The Targeting Incidence of Brazil’s Vale Transporte transit subsidy scheme

doi:10.21203/rs.3.rs-4450662/v1

Download PDF

Research Article

The Targeting Incidence of Brazil’s Vale Transporte transit subsidy scheme

https://doi.org/10.21203/rs.3.rs-4450662/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

We evaluate the targeting properties of the Vale Transporte scheme in Brazil, one of the oldest and most iconic transit subsidies in the Latin American region. We use household survey data for 26 metropolitan areas to calculate the Concentration Index and the Ω Index to determine whether benefits are accruing to workers belonging to low-income households. The results indicate that this program is badly targeted. The Concentration Index is negative in only 7 of the 26 metropolitan areas, indicating a regressive incidence of this benefit in most cities. Likewise, the Ω Index indicates that only in 3 metropolitan areas is the targeting incidence mildly progressive when the target population is defined as workers belonging to the poorest 40% of households in each city. The reason behind these results is that only formal sector workers are eligible for this benefit while many low-income individuals work in the informal sector in Brazil. In addition, since this subsidy is paid for by employers it is reasonable to expect compensating equilibrium wage effects in the formal labor market. We present evidence that suggests that this may have occurred.

transit subsidy

Brazil

targeting incidence

concentration index

Transport inequality, either related to accessibility or affordability, is a perennial research subject.¹ This stems from the importance that transport ―particularly transit― has on the ability of lower income individuals to access essential goods and services, such as education, health and labor opportunities.

Transit subsidies have received particular attention in the transport inequality literature. In part, this is motivated by the large amount of public funds spent subsidizing urban public transport systems in cities around the world. Although transit subsidies can be justified on economic efficiency grounds (Parry and Small, 2009), their distributive incidence is also relevant since equity and social objectives are often an explicit aim of these subsidies (Bondemark, et al., 2021) or figure prominently in policy discussions surrounding these benefits. However, with few exceptions, most research on this topic has been undertaken for developed countries (Silver, et al., 2023; Börjesson, et al., 2020; Bueno Caden, et al., 2016; Asensio, et al, 2003; among others). This is unfortunate since there are several reasons that make the incidence of these subsidies just as important, if not more, in developing countries.

First, transit subsidies in the global south are just as large and ubiquitous as they are in the developed world. For example, yearly operational transit subsidies in Latin America averaged 0.23% of GDP in 2018, ranging from 0.4% in Chile to 0.8% in Argentina, covering between 26% (Bogota) to 69% (Buenos Aires) of operating costs in 2019, a share comparable to developed country cities (Rivas, et al. 2020).

Second, developing countries face unique challenges for transit affordability and subsidies. Income is more unequally distributed, and poverty levels tend to be high. This exerts pressure on public budgets to provide educational, health, public services, housing and direct income support to poor households, increasing the opportunity costs subsidizing transit. Therefore, determining the targeting properties of these benefits is important to design policies that save scarce public resources by reducing leakages to richer non-deserving individuals.

The new post Covid-19 fiscal context adds an additional urgency to this last issue. The financial deficits of transit systems have increased as demand for public transport has yet to recover pre-pandemic levels. Given the overall fiscal strain faced by most countries, financing these deficits is becoming very challenging to transport authorities and fare increases may be inevitable to achieve financial sustainability. Increasing fares without unduly affecting the mobility needs of the poor requires better targeted subsidies.

Finally, the informal sector constitutes a large fraction of the economy in developing countries and policies, such as formal sector subsidies, may not benefit the poor who disproportionally operate in the informal sector. In addition, poor individuals tend to rely more on informal services like vans and micro buses out of reach of many formal transit policies (Cervero and Golub, 2007; Serebrisky, et al, 2009; Golub, et al., 2009). In this context, evaluating the targeting properties of formal transit subsidies is important for social equity considerations and economic welfare.

In this paper we analyze the distributive incidence of the Vale Transporte scheme in Bazil. This is a nationwide benefit whereby employers must provide workers with transit vouchers for work related trips to and from their residence. In exchange, they can deduct up to 6% from worker’s wages.² This last feature implies that only lower wage workers whose commuting costs exceed this wage threshold will benefit from the scheme, while higher paid workers will opt-out of the benefit. In principle, this should make the targeting incidence of this policy progressive. However, as will be discussed below, this is counteracted by the fact that only formal sector workers benefit from the scheme. We also analyze whether general equilibrium effects in the labor market may further affect the incidence of this subsidy.

Brazil is an interesting case study because Vale Transporte is one of the oldest and most cited demand side transit subsidies in Latin America (Gwilliam, 2017; Vasconcelos, 2018). Although Gwilliam (2017) has noted its potential drawbacks, as far as we are aware there is no published empirical research on the economic and distributive properties of this scheme. In addition, Brazil has a large informal sector representing close to 40% of employment. Thus, the incidence of this subsidy in the presence of a large informal sector can be analyzed, highlighting the issues that may restrict the use of such schemes in the developing world. This may differ significantly from the targeting properties and behavioral effects of analogous employer paid transit benefits in cities of developed countries, such as the “Versement Transport” in France³ or the experience from an experiment in Vancouver, Canada (Hall, et al., 2021). In sum, we add to the scarce literature on the distributive incidence of transit subsidies in the developing world by presenting an important and relevant case study.

From a methodological perspective, we use the Concentration Index approach proposed by Kakwani (1977) to determine the distributive incidence of the Vale Transporte scheme. As argued by Karner, et al. (2024), this approach has several advantages over alternative measures of inequality, but it also presents some limitations. Therefore, we also calculate the Ω index commonly used to measure the targeting properties of public utility subsidies (Coady et al.,2004; Komives et al., 2005; Gómez-Lobo, et al., 2023).

Using individual level data for 26 metropolitan areas of Brazil taken from the 2020 PNAD household survey we show that the Vale Transporte is in general poorly targeted. The concentration index shows a moderately progressive targeting incidence in only 7 of the 26 metropolitan areas. Furthermore, if the poorest 40% of households in each city are considered as the target population, the Ω index indicates that only in 3 of the 26 cities is the share of benefits accruing to this group higher than its share in the population. The explanation for these results is not hard to find. Since the vouchers only benefit formal sector workers, informal workers, who are on average poorer, do not benefit. Beneficiaries tend to be middle class workers. Furthermore, we find evidence to suggest that wages in the formal sector may have adjusted to render the subsidy even less effective.

This paper is organized as follows. In the next section we review the literature on the incidence of transit subsidies, including a discussion of the methodological approaches used by previous researchers on this topic. We then describe the Vale Transport scheme applied in Brazil and the data used in this paper. The paper then presents the incidence results together with some suggestive evidence for general equilibrium effects in the labor market that may further reduce the progressivity of the Brazilian scheme. The paper concludes with a summary of the main findings and a discussion of policy lessons for other developing countries.

There is some older literature from the 70’ and 80’ which we will not cover and concentrate on the more recent publications.⁴ The incidence of transit subsidies has been studied for urban areas in Spain by Asensio, et al. (2003). They estimate a transit expenditure equation and then simulate the impact of subsidies on this expenditure for different household incomes. They find that urban transport subsidies are progressive, in the sense that the subsidy received as a percentage of income decreases with income. They also calculate the difference in the Gini coefficient with and without transit subsidies and find that these benefits improve the income distribution, but the change is very modest owing to the relatively small share of transit within total household expenditure.

Bueno Cadena, et al. (2016) used the 2004 Madrid mobility survey to estimate the distributive incidence of transit subsidies by estimating travel card use by neighborhood. They find that travel card use is higher in poorer neighborhoods suggesting a progressive incidence of benefits.

Börjesson, et al. (2020) study the incidence of transit subsidies in Stockholm, Sweden. The novelty of their approach is that they estimate subsidies by trip link considering differences in costs and loads by mode and service. They then impute the subsidy received by passengers according to the fare paid and the cost of the different links used in their trips. This granular approach effectively identifies cross subsidies among transit users and thus gives a more detailed picture of the distributional impact of benefits. They then use a Concentration Curve index to estimate the incidence of these transit subsidies and conclude that they are mildly progressive, but perhaps less than would be expected ex-ante. They also analyze the incidence by different social groups and simulate the distributive effects of different fare structures.⁵

Silver, et al. (2023) study the distributive incidence of the 2019 fare reform In Lisbon, Portugal, that simplified the fare structure from over 300 passes to just one pass with a flat rate. They use a generalized cost approach to analyze the incidence of this reform by demographic groups. They also use the Palma Ratio and conclude that the reform reduced inequality; the transit expenditure reduction of the 40% lowest income individuals was larger than the expenditure reduction for the highest 10% of income.

Studies for developing countries are few. Interestingly, available research often suggests different incidence results to those reported in the above publications for developed countries. For example, Flynn (2007) shows that transport subsidies in Mexico City are not as pro-poor as is commonly believed. The majority of the poor do not benefit from the metro, public bus, trolley, and light railway subsidies that form the bulk of transport subsidies in that city. Bhattacharya and Cropper (2007) find similar results for bus and rail subsidies in Mumbai, India. Gómez-Lobo (2009) studies the distributive impact of a series of transport related taxes and subsidies in Santiago, Chile, including bus and metro fares. The novelty of this last study is that in some cases subsidies are funded through cross subsidies and thus the incidence of revenues is also included in the analysis. Although subsidies are mildly progressive, using cross-subsidies to fund student concessionary fares reduces the positive impact of this benefit. Vecchio, et al. (2024), also for Santiago, study the fairness of different transit policies, including fare subsidies, but only for a specific demographic group (the elderly).

Another literature deals with the transport behavior or labor market outcomes induced by transit subsidies. For example, Brough, et al. (2022) for a voucher experiment in Kings County, Washington, Guzman, et al. (2023) for a randomized controlled experiment with transit voucher and Guzman and Hessel (2022) for a means-tested transit subsidy, both in Bogotá, Colombia, Franklin (2017) for a randomized experiment with transit subsidies for unemployed youth in urban Ethiopia, Hall, et al. (2021) for an employer paid transit subsidy in Vancouver, Canada, and Bull, et al. (2020) on a randomized control trial with free transit fares in Santiago, Chile. Special mention should be given to Moreno-Monroy and Posada (2018) who study the effects of different types of transit subsidies on informality using a linear city urban economic model. They find that subsidies targeted to informal workers do not increase welfare or reduce informality rates. Since all these papers deal only tangentially with subsidy incidence, we will not discuss them further.

For Brazil we have been unable to find any published research analyzing the Vale Transporte scheme. Herszenhut, et al (2022) examine the impact of including monetary costs in access inequality measures in Rio de Janeiro. However, they do not examine the incidence of transit subsidies. Closer to the objective of the present paper is Proque, et al. (2022) who use a dynamic general equilibrium model to simulate the economic and distributive effects of eliminating fuel taxes or using fuel tax revenues to finance lower bus fares in Brazil. Proque, et al. (2023) use a similar model to study the effects of urban transport subsidies. However, in both cases the analysis is very aggregate. For example, Proque, et al. (2023) use five representative households, annual operational subsidies for public transport at the national level and do not consider the large informal sector of the Brazilian economy. The Vale Transporte scheme is not mentioned in these papers.

From a methodological perspective, recent studies have used a Concentration Curve index (CI) approach (Gómez-Lobo, 2009; Börjesson, et al., 2020; Karner, et al., 2024) that is also common in the study of public utility subsidies (Palma, et al., 2018; Gómez-Lobo, et al. 2023).⁶ This consists of graphing a Lorenz curve of the incidence of the subsidy as shown in Fig. 1. Let I represent income and F(I) is the cumulative distribution function of income. Likewise, define S(I) as the cumulative distribution of benefits, that is, the proportion of total benefits received by individuals or households with incomes I or less. The Lorenz curve is then the graph of F(I) and S(I). If we define A as the area under the Lorenz curve, then the CI (Kakwani, 1977) is equal to $CI=1-2\bullet A$ and is bounded between (-1,1).⁷ If the Lorenz curve is below the 45-degree line, the index is positive and indicates a regressive incidence of the benefit since the proportion of total benefits accruing to poorer families is below their proportion in the population. A negative value indicates a progressive targeting of the benefit.

The CI of Kakwani (1977) is similar to the Suits (1977) index, however, this last index is calculated from a Lorenz curve where the horizontal axis measures the proportion of total income accumulated by individuals or households, rather than their proportion in the population. As explained in Börjesson, et al. (2020), in this case the Lorenz curve will coincide with the 45-degree line if benefits are proportional to income, indicating that benefits over income is the same for all individuals or households. In contrast, in the Kakwani (1977) approach the 45-degree line represents a scenario where all individuals or households receive the same absolute amount of the benefit. The Suits (1977) approach would be relevant if one wishes to examine the progressiveness of the subsidy in a vertical equity sense. If everyone receives the same absolute value of benefits it would be a progressive subsidy according to the Suits (1977) index, since lower-income individuals receive a higher proportion of their income in benefits, while it would be a neutral subsidy under the Kakwani (1977) index.

Since our interest lies in the targeting properties of the Vale Transporte subsidy, we will use the Kakwani (1977) Concentration Curve index. We want to explore how well this benefit targets low-income individuals and it is in this sense that we will label the subsidy progressive or regressive. For clarity we call this the targeting incidence of the subsidy to distinguish it from the distributive incidence in a vertical equity sense.

One potential problem with the CI noted by Karner, et al. (2024) is that it is influenced by the incidence of benefits along the whole range of the income distribution with equal weight given to low and high-income beneficiaries.⁸ One alternative to overcome this problem would be to follow Karner, et al. (2024) and use the “Corrected Concentration Index” of Erreygers (2009) whereby weights are inversely proportional to income rankings. Another alternative would be to use the Palma Index used by some researchers and which in the present context implies comparing the average subsidy received by the 10% richest individuals to the average subsidy received by the 40% lowest income individuals. However, we prefer to use the Ω index originally used by Coady et al. (2004) and Komives, et al. (2005). This index compares the average benefit received by a predefined target population (e.g. the poor) to the average benefit received by all beneficiaries and is defined as follows:

$${\Omega }=\frac{\frac{{S}_{P}}{{S}_{T}}}{\frac{P}{T}}$$

Where P is the number of individuals in the target group, T is the total number of individuals in the relevant population, ${S}_{p}$ is the total subsidy received by the targeted population and ${S}_{T}$ represents the total amount of the subsidy distributed. This means that ${\Omega }$ is the share of the subsidy received by the target population over the share of this group in the total population.⁹ If this indicator is less than one, then it implies that the share of the subsidy accruing to the target population is less than proportional to the share of this group in the population, an indication of regressivity in the targeting of the benefit. If it is greater than 1 then it is progressive (and more so the higher above 1 is this indicator).

One can calculate the Ω index for different definitions of the target population using alternative cut-off points of the income distribution. For example, one target population could be individuals in the lowest three deciles of the income distribution, while an alternative target population could be individuals in the four lowest deciles of the income distribution, and so on. If deciles of the income distribution are used to define the target population, then it is useful to note that the Ω index can be read directly from the Lorenz curve.¹⁰ It is equal to the slope of the ray from the origin to the point on the Lorenz curve for the rank threshold defining the target group. It will be less than one if the Lorenz curve is below the 45° line at that point and greater than 1 in the opposite case.

One final point to make is that the incidence of the funding side of the subsidy is also relevant. If some firms increase prices as a result of the Vale Transporte scheme, then how these price increases affect different income groups should be analyzed. Likewise for the public revenue sources for the proportion of benefits funded through tax revenues or deductions. Although analyzing these issues is beyond the scope of this paper, we will present evidence for a third possible funding channel, namely that wages in the formal sector decreased because of the policy. In this case, it would be beneficiary workers themselves who are funding to some extent their own benefits, reducing the effectiveness of the scheme.

Supply side subsidies in Brazil cover on average 29% of operational costs, with this share reaching up to 50% in some metropolitan areas such as Sao Paulo and Brasilia (NTU, 2023). However, supply-side subsidies may induce operational inefficiencies that impede all these resources reaching passengers. In addition, these operational subsidies are city specific and rely on municipal or state funding. Thus, there is a wide range of experiences with some cities subsidizing close to 50% of operational costs while others none.

In contrast, Vale Transporte is a Federally mandated policy and is a demand side subsidy to help workers pay their commuting costs. It was introduced in 1985 through a national law (Law Nº 7.418 of 1985 and regulated by the Decree Nº 95.247 of 1987). Employers must provide workers with transit vouchers for work related trips to and from their residence. In exchange, they can deduct up to 6% from worker’s wages. Workers who earn low wages and whose commuting transport expenditure is more than 6% of earnings will benefit from the scheme. Those with higher wages will be indifferent since their expenditure, if lower than 6% of earnings, will be exactly offset by the wage reduction.¹¹ To give an idea of the scale of the benefit, in 2019, according to information obtained directly from the transport authority (SPTrans), Vale Transporte trips in the Sao Paulo transit system represented 30.3% of total passengers and 14% of total revenues.

The employer must provide an employee with vouchers for the full trip to and from work, including all segments when a trip has multiple links with different operators and fares. Consequently, employees who live further away receive a higher benefit. The scheme applies to all urban public transport modes as well as to intermunicipal or interstate transport that have the characteristics of urban services.

If the cost of the transit vouchers for a particular worker is higher than the wage deduction limit of 6%, then it is the employer who must pay the difference. However, since this net excess expenditure on vouchers is tax deductible, about one third of the cost is funded by the government in the form of lower tax revenues. The cost (above the wage deduction) may be substantial. Gwilliam (2017) reports that in the second income quintile, individuals benefit to the amount of 30% of their incomes, much higher than the 6% deduction.

In practice, the scheme works as follows. At the beginning of each month, firms purchase vouchers from a syndicate of transit companies, a clearing house operated by the government or a private bank that acts as a “selling agency” (Gwilliam, 2017). The vouchers are then transferred to workers who then can use them to pay for their trips. The transit companies later exchange the Vale Transporte vouchers received from passengers for cash from the selling agency. Depending on the city, the benefit can take the form of tickets, tokens or electronic credits in transit systems that have a smartcard payment technology.

One of the unintended consequences of the scheme is the emergence of an active market for vouchers. Gwilliam (2017) reports that in the year 2000 about 25% of recipients sold their vouchers on the black market rather than using them. The introduction of non-transferable individual smartcards has decreased this phenomenon. However, as noted by Gwilliam (2017) there could be other non-intended effects of the subsidy. For example, firms may be less willing to hire workers who reside far away from their premises. This could reduce productivity through lower matching quality in the labor market. Firms may also pass- through to prices part or all the costs of the scheme.

Another unintended effect may be through wage accommodation in the formal sector. Low-skilled workers may internalize the lower commuting costs if hired and may thus be willing to work for a lower wage compared to a counterfactual without the scheme. If the minimum wage is binding in the formal sector, a possible consequence is higher unemployment for low-skilled workers who search for formal sector jobs.

4.1. Survey data

We use individual level data for 26 metropolitan areas of Brazil taken from the 2020 PNAD household survey. This survey is statistically representative for these metropolitan areas. We use data for the first quarter of that year to avoid any distortions caused by the Covid pandemic.

The data was trimmed eliminating observations from households with incomes in the lowest and highest 5% percentile of the income per capita distribution in each city. This eliminates individuals from zero income households. Excluded observations in the bottom 5% percentile are for the most part (99.4%) not employed and 0.6% are employed but in informal work.

For commuting expenditure, we use data from Expatistan for 2021. This is a crowdsourced platform that registers harmonized cost of living information for different countries and cities around the world. For this research, we use the public transport data, which indicates the average value spent in public transportation by residents of different cities in Brazil. These expenditures were deflated using the national inflation rate between 2020 and 2021. Results are shown in Fig. 2.

The criteria to define if an individual is eligible to receive the subsidy is that the individual was employed in a formal job for at least 15 hours a week. Formality is defined as those workers who pay social security contributions or those who own firms where financial results of productive activities can be separated from those of its owner’s. According to the International Labour Organization 38.2% of employment is informal in Brazil (ILOSTAT Database, 2022).

According to our estimates, 26% of the Brazilian population in these metropolitan areas are eligible to receive the subsidy, e.g., 21.4 million people. This represents 55% of employed individuals. However, not all individuals that are in the category Employed and eligible necessarily take up the subsidy, since it will only be rational for those whose cost savings is higher than 6% of wage earnings. We assume a best-case scenario where all eligible formal sector workers who stand to benefit from the subsidy take it up.

4.2. Imputing the subsidy benefit

The Brazilian PNAD household survey reports the monetary gross earnings¹² including all monetary subsidies. We assume that gross labor income for workers who take-up the Vale Transporte scheme includes the value of the transport vouchers net of the wage reduction.¹³ The reported labor earnings for a formal sector worker would then be:

$${\stackrel{-}{w}}_{i}=\text{max}\left[{M}_{i}, {M}_{i}\left(1-0.06\right)+\theta *{S}_{i}\right]$$

where ${\stackrel{-}{w}}_{i}$ stands for the gross labor income as reported in the PNAD; ${M}_{i}$ represents the monetary gross labor earnings received by the worker before the Vale Transporte benefit and deduction; $\theta$ is a parameter that equals one for full-time workers and a proportion less than one for part-time workers; and ${S}_{i}$ represents the transit commuting expenditure for a full time worker. This last parameter is equal to the monetary value of transport vouchers given by employers to their full-time workers. This equation assumes that all workers that stand to benefit from the scheme take-up the subsidy.

${M}_{i}$ is not observed in the PNAD surveys but can be deduced from reported earnings through the following expression deduced from Eq. (1):

$${M}_{i}=min\left[{\stackrel{-}{w}}_{i},\frac{{\stackrel{-}{w}}_{i}-\theta \bullet {S}_{i}}{1-0.06}\right]$$

Equivalently, the worker will take the subsidy as long as:

$${\stackrel{-}{w}}_{i}\le \frac{{\theta \bullet S}_{i}}{0.06}$$

This means that a formal sector worker will take the subsidy if her/his reported wage is below 16.67 times the commuting cost —multiplied by the proportion of full-time work the individual undertakes.

The net monetary amount of the benefit received by worker i in city c is:

$${B}_{i}=Max\left[\theta \bullet {S}_{i}-0.06\bullet {M}_{i},0\right]\bullet I(i=formal)$$

where ${S}_{i}$ represents the transit commuting expenditure required by worker i, and $I$ is an indicator function that takes the value of one if the worker is in formal employment and cero otherwise. Using Eq. (2) for those who benefit from the subsidy we obtain an expression of this amount as a function of observable variables:

$${B}_{i}=\frac{1}{(1-0.06)}\bullet Max[\theta \bullet {S}_{i}-0.06\bullet {\stackrel{-}{w}}_{i},0]\bullet I(i=formal)$$

If contrary to our assumption above, the surveys register gross earnings before the effects of the Vale Transporte scheme, then condition (3) is unchanged and the monetary benefit of the scheme to each worker is the same as (4) except that it does not have the leading fraction.

We do not have information on individual worker’s commuting expenditure and so ${S}_{i}$ is replaced by the average transit expenditure shown in Fig. 2. This may overestimate the benefit to those workers residing close to their employment and underestimate it for those residing far away. However, since distance-based transit fares are uncommon and many Metropolitan Areas in Brazil have integrated transit fares, using average city transit expenditure may not significantly bias our results.

We estimate that out of the 21.4 million eligible workers, those with sufficiently low income to obtain a positive benefit from the scheme amount to 16 million in the 26 metropolitan areas. The subsidy amount was imputed for each of these observations using Eq. (4) above. In the aggregate, the subsidy ―net of the part paid by workers― in these cities amounted to 1.8 billion Reals (US$410 million) a month, which annually represents 0.3% of national GDP.¹⁴ Sao Paulo and Rio de Janeiro account for 943 million reals in subsidies, which is more than half the total amount for the 26 metropolitan areas.

We present the incidence results by city, although aggregating the data for the 26 metropolitan areas gives qualitatively similar results. We only compare the targeting incidence among workers (formal and informal). However, we rank workers in the income distribution according to their household’s income per capita in each city. This is a better gauge of the relative socioeconomic standing of each worker compared to using their individual wage earnings.

Figure 3 presents the Concentration Curve index (CI) by city. Only 7 out of the 26 metropolitan areas have a negative CI. In all the others the targeting incidence of the subsidy is regressive with the average over the 26 metropolitan areas slightly positive. Florianopolis has the most progressive targeting while Manaus the most regressive. It is also useful to note that all the CIs are quite small. Therefore, although some cities exhibit regressive while others progressive targeting properties, the CIs are not too far from a neutral targeting result.

Figure 4 presents the Lorenz curves for each metropolitan area. The CI results shown above are related to the regressive properties of the Lorenz curve at the bottom of the income distribution. In all cities the Lorenz curve implies a regressive distribution for the poorest workers. Only when the curve crosses the 45-degree line is there some progressivity. In none of the metropolitan areas does the Lorenz curve cut this line before the 30th percentile of the income distribution and in many not even before the 60th percentile.

As for the ${\Omega }$ index, this can be read right off the Lorenz curve for each city as mentioned above. If the target group is defined as workers in the lowest 30% of the distribution, then the ${\Omega }$ index would be below 1 for all metropolitan areas, indicating that the amount of the subsidy going to this group is less than 30% of total benefits.

Figure 5 presents this last index assuming that the target population is 40% of workers in the lowest part of the income distribution. Only three cities present a value over 1.0 —Florianopolis, Curitiba, and Brasilia—, meaning that most of the cities give a lower share of the subsidy to workers in the bottom 4 deciles of the income distribution compared to higher income groups. Manaus is the metropolitan area with the lowest coefficient presenting a value of 0.5.

The results from the Concentration Index and the ${\Omega }$ index all imply that the Vale Transporte scheme is not well targeted in general. It tends to benefit more workers from middle to high-middle income households. The main reason behind these results is that there is a high proportion of informal workers among lower income households. Figure 6 graphs the labor market informality (informal workers as proportion of all employed workers) and the Concentration Index for each city, with the dotted line showing the linear regression between these two variables. Notably, the ${R}^{2}$ is 0.62 and the slope is positive and strongly significant to a 1% confidence level, suggesting that informality tends to explain particularly well the lack of progressiveness of the subsidy.

Finally, our qualitative results do not change if we increase or decrease the average transit expenditure in each city by 10%. The Concentration Index is not very sensitive to this parameter.

The above results assume that there are no equilibrium market effects. This is unrealistic considering that the net subsidy that firms must pay represents 0.3% of national GDP.

The scheme may generate various market equilibrium changes. First, firms may increase the sales price charged for their goods and services as the Vale Transporte scheme amounts to an increase in labor costs. Second, firms in the formal sector may lower their demand for low skilled labor. In addition, as mentioned above, if low skilled workers internalize the benefit of the scheme, they may be willing to work in the formal sector for a lower salary. Both effects may decrease wages for lower wage earners in the formal sector. If these wages cannot decrease (due, for example, to a minimum wage binding constraint) then this may generate higher unemployment among lower income workers or incentivize their migration to the informal sector.

With the data at hand, we can explore whether there is evidence for an equilibrium effect on wages. For this purpose, we estimate nearest neighbor regressions based on a propensity score matching (PSM) on the hourly wages between formal and informal sector workers.¹⁵ Furthermore, we undertake this regression analysis using several subsamples of the data to see if there are different results along different segments in the labor earnings distribution. The subsamples are presented in Table 1 and are based on the relative labor earnings with respect to the wage cutoff that determines a positive subsidy amount.

Table 1

Sample used in the wage comparison model
Sample	Observations in each sample
Baseline	Whole sample
Sample 1	100–110% wage cut-off point
Sample 2	90–100% wage cut-off point
Sample 3	70–90% wage cut-off point
Sample 4	50–70% wage cut-off point
Sample 5	Below minimum wage

Table 2 presents the regression results. The covariates to determine the matching propensity score in every model were: experience; experience squared; years of education; gender; economic sector; and city of residence. The dependent variable is the wage per hour —the first row of the ATET coefficient indicates that the wage used included the subsidy, whereas the second row excluded the subsidy—. Formality has a positive and significant effect on hourly wages in most of the models.

The results are suggestive. If we consider workers that are above but close to the wage threshold, we see that wages before the subsidy are higher in the formal sector than in the informal sector. However, as we move down the wage gradient, the wage premium for formal sector workers decreases. This is consistent with a negative wage effect of the Vale Transporte scheme since the transport benefit is higher for lower wage earners. For wages close to the threshold the benefit is very small and thus one would not expect any compensating impact on wages.

For very low wage earners (sample 5) the formal wage premium increases compared to workers in sample 4. It is possible that the minimum wage restriction is active for these low-income workers and thus formality has a large impact on formal wages compared to informal wages. What we cannot test is whether unemployment is higher for workers in this group, as would be expected if the minimum wage restriction is active for these formal sector workers.

Table 2

Nearest neighbor regression results
	Baseline	Sample 1	Sample 2	Sample 3	Sample 4	Sample 5
ATET (Formal vs. informal)
Wage after subsidy	-0.85***	5.78***	7.91***	4.88***	2.71***	3.87***
Wage before subsidy	-2.74***	5.78***	7.65***	3.77***	0.59***	0.69***
Observations	22,674,581	818,401	839,906	2,508,821	3,976,702	5,140,006
Min neighbors	9	34	54	34	31	22
Max neighbors	113,874	9,716	8,690	19,876	41,744	39,597
Min requested	3	3	3	3	3	3

*p < 0.1; **p < 0.05; ***p < 0.01.

Our propensity score matching estimator of wage differences between formal and informal workers rests on the assumption that formality or informality, conditional on the observables that determine the propensity score, is random. This may not be realistic since there may be unobservable characteristics that determine whether a given worker finds formal employment or not. However, our results show that the formal wage premium is U shaped according to the expected wage rate. This is consistent with prior expectations if the Vale Transporte scheme does have a compensating effect on equilibrium formal sector wages. The lower the wage rate the stronger the compensating effect on equilibrium wages since the subsidy is higher for low-income workers, thus lowering the formality wage premium. However, for very low wage earners the minimum wage restriction kicks in and the formal sector wage premium rises.

Although more research is warranted on this issue, we can tentatively say that there is suggestive evidence that the Vale Transporte scheme does generate some compensating effects on equilibrium wages in the formal sector. If this is so, then any distributive effects of the scheme will be more neutral than what was estimated above.

In this paper we analyze the Vale Transporte transit subsidy in Brazil. Although it is not a formal means-tested subsidy, the institutional design is such that only lower wage earners benefit from this program and thus it is indirectly targeted. One important drawback of the Brazilian scheme is that this subsidy only benefits formal sector workers. In a country with an important informal labor market, this last feature may imply that the distributive incidence of benefits is regressive.

To test this last proposition, we used the PNAD household surveys for 26 metropolitan areas of Brazil, impute the subsidy benefit to eligible workers and estimate the distributive incidence of the benefit. To the best of our knowledge, this is the first attempt to estimate the targeting properties of this transit subsidy.

We find that the Vale Transporte scheme is badly targeted in general. In some cities, particularly the larger ones, the subsidy has a slightly progressive impact measured by the Concentration Index. However, if we define the target population as individuals belonging to 40% of lowest income households in each city, this group receives a lower share of the subsidy benefit compared to higher income households. The exceptions are three cities: Florianópolis, Curitiba and Brazilia. These cities are also the ones with the lowest share of informal employment to total employment, pointing to the main cause of the poor targeting properties of this subsidy; namely, the exclusion of informal sector workers from the scheme. This effect is usually not present in a developed country context. Therefore, the impact of transit subsidy policies in the global south may be very different from those in developed countries and policy design must take this into account.

We also present evidence that there may be a compensating equilibrium wage adjustment in the formal labor market caused by the subsidy, further reducing its targeting properties. Our evidence is only suggestive, and more research is warranted on the potential labor market equilibrium effects of the Vale Transporte scheme.

Is there a better way to help low-income workers in paying their commuting costs? Transit subsidies are usually motivated by the need to address affordability problems for the poor who depend on public transport for their daily activities. However, monetary transfers could be a better way to help this group than sectoral subsidies. As argued by Gwilliam (2017), low incomes may be the root cause of transport affordability problems faced by the poor rather than high transit fares.

Nevertheless, there may be special reasons to subsidize transit directly. For example, to promote the use of public transport when private transport externalities are not priced or to reflect the economies of scale present in transit systems (Mohring effect). Also, in some countries the welfare system may be underdeveloped, hindering the administration of monetary transfers. Sectoral subsidies may also be a better way to distribute resources to certain groups that, due to intrahousehold resource allocation problems, may be difficult to target using monetary transfers (e.g. school-age students). For these and other reasons, monetary transfers to the poor may not always be a good policy substitute for transit subsidies.

If this is the case in Brazil, how then could commuting affordability be improved for lower income workers in Brazil? One alternative is to consider means-tested demand side subsidies such as the one introduced in Bogotá, Colombia (see Gómez-Lobo, et al. 2022, for details of this experience). The advantage of these type of subsidies is that they could also benefit informal sector workers, substantially improving the targeting properties of benefits.

Author Contribution

Andrés Gómez-Lobo supervised the research, wrote the literature review and developed the methodological approach. Vileydy González and Santiago Sanchez obtained the data and undertook all the empirical estimations. All three authors contributed to the writing of the original manuscript and the review of subsequent versiones of the manuscript.

Acknowledgement

We thank comments and suggestions provided by Tomas Serebrisky, Agustina Calatayud and an anonymous reviewer from the Inter-American Development Bank.

Asensio, J., A. Matas and J.L. Raymond (2003), ‘Redistributive effects of subsidies to urban public transport in Spain’, Transport Reviews, 23:4, 433-452, https://doi.org/10.1080/01441647.2002.10823176.
Bhattacharya, S. and M. Cropper (2007), ‘Public Transport Subsidies and Affordability in Mumbai, India, Policy Research Working Paper 4395, The World Bank.
Börjesson, M., J. Eliasson and I. Rubensson (2020), ‘Distributional effects of public transport subsidies’, Journal of Transport Geography, 102674, https://doi.org/10.1016/j.jtrangeo.2020.102674.
Bondemark, A., H. Andersson, A. Wretstrand, and K. Brundell‑Freij (2021), ‘Is it expensive to be poor? Public transport in Sweden’, Transportation, 48:2709–2734. https://doi.org/10.1007/s11116-020-10145-5.
Brough, R., M. Freedman and D.C. Phillips (2022), ‘Experimental evidence on the effects of means-tested public transportation subsidies on travel behavior’, Regional Science and Urban Economics, 96, 103803, https://doi.org/10.1016/j.regsciurbeco.2022.103803.
Bueno Cadena, P.C., J. M. Vassallo, I. Herraiz and M. Loro (2016), ‘Social and Distributional Effects of Public Transport Fares and Subsidy Policies: Case of Madrid, Spain’, Transportation Research Record: Journal of the Transportation Research Board, No. 2544, Transportation Research Board, Washington, D.C., 2016, pp. 47–54. DOI: 10.3141/2544-06.
Bull, O., J.C. Muñoz, and H.E. Silva (2021), ‘The impact of fare-free public transport on travel behavior: Evidence from a randomized controlled trial’, Regional Science and Urban Economics, 86, 103616. https://doi.org/10.1016/j.regsciurbeco.2020.103616.
Cervero, R. and R. Golub (2007), ‘Informal transport: A global perspective’, Transport Policy, 14, pp. 445-457.
Chen, H., W. Du, N. Li, G. Chen and X. Zheng (2013), ‘The socioeconomic inequality in traffic-related disability among Chinese adults: the application of concentration index’, Accid. Anal. Prev., 55, 101–106. https://doi.org/10.1016/j.aap. 2013.02.025.
Coady, D., M. Grosh and J. Hoddinott. (2004), ‘The Targeting of Transfers in Developing Countries: Review of Experience and Lessons’, Social Protection Discussion Paper, Washington, DC: World Bank.
Eliasson, J., R. Pyddoke, J.-E. Swärdh (2018), ‘Distributional effects of taxes on car fuel, use, ownership and purchases’, Economics of Transportation, https://doi.org/10.1016/j.ecotra.2018.03.001.
Erreygers, G. (2009), ‘Correcting the concentration index’, Journal of Health Economics, 28, pp. 504–515. https://doi.org/10.1016/j.jheal eco.2008.02.003.
Flynn, J. (2007), ‘Measures to make Urban Transport Affordable to the Poor: Mexico City Case Study’, John F. Kennedy School of Government, Harvard University, Report prepared for The World Bank, February.
Franklin, S. (2017), ‘Location, search costs and youth unemployment: experimental evidence from transport subsidies’, The Economic Journal, 128(614), pp. 2353-2379.
Golub, A., R. Balassiano, A. Araújo and E. Ferreira (2009), ‘Regulation of the informal transport sector in Rio de Janeiro, Brazil: welfare impacts and policy analysis’, Transportation, 36, pp. 601–616.
Gómez-Lobo, A., S. Sánchez González and M. González Mejia (2022), ‘Means-tested transit subsidies in Latin America’, IDB Working Paper Series IDB-WP-01365, Interamerican Development Bank, November.
Gómez-Lobo, A. (2009), ‘A New Look at the Incidence of Public Transport Subsidies: a case study of Santiago, Chile’, Journal of Transport Economics and Policy, vol. 43(3), pp. 405-425.
Gómez-Lobo, A., T. Serebrisky, B. Solís, H. Cárdenas, M. Gutiérrez and S. Huamaní (2023), ‘Reducing Leakage: Subsidies and Tariff Reform in Water and Sanitation Services in Metropolitan Lima, Peru’, Water Economics and Policy, 2350004 (37 pages).
Guzman, L.A., S. Gómez Cardona and J.L. Ochoa (2023), ‘The effect of public transport subsidies on travel behavior and welfare: a large-scale randomized controlled field experiment in a developing context’, available at SSRN: https://ssrn.com/abstract=4330371 or http://dx.doi.org/10.2139/ssrn.4330371.
Guzman, L.A. and P. Hessel (2022), ‘The effects of public transport subsidies for lower-income users on public transport use: A quasi-experimental study’, Transport Policy, 126, pp. 215-224.
Gwilliam, K. (2017), Transport Pricing and Accessibility, Moving to Access Initiative, The Brookings Institution.
Hall, P., A. Perl, K. Sawatzky and S. Tornes (2021), ‘Employer-paid transit subsidies and travel behaviour: Experimental evidence from Vancouver hotels’, Journal of Urban Mobility, 1, pp. 1-8.
Herszenhut, D., R. H.M. Pereira, L. da Silva Portugal and M. Henrique de Sousa Oliveira (2022), ‘The impact of transit monetary costs on transport inequality’, Journal of Transport Geography, 99, 103309. https://doi.org/10.1016/j.jtrangeo.2022.103309.
Kakwani, N.C. (1977), ‘Applications of Lorenz curves in economic analysis’, Econometrica, 45, pp. 719–727. https://doi.org/10.2307/1911684.
Karner, A., R.H.M. Pereira and S. Farber (2024), ‘Advances and pitfalls in measuring transportation equity’, Transportation. https://doi.org/10.1007/s11116-023-10460-7.
Komives, K., V. Foster, J. Halpern y Q. Wodon (2005), ‘Agua, electricidad y pobreza. ¿Quién se beneficia de los subsidios a los servicios públicos?’, Banco Mundial.
Moreno-Monroy, A.I. and H.M. Posada (2018), ‘The effect of commuting costs and transport subsidies on informality rates’, Journal of Development Economics, 130, pp.99-112.
NTU (2023), Anuário NTU: 2022-2023, Associação Nacional das Empresas de Transportes Urbanos. – Brasília.
Parry, I.W.H and K.A. Small (2009), ‘Should Transit Subsidies be Reduced?’, American Economic Review, 99(3), pp. 700-724.
Porque, A.L., A.A. Betarelli Junior and F. Salgueiro Perobelli (2022), ‘Fuel tax, cross subsidy and transport: Assessing the effects on income and consumption distribution in Brazil’, Research in Transportation Economics, 95, 101204. https://doi.org/10.1016/j.retrec.2022.101204.
Porque, A.L., A.A. Betarelli Junior, D. Domingos Ribeiro and G. Filgueiras Di Filippo (2023), ‘Redistributive effects of consumption and income from subsidies to passenger transportation in the Brazilian economy’, paper presented at the 29th International Input-Output Association, Alghero, Italy, June 26-30.
Rivas, M.E., J.P. Brichetti and T. Serebrisky (2020), Operating subsidies in urban public transit in Latin America: a quick view, IDB Monograph 786, InterAmerican Development Bank.
Rubensson, I., Y. Susilo and O. Cats (2020), ‘Fair accessibility–operationalizing the distributional effects of policy interventions’, Journal of Transport Geography, 89, 102890. https://doi.org/10.1016/j.jtrangeo.2020.102890.
Serebrisky, T., A. Gómez‐Lobo, N. Estupiñán, and R. Muñoz‐Raskin. 2009. “Affordability and Subsidies in Public Urban Transport: What Do We Mean, What Can Be Done?” Transport Reviews 29(6): 715–39.
Silver, K., A. Lopes, D. Vale and N. Marques da Costa (2023), ‘The inequality effects of public transport fare: The case of Lisbon's fare reform’, Journal of Transport Geography, 112, 103685. https://doi.org/10.1016/j.jtrangeo.2023.103685.
Suits, D.B. (1977), ‘Measurement of tax progressivity’, American Economic Review, 67, pp. 747–752.
Vasconcellos, E.A. (2018), ‘Urban transport policies in Brazil: The creation of a discriminatory mobility system’, Journal of Transport Geography, 67, pp. 85-91.
Vecchio, G., I. Tiznado‑Aitken, B. Castillo and S. Steiniger (2024), ‘Fair transport policies for older people: accessibility and affordability of public transport in Santiago, Chile’, Transportation, 51, pp. 689–715. https://doi.org/10.1007/s11116-022-10346-0.
West, J. and M. Börjesson (2020), ‘The Gothenburg congestion charges: cost-benefit analysis and distribution effects’, Transportation, 47, pp. 145–174. https://doi.org/10.1007/s11116-017-9853-4.

Karner, et al. (2024) present an up-to-date review of the concepts, methods and applications in the literature with an applied example for Washington DC.
Although this benefit is formally paid by employers, the vouchers are tax deductible as discussed further below. Thus, a fraction of the cost is funded by the public sector in the form of lower tax revenues.
This is a hypothecated payroll tax paid by firms with 11 or more workers. The proceeds are used to fund capital and operating expenditure of the public transport systems. It was originally introduced in the Paris region but later extended to all major urban areas.
Interested reader may find these older references in the papers reviewed here.
In the case of Sweden, a related paper is Bondemark, et al. (2021) who show that the probability of purchasing a travel card increases with income. Thus, liquidity constrained low-income individuals often pay a higher fare per trip compared to higher income individuals who buy a travel card. However, they do not estimate how this changes the overall distributive incidence of transit subsidies.
Karner, et al. (2024) report other transport related uses of the Concentration Curve index (either the Kakwani (1977) or the Suits (1977) approach discussed further below). These include Chen et al. (2013), for inequality in crash-related disabilities and Rubensson et al. (2020), who analyzed the socioeconomic inequality of accessibility and the progressivity of a skip-stop train policy introduced in Stockholm, Sweden. To these we can add West and Börjesson (2004) who estimate the distributive impact of the Gottenburg congestion charge and Eliasson, et al. (2018) for the incidence of fuel prices.
To be precise, this is what Kakwani (1977) calls the Lorenz curve and is a particular case of his more general concentration curve approach. The Concentration Curve index is analogous to the Gini coefficient.
Two Lorenz curves that cross can have the same CI value but very different targeting properties.
It can also be interpreted as the average subsidy received by a targeted individual relative to the overall average subsidy across all individuals.
More generally, all the information regarding the distribution of benefits is contained in the Lorenz curve. For example, defining S(p) as the share of total subsidies received by the first p centiles of the income distribution, then the Palma Index is (1-S(90))/S(40) and can calculated from the slope of the ray to each of these points in the Lorenz curve.
This last type of worker could opt to receive Vale Transporte vouchers if they so wish. However, this is unlikely since the price of the vouchers sold to employers is somewhat higher than the price individual passengers would have to pay (see further below on how employers purchase vouchers). Therefore, the wage deduction would be larger than their commuting expenditure without Vale Transporte.
The income after payment of social security contributions but before income taxes.
Further below we discuss the implications if this assumption does not hold, and gross labor earnings are registered before the effects of the Vale Transporte scheme are considered.
The subsidy value in dollars was calculated using the average exchange rate of the first quarter of 2020 (IMF, IFS Statistics, 2022).
We compute this estimate with the teffects psmatch command in Stata.

No competing interests reported.

Download PDF

Reviews received at journal
03 Sep, 2024
Reviewers agreed at journal
31 Jul, 2024
Reviewers invited by journal
29 Jul, 2024
Editor assigned by journal
01 Jul, 2024
Submission checks completed at journal
21 May, 2024
First submitted to journal
20 May, 2024

You are reading this latest preprint version

The Targeting Incidence of Brazil’s Vale Transporte transit subsidy scheme

Status:

Version 1

Abstract

Figures

1. Introduction

2. Literature review

3. Description of the Vale Transporte

4. Data and subsidy imputation

4.1. Survey data

4.2. Imputing the subsidy benefit

5. Distributive incidence results

6. Possible labor market adjustments

7. Conclusions

Declarations

Author Contribution

Acknowledgement

References

Footnotes

Additional Declarations

Status:

Version 1