saving more than r$ 85 billion of the brazilian social

BRAZILIAN JOURNAL OF PUBLIC ADMINISTRATION | Rio de Janeiro 54(6):1729-1746, Nov. - Dec. 2020

1729

ISSN: 1982-3134

Forum: Practical Perspectives

Saving more than R$ 85 billion of the Brazilian Social Security System: the case of the PRBI

Vinícius Botelho ¹Raquel Maria Soares Freitas ²Alberto Beltrame ³

¹ Fundação Getulio Vargas / Instituto Brasileiro de Economia, Brasília / DF – Brazil² Ministério da Cidadania, Brasília / DF – Brazil ³ Desenvolvimento Social, Brasília / DF – Brazil

Since 2016, the number of recipients of incapacity allowance in Brazil has been continuously falling. This article presents the program of incapacity benefits assessment (PRBI) to help understand the dynamics around incapacity allowance and similar benefits. The study shows that the PRBI can save more than R$ 85 billion of the budget allocated to social security in the country.Keywords: social security; incapacity benefits; program assessment.

Economizando mais de R$ 85 bilhões ao Regime Geral de Previdência Social do Brasil: o caso do PRBIO número de benefícios de auxílio-doença vem caindo drasticamente desde 2016. Este artigo mostra que o Programa de Revisão dos Benefícios por Incapacidade (PRBI) é fundamental para entender essa dinâmica, e estima que o Programa seja responsável por uma economia de mais de R$ 85 bilhões ao Regime Geral de Previdência Social.Palavras-chave: previdência social; benefícios por incapacidade; auxílio-doença.

Economía de más de R$ 85 mil millones en el Régimen General de Previsión Social de Brasil: el caso del PRBI

El número de beneficiarios de subsidios por incapacidad laboral ha disminuido drásticamente desde 2016. Este artículo muestra que el Programa para la Evaluación de Subsidios por Incapacidad Laboral (PRBI) es clave para entender esta dinámica y es responsable de una economía de más de R$ 85 mil millones para el Régimen General de Previsión Social de Brasil.Palabras clave: seguridad social; subsidios por incapacidad laboral; evaluación de programas.

DOI: http://dx.doi.org/10.1590/0034-761220190093xArticle submitted on March 08, 2019 and accepted on September 02, 2020.[Original Version]


RAP | Saving more than R$ 85 billion of the Brazilian Social Security System: the case of the PRBI

1730

1. INTRODUCTION

During recent years, Brazilian Social Security has expanded at an accelerated pace (Graph 1), which led to an intense public debate about a reform of the system, carried out and approved by Congress in 2019. Several measures to improve the management of social security benefits emerged from this debate. One example is the continuous assessment of incapacity benefits, an activity that had been neglected despite legal provision.

GRAPH 1 SOCIAL SECURITY BENEFITS (IN MILLIONS), 1999-2019 (DECEMBER)

de Subsidios por Incapacidad Laboral (PRBI) es clave para entender esta dinámica

y es responsable de una economía de más de R$ 85 mil millones para el Régimen

General de Previsión Social de Brasil.

Palabras clave: seguridad social; subsidios por incapacidad laboral; evaluación de

programas.

1. INTRODUCTION

During recent years, Brazilian Social Security has expanded at an accelerated pace (Graph

1), which led to an intense public debate about a reform of the system, carried out and

approved by Congress in 2019. Several measures to improve the management of social

security benefits emerged from this debate. One example is the continuous assessment of

incapacity benefits, an activity that had been neglected despite legal provision.

Graph 1

Social security benefits (in millions), 1999-2019 (December)

Source: Elaborated by the authors based on data from Boletim Estatístico da Previdência Social – BEPS

(Social Security Statistical Bulletin).

There are two kinds of incapacity benefit: incapacity allowance (IA), a temporary

benefit targeted to formal workers who are ill and cannot perform their labor activities,

regardless of whether their occupation caused the illness, and incapacity retirement,

Source: Elaborated by the authors based on data from Boletim Estatístico da Previdência Social – BEPS (Social Security Statistical Bulletin).

There are two kinds of incapacity benefit: incapacity allowance (IA), a temporary benefit targeted to formal workers who are ill and cannot perform their labor activities, regardless of whether their occupation caused the illness, and incapacity retirement, which is a permanent benefit. Graph 2 shows that the benefits present different patterns, as incapacity retirement is consistently growing over time.



1731

GRAPH 2 IA AND INCAPACITY RETIREMENT (IN MILLIONS), 1999-2019 (DECEMBER)

O sistema de previdência social brasileiro apresenta dois tipos principais de

benefícios por incapacidade: o auxílio-doença e a aposentadoria por invalidez. A

evolução desses dois benefícios ao longo do tempo é bastante distinta (Gráfico 2), visto

que a aposentadoria por invalidez cresce continuamente.

Gráfico 2

Auxílio-doença e aposentadoria por invalidez (número de benefícios, em milhões), 1999-2019

(dezembro)

Fonte: Elaborado pelos autores com base em dados do Boletim Estatístico da Previdência Social (BEPS).

Uma análise de picos e vales (Gráfico 3) sobre o número de beneficiários do

auxílio-doença revela dois picos (outubro de 2005 e fevereiro de 2010) e um vale (agosto

de 2016). Esse resultado é um tanto surpreendente, pois esse tipo de oscilação na

incidência de incapacidade na população não é algo esperado. Para Marinho, Resende, e

Lucas (2017), o auxílio-doença é um benefício de risco, ou seja, não depende diretamente

das escolhas dos beneficiários. A oscilação, portanto, coloca em dúvida a credibilidade

do sistema, ensejando discussões sobre a necessidade de reformá-lo, como observado nos

artigos de Liebman (2015) e Banks, Blundell, e Emmerson (2015).

4,0

3,5

3,0

2,5

2,0

1,5

1,0

0,5

-

Num

ber o

f ben

efits

(in

milli

on)

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019Years

Incapacity allowance Disability retirement benefits

Source: Elaborated by the authors, based on data from Boletim Estatístico da Previdência Social – BEPS (Social Security Statistical Bulletin).

A visual analysis of peaks and valleys on the number of IA beneficiaries (Graph 3) shows two peaks (October 2005 and February 2010) and one valley (August 2016). These results are somewhat surprising because such volatility in the incidence of incapacity among the population is unexpected: as Marinho, Resende, and Lucas (2017) point out, IAs are benefits that do not depend directly on beneficiaries’ choices. In many countries, such as in the US (Liebman, 2015) and the UK (Banks, Blundell, & Emmerson, 2015), this volatility led to debates about reforming the benefits system.

GRAPH 3 IA BENEFITS (2001-2019)

Gráfico 3

Número de benefícios de auxílio-doença pagos entre 2001 e 2019


A queda no número de benefícios desde 2016 foi tão expressiva que a folha de

pagamento do auxílio-doença tem caído todos os anos: o número de benefícios pagos ao

final de 2019 foi o menor desde 2003, custando ao sistema de previdência o montante de

R$ 20,1 bilhões anuais.

O presente artigo tem por objetivo discutir o quanto o Programa de Revisão dos

Benefícios por Incapacidade (PRBI) explica esse movimento. Entendê-lo é importante,

especialmente à luz de autores como Coile, Milligan, e Wise (2014), Autor, Duggan,

Greenberg, e Lyle (2015), e Mullen e Staubli (2016), que oferecem evidências de que

mudanças na generosidade dos benefícios por incapacidade têm impactos

socioeconômicos, afetando o mercado de trabalho e, particularmente, a participação da

população na força de trabalho.

Este artigo traz três contribuições principais. Primeiro, fazemos uma extensão do

modelo de Liebman (2015), para avaliar o impacto de mudanças não lineares nas regras

de duração dos benefícios sobre o número de beneficiários do auxílio-doença. Segundo,

demonstramos como o modelo pode ser usado para calcular o impacto econômico do

1,900,000

1,700,000

1,500,000

1,300,000

1,100,000

900,000

700,000

500,000

Num

ber o

f IA

bene

fits

Dec-

00

Dec-

01

Dec-

02

Dec-

03

Dec-

04

Dec-

05

Dec-

06

Dec-

07

Dec-

08

Dec-

09

Dec-

10

Dec-

11

Dec-

12

Dec-

13

Dec-

14

Dec-

15

Dec-

16

Dec-

17

Dec-

18

Dec-

19

Periodo 1 Periodo 2 Periodo 3 Periodo 4




1732

The fall in the fourth period is so significant that IA payroll has fallen every year since 2017. Moreover, the number of benefits at the end of 2019 was the lowest since 2003. In 2019, the Brazilian social security system spent R$ 20.1 billion on IA.

This article aims to discuss how much of the fall in the number of IA benefits since August 2016 can be explained by the program of incapacity benefits assessment (PRBI). Launched in mid-2016, it changed important aspects of IA design. International evidence (Coile, Milligan, & Wise, 2014; Autor, Duggan, Greenberg, & Lyle, 2015; Mullen & Staubli, 2016) shows that changes in the generosity of incapacity benefits affect the labor market – particularly the participation of the population in the labor force – and have socioeconomic impacts.

This study contributes to the literature by extending the model employed by Liebman (2015) to evaluate the impact of nonlinear changes in benefit duration rules on the number of IA beneficiaries. Another contribution lies in showing how to use the model to calculate the economic impact of reevaluating benefits. Finally, this research applies the extended model to provide the first (to our knowledge) evaluation of the PRBI.

The article is divided into five sections, including this introduction. The second section presents and discusses IA and the PRBI. The third part extends and calibrates Liebman’s (2015) model. The fourth section estimates the PRBI impact on the social security budget and the number of IA beneficiaries, followed by the conclusion.

2. IA AND PRBI

When social security contributors believe they are unable to work, they can apply for IA. The National Social Security Institute (INSS) evaluates their medical condition and may grant the benefit. The IA is always temporary and has a termination date. Since it is hard to anticipate the incapacitation period, if the beneficiary believes they cannot return to work by the termination date, they can schedule another medical examination and ask for a benefit extension. Termination dates cannot exceed two years, so benefits are reevaluated timely. Very often, beneficiaries try to revert application rejections in court. In many cases, successfully.

Periodic medical examinations are important since many incapacity determinants are temporary. Okpatu, Sibulkin, and Schenzler (1994) and Marasciulo (2004) show that granting incapacity benefits is not an objective process because incapacity is more subjective than the concept of sickness. Moreover, there is a recent rise in several incapacity determinants that are hard to diagnose, such as back pain (Meziat & Silva, 2011).

As previously discussed, there is no obvious explanation for the high Brazilian IA volatility. Although weak, since the 90s (Mueller, Rothstein, & Watcher, 2016), there is international evidence IA applications are countercyclical, related to unemployment rates (Coe, Haverstick, Munnell, & Webb, 2011; Duggan & Imberman, 2009; Mueller et al., 2016; Von Watcher, Song, & Manchester, 2011). However, according to Maestas, Mullen, and Strand (2015), although recessions may impact IA applications, many are rejected. Liebman (2015) and Banks et al. (2015) recognize that incapacity benefits and employment may be linked. They also suggest that demographic trends, population medical conditions, and benefit rules are important factors to explain IA trends. In particular, Coile et al. (2014) argue that benefit generosity explains the majority of variation in incapacity benefit incidence across countries.



1733

In Brazil, there is no clear association between the business cycle and IA applications: formal employment and IA trends do not correlate (Graph 4); IA applications and the unemployment rate do not correlate either (Graph 5).

GRAPH 4 IA APPLICATIONS AND THE NUMBER OF FORMAL WORKERS (2012-2018)

a maior parte delas é rejeitada. Além disso, Liebman (2015) e Banks et al. (2015) sugerem

que as tendências demográficas, as condições médicas da população e as regras aplicadas

à concessão dos benefícios são os fatores que melhor explicam a variação da incidência

do auxílio-doença entre os países. Reforçando essa visão, para Coile et al. (2014) a

generosidade na concessão dos benefícios pode explicar a maior parte da variação

observada na quantidade de pessoas recebendo benefícios por incapacidade nos diferentes

países.

No Brasil, não há associação clara entre o ciclo econômico e os pedidos por

auxílio-doença: não há correlação entre emprego formal e as tendências sobre o auxílio

(Gráfico 4), nem entre os pedidos e a taxa de desocupação (Gráfico 5).

Gráfico 4

Pedidos de auxílio-doença e número de trabalhadores formais (2012 a 2018)

Fonte: Elaborado pelos autores com base em dados do Boletim Estatístico da Previdência Social (BEPS) e

da PNAD Contínua (IBGE).

500,000

450,000

400,000

350,000

300,000

250,000

200,000

150,000

100,000

50,000

0

IA A

pplic

atio

ns (3

-mon

th m

ovin

g av

erag

e)

Jan-

12

Jul-1

2

Jan-

13

Jul-1

3

Jan-

14

Jul-1

4

Jan-

15

Jul-1

5

Jan-

16

Jul-1

6

Jan-

17

Jul-1

7

Jan-

18

Jul-1

8

49,000,000

48,000,000

47,000,000

46,000,000

45,000,000

44,000,000

43,000,000

42,000,000

41,000,000

40,000,000

Num

ber o

f for

mal

wor

kers

(3-m

onth

mov

ing

aver

age)

IA Applications Number of formal workers

Source: Elaborated by the authors, based on data from Boletim Estatístico da Previdência Social – BEPS (Social Security Statistical Bulletin) and Continuous National Household Sample Survey (PNADC – IBGE).

GRAPH 5 IA APPLICATIONS AND UNEMPLOYMENT (2012-2018)

Gráfico 5

Pedidos de auxílio-doença e desocupação (2012 a 2018)


da PNAD Contínua (IBGE).

Além disso, o número de requerimentos não parece ter mudado antes, durante ou

depois das quatro crises econômicas ocorridas no Brasil desde 2001 (CODACE, 2017),

como mostra o Gráfico 6.

500,000

450,000

400,000

350,000

300,000

250,000

200,000

150,000

100,000

50,000

0

IA A

pplic

atio

ns (3

-mon

th m

ovin

g av

erag

e)

Jan-

12

Jul-1

2

Jan-

13

Jul-1

3

Jan-

14

Jul-1

4

Jan-

15

Jul-1

5

Jan-

16

Jul-1

6

Jan-

17

Jul-1

7

Jan-

18

Jul-1

8

16.0%

14.0%

12.0%

10.0%

8.0%

6.0%

4.0%

2.0%

0.0%

Unem

ploy

men

t rat

e (%

)

IA Applications Unemployment rate

Source: Elaborated by the authors, based on data from Boletim Estatístico da Previdência Social – BEPS (Social Security Statistical Bulletin) and Continuous National Household Sample Survey (PNADC – IBGE).



1734

Additionally, the number of benefit applications does not seem to change before, during, or after the four economic crises that have occurred in Brazil since 2001 (CODACE, 2017). Graph 6 does not show a clear change in the pattern of IA applications before (from -6 to 0), during, or after recessions (from 0 to 38, depending on recession duration).

GRAPH 6 IA APPLICATIONS BEFORE, DURING, OR AFTER RECESSIONS

Gráfico 6

Pedidos de auxílio-doença antes, durante ou depois de momentos de recessão no Brasil


do Comitê de Datação de Ciclos Econômicos (FGV CODACE).

Por outro lado, os requerimentos de auxílio-doença (canal pelo qual se observa

como o ciclo econômico afeta o número de benefícios concedidos), bem como o seu

número de concessões e cancelamentos, oscilam em torno da mesma média desde 2005

(Gráfico 7).

500,000

450,000

400,000

350,000

300,000

250,000

200,000

150,000

100,000

50,000

0

IA A

pplic

atio

ns (3

-mon

th m

ovin

g av

erag

e)

-6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38Months since recession started

Recession 1 (Jul/01 - Dec/01) Recession 2 (Jan/03 - Jun/03) Recession 3 (Oct/08 - Mar/09) Recession 4 (Apr/14 - Dec/16)

Source: Elaborated by the authors, based on data from Boletim Estatístico da Previdência Social – BEPS (Social Security Statistical Bulletin) and the Brazilian Business Cycle Dating Committee (FGV CODACE).

Furthermore, the number of IA applications (which is the indicator of how the business cycle affects the number of IAs granted), and grants and terminations, oscillate around the same mean since 2005 (Graph 7).

GRAPH 7 IA APPLICATIONS, GRANTS, AND TERMINATIONS (2001-2017)

Gráfico 7

Pedidos de auxílio-doença, benefícios concedidos e cancelados (2001-2017)


Com essas evidências, é difícil explicar a oscilação nos números referentes ao

auxílio-doença, ou sua recente queda, com base apenas no ciclo econômico.

A trajetória mais recente de queda nos benefícios do auxílio-doença se iniciou por

volta de agosto de 2016, quando o Ministério do Desenvolvimento Social (MDS)

identificou, de um total de 1.827.225 benefícios, 563.771 auxílios-doença concedidos há

mais de dois anos e sem data de rescisão programada (Ministério do Desenvolvimento

Social, 2018).

Obtivemos esse banco de dados no MDS (Ministério do Desenvolvimento Social,

2018) e, após a exclusão de dados ausentes, inconsistentes ou incompletos, encontramos

476.163 benefícios (Gráfico 8). Desses, 99,7% haviam sido concedidos mais de quatro

anos antes da data de lançamento do PRBI (agosto 2016) e 238.902 haviam sido

concedidos por decisão judicial (50,2%). Tais números chamam a atenção, uma vez que

a média mensal de concessões de auxílio-doença nos dez anos anteriores à data de início

do PRBI (entre setembro de 2004 a agosto de 2014) foi de 189.145, sendo apenas 2,23%

delas judiciais.

Dec-

00Ju

n-01

Dec-

01Ju

n-02

Dec-

02Ju

n-03

Dec-

03Ju

n-04

Dec-

04Ju

n-05

Dec-

05Ju

n-06

Dec-

06Ju

n-07

Dec-

07Ju

n-08

Dec-

08Ju

n-09

Dec-

09Ju

n-10

Dec-

10Ju

n-11

Dec-

11Ju

n-12

Dec-

12Ju

n-13

Dec-

13Ju

n-14

Dec-

14Ju

n-15

Dec-

15Ju

n-16

Dec-

16Ju

n-17

Dec-

17

Grants

500,000

450,000

400,000

350,000

300,000

250,000

200,000

150,000

100,000

50,000

0

Num

ber o

f IA

mon

thly

gran

ts, t

erm

inat

ions

and

app

licat

ions

(3

-mon

th m

ovin

g av

erag

e)

Terminations Applications




1735

Based on this evidence, it is difficult to explain IA volatility or recent fall based only on the business cycle.

In August 2016, the Brazilian Ministry of Social Development (MDS) identified 563,771 IAs (out of 1,827,225) without a medical examination within two years. They had no termination dates. Therefore, these benefits were granted before August 2014 and had not been evaluated since (Ministério do Desenvolvimento Social, 2018).

We obtained this database from MDS and, after excluding missing, inconsistent, or incomplete data, we found 476,163 benefits (Graph 8). Of them, 99.7% started more than four years before the reference date, and 238,902 were granted by judicial rulings (50.2%). Such numbers drew attention, since average monthly IA grants during the ten years before the reference date (September 2004 to August 2014) was 189,145, and only 2.23% of grants were judicial.

GRAPH 8 DURATION OF ALL 476,163 BENEFITS (YEARS)

Gráfico 8

Duração de todos os 476.163 benefícios de auxílio-doença (número de anos)

Fonte: Elaborado pelos autores com base em dados do Ministério do Desenvolvimento Social (MDS).

Análises posteriores revelaram a causa do problema: as decisões judiciais

normalmente não especificam a duração dos benefícios de auxílio-doença que concedem

e, portanto, esses segurados nunca solicitaram reavaliação médica do INSS, já que

poderiam manter seus benefícios sem fazê-lo. Muitos desses trabalhadores continuaram

recebendo o auxílio por vários anos, mesmo depois de recuperados da sua condição

inicial.

Como resultado, uma parcela dos segurados com o auxílio recebia o benefício por,

em média, menos de um ano (Ministério do Trabalho e da Previdência Social, 2014),

enquanto o grupo que obteve a concessão via decisão judicial recebia os valores por

tempo indeterminado.

O Programa de Revisão dos Benefícios por Incapacidade (PRBI) foi lançado em

agosto de 2016 para lidar com essa questão. Ele permitiu ao INSS pagar horas extras aos

médicos peritos para a realização das perícias médicas dos segurados do auxílio-doença

que não eram examinados há mais de dois anos (um grupo de 563.771, o “grupo PRBI”).

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

90,000

80,000

70,000

60,000

50,000

40,000

30,000

20,000

10,000

-

Num

ber o

f ben

efits

IA duration (Years)Over 20

Source: Elaborated by the authors based on data from MDS.

Further investigation revealed the cause of the problem: judicial rulings typically do not specify a duration for IA benefits and, therefore, beneficiaries whose benefits were granted by judicial rulings never requested the INSS medical reevaluation because they could keep their benefits indefinitely without doing so. These workers continued to receive IAs several years after they had recuperated. It is a clear case in which a small problem persisted for a very long time, causing a big distortion many years later.

As a result, there was a group of IA beneficiaries receiving benefits, on average, for less than a year (Ministério do Trabalho e da Previdência Social, 2014), and another group, whose benefits were granted by judicial rulings, receiving benefits indefinitely.



1736

To solve these problems, the program for incapacity benefits assessment (PRBI) was launched in August 2016. The program allowed INSS to pay extra hours for INSS medical experts to evaluate IA benefits that had not been examined for more than two years, a group of 563,771 benefits, which we call the PRBI group. It also determined that new IA benefits granted by judicial rulings without explicit termination dates must be terminated or reevaluated 120 days after being granted.

PRBI started with Provisional Measure 739/2016. Congress did not vote in time, so it expired and had to be reissued in 2017 (Provisional Measure 767/2017). After intense discussions, Congress voted the new Provisional Measure (almost identical to the first one) and converted it into Law 13,457/2017. In the process, Congress made almost no relevant changes to the PRBI design.

Nevertheless, PRBI faced resistance in INSS. INSS administrative staff required extra payment for the extra hours they worked (without success), and medical staff needed to be motivated to do the evaluations since they could choose not to do them (the payment for extra hours were an important incentive, as well as the engagement from the National Association of INSS Medical Experts).

As Liebman (2015) points out, the United States tried a similar program, but it was discontinued because benefit terminations were perceived as unfair. From this perspective, one of the core elements for PRBI success was communication: monthly reports to the press about PRBI results and public disclosure on the most striking findings, such as people receiving IAs for over ten years for risky pregnancies, IAs for allegedly blind people who were found to be informally working as drivers, IAs for allegedly paraplegic beneficiaries who were professional runners (Ministério do Desenvolvimento Social, 2018). Communication was essential for PRBI sustained political support.

Despite the resistance, between August 2016 and October 2018, 91.7% of the PRBI group had been evaluated (436,642 benefits), and 74.8% of evaluated benefits were terminated (Table 1).

TABLE 1 PRBI RESULTS

IA benefits

PRBI benefits (database obtained from MDS) 476,163

Evaluated 436,642

Terminated 326,786

Not terminated 109,856

Source: Elaborated by the authors based on data from MDS.



1737

3. METHODOLOGY

To evaluate PRBI impacts on IA, this section extends the model from Liebman (2015).

3.1 Model

Number of IA beneficiaries in time t(B(t)) are a function of new IA granted (C(t)) and terminations (T(t)), as in equation (1).

B(t) = B(t – 1) + C(t) – T(t) (1)

Liebman (2015) assumes termination numbers depend on fixed death and recovery rates (D(t)), as in equation (2).

T(t) = DTB(t – 1) (2)

However, to understand the effects of nonlinear benefit duration changes on benefits, it is not possible to use equation (2). It is necessary to generalize the Liebman (2015) model considering nonlinear benefit termination structures.

Where p(i) is the probability of a benefit being terminated, i is the number of months after it was granted and P(i) is the cumulative distribution of p(i), we assume n is the smallest number for which P(i) = 1, for every i ≥ n. Therefore, the number of terminated benefits in time t is expressed by equation (3) and the number of active benefits by equation (4). Average benefit duration (D) is expressed by equation (5).

T(t) = Σn

i= 1 C(t – i) p(i) (3)

B(t) = C(t) + Σn

i= 1 C(t – i) (1 – P(i)) (4)

D = n – Σn – 1

i= 1 P(i) (5)

Liebman (2015) original model is a particular case of equations (3), (4), and (5) in which equations (2) and (6) hold.

DT =1

(6)D

Applying the expectation operator (E[■]) on equations (1), (4), and (5), it is possible to write the expected number of IA benefits (B) as equation (7).

B = DC (7)



1738

3.2 PRBI economic impacts

After PRBI was implemented, the cumulative probability distribution of benefit termination for new benefits changed from to P(i) to P'(i). This change implies new benefits follow another trend (B'(t)) and have another duration (D'). In this case, equations (1), (4), and (5) lead to equation (8). The first term of equation (8) is the long-term impact of this change assuming that all benefits were instantly affected. The second term represents the transition cost of having active benefits granted according to previous regulation.

E[B(t) – B' (t)] = C(D – D') – CΣn

i= t – k+1 (P'(i) – P(i)) (8)

To estimate the impact of medical reexaminations, it is necessary to compare medical evaluation costs (α) with the expected cash flow of terminated benefits, considering their probability of natural termination (P(i)). It is noteworthy that medical exams can increase expenditures if they only cease benefits with a high probability of natural termination.

The probability (Lt,j) a benefit conceded on date cj is still active on date t, considering it is active when the medical examination occurs (dj), is expressed in equation (9).

𝐿𝐿�𝑖� = 1 − �� (9)

Sendo 𝑉𝑉�𝑡𝑡� o valor médio dos auxílios-doença no tempo 𝑡𝑡, 𝑟𝑟 a taxa de juros real,

e 𝑘𝑘 o momento em que o PRBI foi implementado, a equação (10) mede o valor presente

líquido do impacto do PRBI no orçamento da Previdência Social.

𝐽𝐽�𝑘𝑘� = 𝑉𝑉�𝑘𝑘� �∑ ∑ �� 1 − ��

�� + 𝐶𝐶 �� −

∑ ∑ ��

�� – 𝛼𝛼 (10)

3.3 Calibração do modelo

Para simular o modelo, os parâmetros da equação (10) foram calibrados com o banco de

dados de benefícios anteriormente mencionado. A menos que explicitamente declarado,

todas as estimativas são baseadas nos 238.902 auxílios-doença concedidos por decisões

judiciais. Seja 𝐴𝐴�𝑖𝑖𝑖 𝑡𝑡� a quantidade de benefícios concedidos na data 𝑡𝑡 − 𝑖𝑖 e ainda ativos

na data 𝑡𝑡; e 𝑡𝑡̅ a data de lançamento do PRBI. A partir de 𝐴𝐴�𝑖𝑖𝑖 𝑡𝑡̅� e 𝐶𝐶�𝑡𝑡�, é possível calcular

a proporção dos benefícios efetivamente rescindidos concedidos há 𝑖𝑖 meses, 𝜇𝜇�𝑖𝑖�, usando

a equação (11).

𝜇𝜇�𝑖𝑖� = 𝐶𝐶��̅�𝑡−𝑖𝑖�−𝐴𝐴�𝑖𝑖𝑖�̅�𝑡�𝐶𝐶��̅�𝑡−𝑖𝑖� (11)

Como P(i) é uma distribuição de probabilidade cumulativa, é necessário que os

dados sejam filtrados para que se obtenha uma função em que 𝑃𝑃�𝑖𝑖� ≤ 𝑃𝑃�𝑖𝑖 + 1�. O Gráfico

9 mostra que nossos dados têm um padrão linear. Portanto, filtraremos μ(i) estimando a

tendência linear (12) por mínimos quadrados ordinários, em que 𝑢𝑢�𝑖𝑖� é um erro aleatório.

𝜇𝜇�𝑖𝑖� = 𝑎𝑎 + 𝑎𝑎𝑖𝑖 + 𝑢𝑢�𝑖𝑖� (12)

Os resultados da estimativa são mostrados na Tabela 2. Como esperado, o valor

estimado para 𝑎𝑎 é maior que zero, o que significa que 𝐸𝐸�𝜇𝜇�𝑖𝑖�� ≤ 𝐸𝐸�𝜇𝜇�𝑖𝑖 + 1��. As

(9)

Defining V(t) as the average value of IA benefits in time t, r as the real interest rate, and k as the moment PRBI was implemented, equation (10) measures the net present value of the overall PRBI economic impact on the social security budget.

𝐿𝐿�𝑖� = 1 − �� (9)




𝐽𝐽�𝑘𝑘� = 𝑉𝑉�𝑘𝑘� �∑ ∑ �� 1 − ��

�� + 𝐶𝐶 �� −

∑ ∑ ��

�� – 𝛼𝛼 (10)








a equação (11).









(10)

3.3 Model Calibration

To simulate the model, equation (10) parameters must be calibrated. To do so, we use the database of PRBI benefits. Unless explicitly stated, all estimates are based on the 238,902 benefits marked as being granted by judicial rulings. Where A(i,t) is the number of benefits conceded on date t – i still active on date t, and t is the date PRBI was launched, from an extraction of A(i,t) and C(t) it is possible to calculate the proportion of effectively terminated benefits granted i months ago, μ(i) , using equation (11).

𝐽𝐽�𝑘𝑘� = 𝑉𝑉�𝑘𝑘� �∑ ∑ �� 1 − ��

�� + 𝐶𝐶 �� −

∑ ∑ ��𝒋𝒋��𝒋𝒋��𝒏𝒏𝒋𝒋��

�� – 𝛼𝛼 (10)


To simulate the model, equation (10) parameters must be calibrated. To do so, we use the

database of PRBI benefits. Unless explicitly stated, all estimates are based on the 238,902

benefits marked as being granted by judicial rulings. Where 𝐴𝐴�𝑖𝑖𝑖 𝑖𝑖� is the number of

benefits conceded on date 𝑖𝑖 − 𝑖𝑖 still active on date 𝑖𝑖, and 𝑖𝑖 is the date PRBI was launched,

from an extraction of 𝐴𝐴�𝑖𝑖𝑖 𝑖𝑖� and 𝐶𝐶�𝑖𝑖� it is possible to calculate the proportion of

effectively terminated benefits granted 𝑖𝑖 months ago, 𝜇𝜇�𝑖𝑖�, using equation (11).

𝜇𝜇�𝑖𝑖� = ��𝑖�� (11)

Since 𝑃𝑃�𝑖𝑖� is a cumulative probability distribution, it is necessary to filter the data

in order to have a function in which 𝑃𝑃�𝑖𝑖� ≤ 𝑃𝑃�𝑖𝑖 + 1�. Graph 9 shows the data has a linear

pattern, so we will filter 𝜇𝜇�𝑖𝑖� by estimating the linear trend (12) by ordinary least squares,

in which u�𝑖𝑖� is a random term.

𝜇𝜇�𝑖𝑖� = a + bi + u�𝑖𝑖� (12)

The estimation results are shown in Table 2. As expected, the estimated value for

b is greater than zero, implying 𝐸𝐸�𝜇𝜇�𝑖𝑖�� ≤ 𝐸𝐸�𝜇𝜇�𝑖𝑖 + 1��. Estimates suggest the cumulative

benefit termination probability increases by 0.3 percentage points every month.

Since 𝑃𝑃�𝑖𝑖� must also follow the inequality 𝑃𝑃�𝑖𝑖� ≤ 1 and 𝑃𝑃�𝑛𝑛� = 1, we suppose

𝑃𝑃�𝑖𝑖� follows a linear trend between the largest 𝑖𝑖 for which 𝑃𝑃�𝑖𝑖� ≤ 1 and 𝑛𝑛. Where 𝑗𝑗 is the

greatest parameter for which 𝐸𝐸�𝜇𝜇�𝑗𝑗�� ≤ 1, such assumption means 𝑃𝑃�𝑖𝑖� follows equation

(13).

𝑃𝑃�𝑖𝑖� = �𝑖𝑖𝑛𝑛 �𝐸𝐸�𝜇𝜇�𝑖𝑖��𝑖 1 − �� 1 − 𝐸𝐸�𝜇𝜇�𝑗𝑗��𝑖 1� (13)

(11)



1739

Since P(i) is a cumulative probability distribution, it is necessary to filter the data in order to have a function in which P(i) ≤ P(i + 1). Graph 9 shows the data has a linear pattern, so we will filter μ(i) by estimating the linear trend (12) by ordinary least squares, in which u(i) is a random term.

𝐿𝐿�𝑖� = 1 − �� (9)




𝐽𝐽�𝑘𝑘� = 𝑉𝑉�𝑘𝑘� �∑ ∑ �� 1 − ��

�� + 𝐶𝐶 �� −

∑ ∑ ��

�� – 𝛼𝛼 (10)








a equação (11).









(12)

The estimation results are shown in Table 2. As expected, the estimated value for is greater than zero, implying E[μ(i)] ≤ E[μ(i + 1)]. Estimates suggest the cumulative benefit termination probability increases by 0.3 percentage points every month.

Since P(i) must also follow the inequality P(i) ≤ 1 and P(n) = 1, we suppose P(i) follows a linear trend between the largest i for which P(i) < 1 and n. Where j is the greatest parameter for which E[μ(j)] ≤ 1, such assumption means P(i) follows equation (13).

𝐽𝐽�𝑘𝑘� = 𝑉𝑉�𝑘𝑘� �∑ ∑ �� 1 − ��

�� + 𝐶𝐶 �� −

∑ ∑ ��𝒋𝒋��𝒋𝒋��𝒏𝒏𝒋𝒋��

�� – 𝛼𝛼 (10)


To simulate the model, equation (10) parameters must be calibrated. To do so, we use the

database of PRBI benefits. Unless explicitly stated, all estimates are based on the 238,902

benefits marked as being granted by judicial rulings. Where 𝐴𝐴�𝑖𝑖𝑖 𝑖𝑖� is the number of

benefits conceded on date 𝑖𝑖 − 𝑖𝑖 still active on date 𝑖𝑖, and 𝑖𝑖 is the date PRBI was launched,

from an extraction of 𝐴𝐴�𝑖𝑖𝑖 𝑖𝑖� and 𝐶𝐶�𝑖𝑖� it is possible to calculate the proportion of

effectively terminated benefits granted 𝑖𝑖 months ago, 𝜇𝜇�𝑖𝑖�, using equation (11).

𝜇𝜇�𝑖𝑖� = ��𝑖�� (11)

Since 𝑃𝑃�𝑖𝑖� is a cumulative probability distribution, it is necessary to filter the data

in order to have a function in which 𝑃𝑃�𝑖𝑖� ≤ 𝑃𝑃�𝑖𝑖 + 1�. Graph 9 shows the data has a linear

pattern, so we will filter 𝜇𝜇�𝑖𝑖� by estimating the linear trend (12) by ordinary least squares,

in which u�𝑖𝑖� is a random term.

𝜇𝜇�𝑖𝑖� = a + bi + u�𝑖𝑖� (12)

The estimation results are shown in Table 2. As expected, the estimated value for

b is greater than zero, implying 𝐸𝐸�𝜇𝜇�𝑖𝑖�� ≤ 𝐸𝐸�𝜇𝜇�𝑖𝑖 + 1��. Estimates suggest the cumulative

benefit termination probability increases by 0.3 percentage points every month.

Since 𝑃𝑃�𝑖𝑖� must also follow the inequality 𝑃𝑃�𝑖𝑖� ≤ 1 and 𝑃𝑃�𝑛𝑛� = 1, we suppose

𝑃𝑃�𝑖𝑖� follows a linear trend between the largest 𝑖𝑖 for which 𝑃𝑃�𝑖𝑖� ≤ 1 and 𝑛𝑛. Where 𝑗𝑗 is the

greatest parameter for which 𝐸𝐸�𝜇𝜇�𝑗𝑗�� ≤ 1, such assumption means 𝑃𝑃�𝑖𝑖� follows equation

(13).

𝑃𝑃�𝑖𝑖� = �𝑖𝑖𝑛𝑛 �𝐸𝐸�𝜇𝜇�𝑖𝑖��𝑖 1 − �� 1 − 𝐸𝐸�𝜇𝜇�𝑗𝑗��𝑖 1� (13) (13)

Our data indicates j = 195. There are only 104 judicial benefits with i > 195. Since n is the highest possible duration of benefits, the whole dataset of 476,163 benefits will be used to calibrate it: n = 463 . Graph 9 shows μ(i) and P(i) for i < j.

GRAPH 9 μ(i) E P(i) PARA i < j

Our data indicates 𝑗𝑗 𝑗 𝑗𝑗𝑗. There are only 104 judicial benefits with 𝑖𝑖 � 𝑗𝑗𝑗.

Since 𝑛𝑛 is the highest possible duration of benefits, the whole dataset of 476,163 benefits

will be used to calibrate it: 𝑛𝑛 𝑗 𝑛𝑛𝑛. Graph 9 shows 𝜇𝜇�𝑖𝑖� and 𝑃𝑃�𝑖𝑖� for 𝑖𝑖 � 𝑗𝑗.

Graph 9 𝜇𝜇�𝑖𝑖� and 𝑃𝑃�𝑖𝑖� for 𝑖𝑖 � 𝑗𝑗

Source: Elaborated by the authors, based on data from MDS.

Table 2

Estimates

Dependent Variable: 𝝁𝝁�𝒊𝒊�

Constant 0.347***

(0.0050)

Trend 0.003***

(0.0000)

Observations 171

R-squared 0.974 Source: Elaborated by the authors.

p-values in parentheses. *** significant at 1%.

Table 3 shows the values for the other parameters of equation (10).

Source: Elaborated by the authors, based on data from MDS.



1740

TABLE 2 ESTIMATES

Dependent Variable: μ(i)

Constant0.347***(0.0050)

Trend0.003***(0.0000)

Observations 171

R-squared 0.974

Source: Elaborated by the authors.p-values in parentheses. *** significant at 1%.

Table 3 shows the values for the other parameters of equation (10).

TABLE 3 MODEL CALIBRATION

Parameter Value Reference

V(k) R$ 1,303.83/month Average IA benefit (2016)

C 8,654 benefits/month Monthly average of judicial grants (2016 and 2017)

r 5.8% o.y. Public debt interest rates (NTN-B 2050) in September 2018.

α R$ 71.8 million Overall medical evaluation costs, according to MDS (Ministério do Desenvolvimento Social, 2018)

Source: Elaborated by the authors.

4. RESULTS

P(i) values and equation (5) suggest that the average duration of benefits granted judicially is 67.83 months. Therefore, the probability of an IA benefit lasting for ten years is significant. The average monthly judicial granted IAs, from September 2004 to August 2014 (exactly two years before the PRBI date), was 4,226. Therefore, equation (7) suggests that the PRBI group should have around 286,650 judicial benefits. Since there were 238,902 judicially granted benefits in the PRBI database, it seems the model is replicating real data well.

Equation (7) suggests the rise in monthly judicial granted IAs (from 4,226 in the period from September 2004 to August 2014 to 8,654, between 2016 and 2017) would cause the number of



1741

overall judicial IA benefits to rise to 586,963. Such a trend is in line with growing IA beneficiaries during 2016.

Moreover, from 2011 to 2016, overall monthly IA grants were 200,766, and the number of overall benefits varied from 1,374,454 to 1,827,225. Therefore, equation (7) suggests the overall benefit duration varied between 6.8 and 9.1 months. As a result, the average duration of benefits granted judicially is eight times longer than the average duration of overall benefits.

In the most extreme case, PRBI reduced the duration of judicial benefits to exactly four months (120 days). In this case, equation (10) suggests PRBI reduced social security expenditures by R$ 124.2 billion, and equation (7) suggests PRBI reduced the number of benefits by 552,348 in the long run.

However, there may be cases in which judicial benefits granted under the PRBI law are extended after medical reexaminations, bringing their duration close to that of all benefits. Graph 10 evaluates the robustness of the results and shows the estimated impact of PRBI on social security budgets when benefit duration is between four and twenty months. Impacts range from R$ 87.9 billion to R$ 124.2, concluding that PRBI can save more than R$ 85 billion of the social security budgets.

GRAPH 10 PRBI IMPACT ON SOCIAL SECURITY BUDGETS (R$ BILLION)

expenditures by R$ 124.2 billion, and equation (7) suggests PRBI reduced the number of

benefits by 552,348 in the long run.

However, there may be cases in which judicial benefits granted under the PRBI

law are extended after medical reexaminations, bringing their duration close to that of all

benefits. Graph 10 evaluates the robustness of the results and shows the estimated impact

of PRBI on social security budgets when benefit duration is between four and twenty

months. Impacts range from R$87.9 billion to R$ 124.2, concluding that PRBI can save

more than R$ 85 billion of the social security budgets.

Graph 10

PRBI impact on social security budgets (R$ billion)


Equation (7) suggests that the reduction in judicial benefits duration from 68.73

months to between 4 and 20 months caused the number of active IAs to fall by between

413,887 and 552,348, in all scenarios (Graph 11). So, the number of judicial granted IAs,

in the long run, falls from 586,963 to between 34,615 and 173,076.

These results are comparable to the actual overall fall in IA, considering the

reduction of 723,050 benefits between 2016 and 2019. The results also reinforce the link

between our simulations and real data, and the fact that the impact is permanent.




1742

Equation (7) suggests that the reduction in judicial benefits duration from 68.73 months to between 4 and 20 months caused the number of active IAs to fall by between 413,887 and 552,348, in all scenarios (Graph 11). So, the number of judicial granted IAs, in the long run, falls from 586,963 to between 34,615 and 173,076.

These results are comparable to the actual overall fall in IA, considering the reduction of 723,050 benefits between 2016 and 2019. The results also reinforce the link between our simulations and real data, and the fact that the impact is permanent.

GRAPH 11 PRBI IMPACT ON THE NUMBER OF IAS (LONG RUN)Graph 11

PRBI impact on the number of IAs (long run)


Equation (10) allows decomposition of the PRBI impact into three elements: a)

long-term effect (𝐿𝐿𝐿𝐿�𝑘𝑘�), measuring the long-term impact of changing the duration rules;

b) transition costs (𝑇𝑇𝑇𝑇�𝑘𝑘�), which adjust to the fact PRBI rules apply only to new benefits;

and c) medical examinations (𝑀𝑀𝑀𝑀�𝑘𝑘�), which reduce transition costs by reevaluating old

benefits. The decomposition is explained in equations (14) to (17).

𝐽𝐽�𝑘𝑘� = 𝑀𝑀𝑀𝑀�𝑘𝑘� + 𝐿𝐿𝐿𝐿�𝑘𝑘� + 𝑇𝑇𝑇𝑇�𝑘𝑘� (14)

𝑀𝑀𝑀𝑀�𝑘𝑘� = 𝑉𝑉�𝑘𝑘�∑ ∑ �� 1 −

��

�� – 𝛼𝛼 (15)

𝐿𝐿𝐿𝐿�𝑘𝑘� = 𝑉𝑉�𝑘𝑘�𝑇𝑇 �� (16)

𝑇𝑇𝑇𝑇�𝑘𝑘� = −𝑉𝑉�𝑘𝑘�𝑇𝑇 ∑ ∑ ��𝒋𝒋��𝒋𝒋��𝒏𝒏𝒋𝒋��

�� (17)


Equation (10) allows decomposition of the PRBI impact into three elements: a) long-term effect (LR(k)), measuring the long-term impact of changing the duration rules; b) transition costs (TC(k)), which adjust to the fact PRBI rules apply only to new benefits; and c) medical examinations (ME(k)), which reduce transition costs by reevaluating old benefits. The decomposition is explained in equations (14) to (17).

Graph 11 PRBI impact on the number of IAs (long run)








𝑀𝑀𝑀𝑀�𝑘𝑘� = 𝑉𝑉�𝑘𝑘�∑ ∑ �� 1 −

��

�� – 𝛼𝛼 (15)



�� (17)

(14)









𝑀𝑀𝑀𝑀�𝑘𝑘� = 𝑉𝑉�𝑘𝑘�∑ ∑ �� 1 −

��

�� – 𝛼𝛼 (15)



�� (17)

(15)



1743









𝑀𝑀𝑀𝑀�𝑘𝑘� = 𝑉𝑉�𝑘𝑘�∑ ∑ �� 1 −

��

�� – 𝛼𝛼 (15)



�� (17)

(16)









𝑀𝑀𝑀𝑀�𝑘𝑘� = 𝑉𝑉�𝑘𝑘�∑ ∑ �� 1 −

��

�� – 𝛼𝛼 (15)



�� (17)

(17)

Medical examinations account for R$ 13.9 billion of overall savings, while the long-term effect ranges between R$ 114.6 and R$ 152.9 billion, and transition costs are between R$ 40.6 and R$ 42.6 billion (Graph 12). Therefore, medical examinations reduced transition costs by approximately one-third and were responsible for 11% to 16% of the overall financial impact.

GRAPH 12 DECOMPOSITION OF PRBI IMPACT ON SOCIAL SECURITY BUDGETS (R$ BILLION)

Medical examinations account for R$ 13.9 billion of overall savings, while the

long-term effect ranges between R$ 114.6 and R$ 152.9 billion, and transition costs are

between R$ 40.6 and R$ 42.6 billion (Graph 12). Therefore, medical examinations

reduced transition costs by approximately one-third and were responsible for 11% to 16%

of the overall financial impact.

Graph 12

Decomposition of PRBI impact on social security budgets (R$ billion)


These results show that the program promoted equality by providing similar

termination rules to those who apply under administrative protocols and those who

require the benefit judicially. Reducing benefit distortions can improve social security

benefits rationality and reduce inequalities.

5 CONCLUSION

While the number of main Brazilian Social Security benefits is steadily increasing over

time, IAs have been steadily falling since 2016. The number of IAs in December 2019

was the lowest since October 2003. From 2016 to 2019, the number of IA benefits reduced


These results show that the program promoted equality by providing similar termination rules to those who apply under administrative protocols and those who require the benefit judicially. Reducing benefit distortions can improve social security benefits rationality and reduce inequalities.



1744

5. CONCLUSION

While the number of main Brazilian Social Security benefits is steadily increasing over time, IAs have been steadily falling since 2016. The number of IAs in December 2019 was the lowest since October 2003. From 2016 to 2019, the number of IA benefits reduced by 723,050. Understanding such dynamics is important to analyze whether benefit volatility is natural or caused by distortions in benefit rules.

The evolution of IAs over time is very volatile. Our analysis suggests such volatility cannot be explained by business cycles but by changes in benefit generosity. After expanding a canonical model for IA benefits, it was possible to measure the impact of PRBI on the social security budget. It lies between R$ 87.9 and R$ 124.2 billion, with an impact on the number of benefits between 413,887 and 552,348.

Such evidence suggests that a great proportion of the reduction in IA benefits since 2016 can be explained by PRBI, a program that combined the reassessment of old incapacity benefits with a change in IA rules that equalized the termination criteria for new benefits granted judicially or by the administrative protocol. The high rate of termination of the benefit after medical examination suggests that many IA should have been terminated several years before the program started. Therefore, PRBI not only saves social security a significant amount of resources, but it also seems to increase justice in the system.



1745

REFERENCES

Autor, D., Duggan, M., Greenberg, K., & Lyle, D. (2015). The Impact of Disability Benefits on Labor Supply: Evidence from the VA’s Disability Compensation Program (NBER Working Paper Series, 21144). Stanford, CA: Stanford Institute for Economic Policy Research.

Banks, J., Blundell, R., & Emmerson, C. (2015). Disability Benefit Receipt and Reform: Reconciling Trends in the United Kingdom. Journal of Economic Perspectives, 29(2) 173-190.

Coe, N., Haverstick, K., Munnell, A., & Webb, A. (2011). What Expains State Variation in SSDI Application Rates? (CRR WP 2011-23). Newton, MA: Center for Retirement Research at Boston College.

Coile, C., Milligan, K., & Wise, D. (2014). Social Security Programs and Retirement Around the World: Disability Insurance Programs and Retirement – Introduction and Summary (NBER Working Paper Series, n. 20120). Cambridge, MA: National Bureau of Economic Research.

Duggan, M., & Imberman, S. (2009). Why are the Disability Rolls Skyrocketing? The Contribution of Population Characteristics, Economic Conditions, and Program Generosity. In: D. Cutler, & D. Wise (Ed.), Health at Older Ages: The Causes and Consequences of Declining Disability among the Elderly. Chicago, IL: University of Chicago Press.

Economic Cycles Dating Committee. (2017). Economic Cycles Dating Report. Rio de Janeiro, RJ: Author.

Instituto Nacional do Seguro Social. (2002). Normative Ruling No. 78, of July 16, 2002. Union Official Journal, Brasília, DF, July 18, 2002.

Liebman, J. (2015). Understanding the Increase in Disability Insurance Benefit Receipts in the United States. Journal of Economic Perspectives, 29(2) 123-150.

Maestas, N., Mullen, K., & Strand, A. (2015). Disability Insurance and the Great Recession. American Economic Review: Papers & Proceedings, 105(5) 177-182.

Marasciulo, A. C. (2004). Performance assessment of the program of incapacity benefits of the Instituto

Nacional do Seguro Social /INSS, management branch of Florianópolis, SC, 2000-2002 (Doctoral Dissertation). Faculdade de Medicina, Universidade de São Paulo, São Paulo, SP.

Marinho, H., Resende, M., Filho, & Lucas, V. M. (2017). Análise do impacto da alteração normativa na aposentadoria por invalidez no Brasil. Nova Economia, 27(3) 551-576.

Meziat Filho, Ney and Silva, Gulnar. (2011). Invalidez por dor nas costas entre segurados da Previdência Social do Brasil. Revista de Saúde Pública, 45(3) 494-502.

Ministério da Fazenda. (2016). Anuário Estatístico da Previdência Social. Brasília, DF.

Ministério da Previdência Social. (2014). 2° Boletim Quadrimestral sobre Benefícios por Incapacidade. Brasília, DF.

Ministério do Desenvolvimento Social. (2018). PRBI, o maior programa de revisão de benefícios por incapacidade da história do INSS (Caderno de Estudos, 33). Brasília, DF: Author.

Ministério do Trabalho e da Previdência Social. (2014). Anuário Estatístico da Previdência Social Suplemento Histórico. Brasília, DF: Author.

Muellen, K., & Stefan, S. (2016). Disability Benefit Generosity and Labor Force Withdrawal (NBER Working Paper Series n. 22419). Cambridge, MA: National Bureau of Economic Research.

Mueller, A., Rothstein, J., & Watcher, T. (2016). Unemployment Insurance and Disability Insurance in the Great Recession. Journal of Labor Economics, 34(S1) S445-S475.

Okpaku, S., Sibulkin, A. E., & Christoph, S. (1994). Disability determinations for adults with mental disorders: Social Security Administration vs independent judgments. American Journal of Public Health, 84(11) 1791-1795.

Von Watcher, T., Song, J., & Manchester, J. (2011). Trends in Employment and Earnings of Allowed and Rejected Applicants to the Social Security Disability Insurance Program. American Economic Review, 101, 3308-3329.

http://pesquisa.bvsalud.org/portal/?lang=pt&q=au:%22Marasciulo,%20Ant%C3%B4nio%20Carlos%20Estima%22



1746

Vinícius Botelho

https://orcid.org/0000-0002-9627-1846Associated researcher at the Brazilian Institute of Economics (IBRE) of Fundação Getulio Vargas. E-mail: [email protected]

Raquel Maria Soares Freitas

https://orcid.org/0000-0001-8148-231XDegree in economic sciences from the Pontifical Catholic University of Rio de Janeiro (PUC-Rio); Coordinator of indicators production at the Department of Monitoring of the Secretary of Information Evaluation and Management of the Brazilian Ministério da Cidadania (DM/SAGI/MC). E-mail: [email protected]

Alberto Beltrame

https://orcid.org/0000-0001-7275-6120Physician; Former Minister of Social Development of Brazil. E-mail: [email protected]

saving more than r$ 85 billion of the brazilian social

Documents