pes

os fr

om 2

003)

2003 2004 2005 2006 2007 2008 2009 2010Year

Macro aggregatesMicrodata

Source of data:

Note: (a) increase in FA payments as time passes, (b) replicate macro aggregates

using micro-data. Go back

Beneficiaries (number of children) Go back

0

0.5m

1m

1.5m

2m

2.5m

3m

3.5mN

umbe

r of c

hild

ren

(in m

illion

s)

2002 2003 2004 2005 2006 2007 2008 2009 2010 2011Year

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 (τ̄) Employee (τ̄)

Remittance responsibility Employer (τ e) Government (τg )

Information reporting Form 931 Form 931

Salience to employers High =

Employees perception (q) Low ↑

Source of funding Contributory ContributoryEmployer 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

I We digitized 50+ schedule plans recovering approximately60K firms with its corresponding “internal deadline”.

I Note: not all firms appeared in the schedules (or at least, inthe publicly available ones).

I Less than 0.1% of firms appeared in more than one schedule.

I Compare deadline with the effective incorporation date.I ∼ 90% of firms were incorporated before deadline.

Scheduled vs effective incorporation date (micro-data)

Deadline to beincorporated

(ANSES'schedule)

0.0

0.2

0.4

0.6

0.8

1.0C

DF

-20 -15 -10 -5 0 5 10 15 20Months

before/after deadline

Note: ∼ 90% of firms were incorporated before government’s deadline. Go back

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)

I Difficult to track the universe of approval memos

I However, it was possible to do a public query so as to checkthe formal incorporation date

I Random sample of 300 firms

I Compare formal incorporation vs effective incorporation dateI ∼ 80% firms incorporated in the same date as formal approval

I No incentives to delay:I Cannot compensate the money (+ inflation)

I Workers’ notification: Within ten days after the switch, firmsshould inform their workers about the new paymentmechanism and the overall scheme of the FA system

Roll out query Go back

Formal inclusion and observed incorporation (micro-data)

0.00

0.20

0.40

0.60

0.80

1.00C

um s

hare

-20 -16 -12 -8 -4 0 4 8 12 16 20Months

Before/after formal incorporation

Note: ∼ 80% of firms were incorporated the same date as the formal approval.Go back

Notification to employees (affidavit) Go backNotificación del Régimen de Asignaciones Familiares Sistema Único de Asignaciones Familiares

Form.PS.2.61

Frente 1

Versión 1.3

Ministerio de Trabajo,

Empleo y Seguridad Social

Apellido y Nombre Completo

Cuil

Domicilio - Calle - Nuemero

LocalidadCódigo PostalPiso

Teléfono Dirección de Correo Electrónico

Depto. Provincia

Fecha de Nacimiento

Sexo

Nacionalidad

Estado CivilTipo y Nº Doc /CUIL

RUBRO I – DATOS DEL TRABAJADOR (a completar por todos los trabajadores con o sin cargas de familia)

Este Formulario reviste carácter de Declaración Jurada y se debe completar en letra de imprenta, sin tachaduras ni enmiendas

Razón Social

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 plazosde 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 seproduzca 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.

CUIT

Domicilio - Calle - Nuemero

LocalidadCódigo PostalPiso

Teléfono Dirección de Correo Electrónico

Depto. Provincia

RUBRO I I – DATOS DEL EMPLEADOR

Localidad, .................... de .............................……. de ..........……

Firma/Aclaración de Firma del Trabajador

Firma/Aclaración de Firma y Sello del Empleador

Sample construction

I 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).

I 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 analysiswith 4, 5, 7 and 8 months.

I Checks

1. Balanced panel X2. Sensitivity FA payments X

Go back

First difference model BackOutcome variable: monthly wage used as base for employers’ SSC.

We get the mean wage for each firm-group-month and take thedifference across groups


Then, for each firm we have a time series of first differences−→ run regular event study specification (f: firms, t: month-year)

G w̄f ,t = α +12∑

j=−13γj · d jf ,t + �f ,t (1)

...

wmeani ,f ,t = βTi ,f ,t +12∑

j=−13γj · d jf ,t · Ti ,f ,t + µf ,t + �i ,f ,t (2)

Empirical approach: Reduced form BackIn levels

wmeani ,f ,t = βTi ,f ,t +5∑

j=−6γj · d jf ,t · Ti ,f ,t + µf ,t + �i ,f ,t (3)

Pooled γs: Alt (1)

wmeani ,f ,t = β1Ti ,f ,t+β1Ti ,f ,t ·Windowf ,t+β2Ti ,f ,t ·Windowf ,t ·Postf ,t+(4)

β3Ti ,f ,t · (1−Windowf ,t) · Postf ,t + µf ,t + �i ,f ,tPooled γs: Alt (2)

wmeani ,f ,t = βTi ,f ,t + β−6 d−6f ,t · Ti ,f ,t︸︷︷︸

d n6

+β5 d5f ,t · Ti ,f ,t︸︷︷︸

d 5

(5)

βafter d[0;4]f ,t · Ti ,f ,t︸︷︷︸

d after

+µf ,t + �i ,f ,t

Note: in (3) j = −1 omitted category; in (5) months [-5;-1] are omitted. Window = 1if months [-5;4].

Empirical approach: Reduced form Back

First difference

Gwf ,t = α +5∑

j=−6γj · d jf ,t + �f ,t (6)

γ’s should be numerically the same as those estimated in eq.(3)

Pooled γs:

Average Gw before = Gwbefore = (γ−5 + γ−4 + γ−3 + γ−2 + 0)/5

Average Gw after = Gwafter = (γ0 + γ1 + γ2 + γ3 + γ4)/5

Average effect = Gwaverage = Gwafter − Gwbefore

Note: in (6) j = −1 omitted category.

Empirical approach: Reduced form BackFirst difference

Gwf ,t = α +5∑

j=−6γj · d jf ,t + �f ,t (7)

γ’s should be numerically the same as those estimated in eq.(3)

Pooled γs: Alt (1)

Gwf ,t = α + β1Windowf ,t + β2Windowf ,t · Postf ,t (8)

+β3(1−Windowf ,t) · Postf ,t + �f ,tPooled γs: Alt (2)

Gwf ,t = α + β−6 d−6f ,t︸︷︷︸

d n6

+β5 ·d5f ,t︸︷︷︸d 5

+βafter d[0;4]f ,t︸︷︷︸

d after

+�f ,t (9)

Note: in (6) j = −1 omitted category; in (8) months [-5;-1] are omitted. Window = 1if months [-5;4].

Empirical approach: Reduced form Back

Expected values: Alt (1)

1. E (Gw/Win = 0, post = 0) = α

2. E (Gw/Win = 1, post = 0) = α + β1

3. E (Gw/Win = 1, post = 1) = α + β1 + β2

4. E (Gw/Win = 0, post = 1) = α + β3

Expected values: Alt (2)

1. E (Gw/Win = 0, post = 0) = α + β−6

2. E (Gw/Win = 1, post = 0) = α

3. E (Gw/Win = 1, post = 1) = α + βafter

4. E (Gw/Win = 0, post = 1) = α + β5

First difference model Back

Gw

Switch

Window = 0post = 0

Window = 1post = 0

Window = 1post = 1

Window = 0post = 1

-5 -1 0 4

Event window

Distance

Empirical approach: 2sls Back

First stage

GTransferf ,t = α + δ1Windowf ,t + δ2Windowf ,t · Postf ,t

+δ3(1−Windowf ,t) · Postf ,t + �f ,tReduced form

Gwf ,t = α + β1Windowf ,t + β2Windowf ,t · Postf ,t

+β3(1−Windowf ,t) · Postf ,t + �f ,t2sls (Wald estimator)

then we should have −→ Θ = β2δ2

Note: if I drop the binned points I get numerically the same number.

Empirical approach Back

First difference with firm and time FE

Gwf ,t =5∑

j=−6γj · d jf ,t + µf + µt + �f ,t (10)

Pooled γs: Alt (1) ... closer but not exact

Gwf ,t = β1Windowf ,t + β2 ·Windowf ,t · Postf ,t (11)

+β3(1−Windowf ,t) · Postf ,t + µf + µt + �f ,tPooled γs: Alt (2) ... closer but not exact

Gwf ,t = β−6 d−6f ,t︸︷︷︸

d n6

+β5 ·d5f ,t︸︷︷︸d 5

+βafter d[0;4]f ,t︸︷︷︸

d after

+µf + µt + �f ,t (12)

Note: in (10) j = −1 is the omitted category; in (12) months [-5;-1] are the omittedones.Window = 1 if months[-5;4].

Empirical approach

First difference 2sls with firm and time FE

Reduced form (z = instrument)

Gwf ,t = γ−6 d−6f ,t︸︷︷︸

d n6

+γ5 ·d5f ,t︸︷︷︸d 5

+γafter d[0;4]f ,t︸︷︷︸

d after

+ µf + µt + �f ,t

First stage:

G transferf ,t = γ−6 d−6f ,t︸︷︷︸

d n6

+γ5 ·d5f ,t︸︷︷︸d 5

+γafter d[0;4]f ,t︸︷︷︸

d after

+ µf + µt + �f ,t

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

Gwf ,t = α + β1Windowf ,t + β2Windowf ,t · Postf ,t

+β3(1−Windowf ,t) · Postf ,t + +β4Highf + β5Windowf ,t · Highf+β6Windowf ,t ·Postf ,t ·Highf +β7(1−Windowf ,t)·Postf ,t ·Highf +�f ,t

If we were to include firm and time FE (µf and µt), then we need to add also theinteraction between µt and Highf . β4 disappears as it is absorbed by µf .

Specification 2sls with High/low:

Robustness checks Back

I Sample sensitivityI Balanced panel (96 months)I Sensitivity FA payments

I Results not driven by modeling choices:I 6= specs: Simple mean, firm & time FE, firm linear trends

I Event window size: 10, 12, 14,... 24 months, etc.

I Event by event

I Sensitivity to the treatment group definitionI Fully treated vs never/partially treated (life events).I Treatment status based on year of birth.

Balanced panel

Dwf ,t =∑12

j=−13 γj · djf ,t + µf + µt + �f ,t

-10

-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: results remain unchanged when considering the balanced panel. Back

Sensitivity FA payments (2sls)

-0.15

-0.12

-0.09

-0.06

-0.03

0.00

0.03

0.062s

ls c

oeffi

cien

t

4 5 6 7 8Sensitivity FA payments

Note: each dot refers to a different regression of the type:Gwf ,t = β1Windowf ,t+β2 ·Windowf ,t ·Postf ,t+β3(1−Windowf ,t)·Postf ,t+µf +µt+�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

∆transfer(τ e) -0.08*** -0.08*** -0.06***

(0.01) (0.01) (0.01)

Simple mean difference XFirm and time FE X XFirm linear trend XObservations 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)

w̄f ,t =∑12

j=−13 γj · djf ,t + µf + µt + �f ,t

Treat

Control

-10

-5

0

5

10

15

20

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: increase in wage is driven by treated workers. Go back

Window’s dynamic (mean)

Two years

One year around event

-10

-5

0

5

10

15Co

nsta

nt p

esos

(bas

e =

Jan

2004

)


Note: results remain unchanged for a time window of 6 months before/after. Back

Effect’s dynamics – 2sls

-0.12

-0.11

-0.10

-0.09

-0.08

-0.072s

ls co

effic

ient

All postperiods[0:11]

Last-halfpost periods

[6:11]

Only lastperiod

[11]

Post periods included

Note: each dot refers to a different regression of the type:Gwf ,t = β1Windowf ,t+β2 ·Windowf ,t ·Postf ,t+β3(1−Windowf ,t)·Postf ,t+µf +µt+�f ,t ,where we vary the post periods included in the event window (W ).

Event by event – 2sls

-0.150

-0.100

-0.050

0.000

0.0502s

ls c

oeffi

cien

t

Jul03 May04 Mar05 Jan06 Nov06 Sep07 Jul08Excluded month

Note: each dot refers to a different regression of the type:Gwf ,t = β1Windowf ,t+β2 ·Windowf ,t ·Postf ,t+β3(1−Windowf ,t)·Postf ,t+µf +µt+�f ,t ,where we exclude switchers in a given month. Go back

Rolling window of events (whole roll-out)

0

10

20

30

40

# of

firm

s (in

K)

-0.16

-0.12

-0.08

-0.04

0.00

0.042s

ls

Jul03-Jun05

May04-Apr06

Mar05-Feb07

Jan06-Dec07

Nov06-Oct08

Sep07-Aug09

Jul08-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]. Thismeans that these workers are fully treated during the period2003-2010.

January 2003

Born in (age):1992 (11)2002 (1)

December 2010

Born in (age):1992 (18)2002 (8)

The rest of the workers belong to the control group (nevertreated and/or partially treated).

Alternative treatment group definition (mean)

-10

-5

0

5

10

15C

onst

ant p

esos

(bas

e =

Jan

2004

)


Note: results hold when using an alternative definition of the treatment group.Go back

Alternative T group definition (mean) using year of birth

-10

-5

0

5

10

15C

onst

ant p

esos

(bas

e =

Jan

2004

)


Note: results hold when using an alternative definition of the treatment group.Go back

Whole period

Before crisis

After crisis

-10

-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: before crisis refers to firms switching into the new system before August 2008;

after between August 2008 and July 2010. Go back

Other potential behavioral responses

1. Worker composition

2. Employment

3. Full vs part-time workers

4. Bunching at notches

[1.] Worker composition

GNf ,t =∑12

j=−13 γj · djf ,t + µf + µt + �f ,t

-0.75

-0.50

-0.25

0.00

0.25

0.50

0.75

Gap

in n

umbe

r of w

orke

rs(T

reat

men

t - C

ontro

l)


Note: no (short-run) effects on firm’s composition i.e., number of treated minuscontrol workers.

[2.] Employment

Nf ,t =∑12

j=−13 γj · djf ,t + µf + µt + �f ,t

-2.0

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

2.0

Firm

siz

e


Note: no (short-run) effects on employment i.e., firm size.

[3.] Full vs part-time workers

GSFTf ,t =

∑12j=−13 γj · d

jf ,t + µf + µt + �f ,t

-0.010

-0.007

-0.005

-0.003

0.000

0.003

0.005

0.007

0.010

Gap

in th

e sh

are

of fu

ll-tim

e w

orke

rs(T

reat

men

t - C

ontro

l)


Note: no (short-run) effects on full-time job.

[4.]Collusion and bunching at notches

I Hypothesis: bunch not to lose transfer.

I Result: no visible bunching at notches Bunch

⇒ some spikes, but not 6= pattern for workers w/ and w/ochildren Kids

I No-bunching is not due to poor enforcement. Sharpdiscontinuity in transfer’s amount above notches.

I Explanation: (i) difficult E-E coordination; (ii) not largeincentives to bunch (change in ATR ∼ 2)

Bunching at AAFF notchesGo back

Notch 1$ 725

Notch 2$ 1225

Notch 3$ 2025

MW$ 450

0

5

10

15

20

25

30

Tran

sfer

/ Ear

ning

s (%

)

05,

000

10,0

0015

,000

20,0

0025

,000

30,0

00Sa

larie

d w

orke

rs

0 500 1,000 1,500 2,000 2,500Gross Monthly Earnings (pesos)

Freq. (left)

ATR 2 kids (right)

By number of childrenGo back

0.0

05.0

1.0

15.0

2.0

25.0

3D

ensi

ty


No kids

1 kid

2+ kids

Across monthsGo back

Notch 1$ 725

0.0

1.0

2.0

3.0

4D

ensi

ty


Salaries 04-2005

April 2005


Notch 1$ 725

0.0

1.0

2.0

3.0

4D

ensi

ty


Salaries 07-2005

April 2005

July 2005


Notch 1$ 725

0.0

1.0

2.0

3.0

4D

ensi

ty


Salaries 08-2005

April 2005

August 2005


Notch 1$ 725

Notch 1$ 1200

Update Sept0

.01

.02

.03

.04

Den

sity


Salaries 08-2005

April 2005

November 2005

Median transfer (Aug-2005)Go back


050

100

150

Tran

sfer

am

ount

(in

Peso

s)

010

,000

20,0

0030

,000

40,0

0050

,000

Sala

ried

Wor

kers

0 1,000 2,000 3,000 4,000 5,000Gross Monthly Earnings (pesos)

Density (left)

Transfer (right)

Mean transfer (Aug-2005)Go back


050

100

150

Tran

sfer

am

ount

(in

Peso

s)

010

,000

20,0

0030

,000

40,0

0050

,000

Sala

ried

Wor

kers

0 1,000 2,000 3,000 4,000 5,000Gross Monthly Earnings (pesos)

Density (left)

Transfer (right)

[1.] Turnover rate

Dwf ,t =∑5

j=−6 γj · djf ,t + µf + µt + �f ,t

High

Low-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


High: above median turnover rate. Go back

[2.] Full-time contract

Dwf ,t =∑5

j=−6 γj · djf ,t + µf + µt + �f ,t

Low

High-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


High: above median share of full-time workers. Go back

[3.] Transfer discrepancy

Dwf ,t =∑5

j=−6 γj · djf ,t + µf + µt + �f ,t

Low

High-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


High: above median share of transfer discrepancies. Go back

Density of Firm Exposure

0.00

0.05

0.10

0.15

0.20

0.25Fr

actio

n

0.00 0.20 0.40 0.60 0.80 1.00Share treated workers

Note: exposure defined as the within-firm share of workers with children. Go back

[1.] Mean wage

Gwf ,t =∑24

j=−13 γj · djf ,t + µf + µt + �f ,t

-10

-5

0

5

10

15

Con

stan

t pes

os (b

ase

= Ja

n 20

04)

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24Months relative to treatment

Note: the effect stabilizes 12 months after the switch. Go back

[2.] p25 and p75

Dwf ,t =∑24

j=−13 γj · djf ,t + µf + µt + �f ,t

p25

p75-10

-5

0

5

10

15

20

25

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: the effect stabilizes 12 months after the switch. Go back

[3.] Pure event (mean)

Wf ,t =∑24

j=−13 γj · djf ,t + µf + µt + �f ,t

Treat

Control

-10

-5

0

5

10

15

20

Con

stan

t pes

os (b

ase

= Ja

n 20

04)


Note: effect driven by the treatment group (workers with children). Go back

[4.a] CompositionGNf ,t =

∑14j=−13 γj · d

jf ,t + µf + µt + �f ,t

-1.50

-1.00

-0.50

0.00

0.50

1.00

Gap

in n

umbe

r of w

orke

rs(T

reat

men

t - C

ontro

l)

-12 -10 -8 -6 -4 -2 0 2 4 6 8 10 12 14Months relative to treatment

Note: firms’ composition seem to change when focusing on a longer time window.Go back

[4.b] Employment

Nf ,t =∑14

j=−13 γj · djf ,t + µf + µt + �f ,t

-3

-2

-1

0

1

2

3

Firm

siz

e


Note: firms size increases in the long-run (driven by workers wo/ kids). Go back

[4.c] Employment by treatment status

Nf ,t =∑14

j=−13 γj · djf ,t + µf + µt + �f ,t

Treat

Control

-1.5

-1.0

-0.5

0.0

0.5

1.0

1.5

Num

ber o

f wor

kers


Note: firms size increases in the long-run (driven by workers wo/ kids). Go back

Delinquency rates: debt past due 90+ days

-.04

-.02

0.0

2.0

4D

elin

quen

t deb

t (p.

p.)

-6 -5 -4 -3 -2 -1 0 1 2 3 4Months relative to event (SFC to SUAF)

Point estimate (10,481 firms)

95% C.I.

Note: firms switching btw Oct’03 and Jul’04 and in 2005 (N=10,481). Go back

Theoretical framework

Simple model to rationalize our findings (based on Gruber 1997):

Lsi = Lsi (w̃i ) = L

si (wi (1 + (1− q)τ ei )) (13)

Ldi = Ldi (w) (14)

where w̃ represents the perceived wage as fx of wage (w),perception parameter (q) and transfer by employers (τ e).

with τ e = τ̄ − τg

I perfect perception/knowledge (q=1), the perceived wage isequal to the true wage w̃1 = w

I no knowledge (q=0) −→, the perceived wage includes thetransfer w̃0 = w(1 + τ

e).

Theoretical framework

Totally differentiating supply and demand equations, andrearranging terms we get:

dln(wi )

dln(1 + tei )

∣∣∣∣τ̄=τ e+τg , q̄=q

=ηsi (1− q) · [

(1+tei )(1+(1−q)τ ei )

]

ηdi − ηsi(15)

Extreme cases

I q=1 −→ perfect knowledge, dln(wi )dln(1+tei ) = 0,standard incidence result.

I q=0 −→ no knowledge, dln(wi )dln(1+tei ) =ηsi

ηdi −ηsi

< 0

Change in perception (q)

I ∆ in the remittance responsibility −→ ∆ the informationcontent to employees.

−→ ∆ scheme’s perception (q), therefore on final incidence.

dln(wi )

dln(1 + tei )

∣∣∣∣τ̄=τ e+τg

=(1 + η

(1−q)i ) · ηsi (1− q) · [

(1+τ ei )(1+(1−q)τ ei )

]

ηdi − ηsi(16)

with η(1−q)i =

∂(1−q)∂τ ei

· τei

(1−q) −→ misperception elasticity i.e., howmuch (1− q) changes as the money disbursed by employersincrease (>0 elasticity i.e., reinforces the main effect).

Go back

Graphical analysis

l

wLS1 q 6=1

LS0 = LS1 q=1

LD

e

w0(1 + (1−q)te)

w0

w1

Note: Go back

Collective bargaining agreement (CBA)

Note: this screenshot presents the first page of a collective agreement, ConvenioColectivo de Trabajo, in Argentina. This is a standard type of agreement where thedifferent articles describe what has been discussed/negotiated.

Digitalization of CBA

Note: summary of the information extracted from a given collective agreement(CCT − 1523− 2016− E): it is at firm level (Nivel : Empresa), was celebrated inSeptember 29th 2015 (Celebración: 29-09-2015) and involved workers in the oil sector(Actividad : Petroleros). The main contents discussed are also enumerated(Contenidos discutidos: Adicional tareas de turno; Antiguedad; AporteSolidario, etc). Finally, firm’s name is available within the extracted information.

(Empleador/s: Yel Informatica S.A.). Go back

First stage (change in τ e)

τe τg

0

30

60

90C

onst

ant p

esos

(bas

e =

Jan

2004

)


Note: the red series represents the old system where employers paid the transfer (τ e);the blue one simulates the new system where the government directly disburse the

money to the beneficiaries (τg ). Go back

Documents

Santiago Garriga Dario Tortaroloeconomics.dtortarolo.com.ar/Garriga-Tortarolo-slides-apr2020.pdf · 1. Employer-employee (SICOSS) (2003-2010) I E-E panel available since 1995 month