Ricardo Carvalho de Andrade Lima ORCID logo

The Economic Impact of State Splitting in Brazil

De Gruyter | Published online: October 16, 2020

Abstract

The Brazilian state of Tocantins was splinted from the state of Goiás in 1988. This was the most recent first-tier subnational border reform in Brazil and involved an area that corresponds to 7.6% of the whole national territory. Using the synthetic control method, this paper estimates that the split increased the per capita GDP of the affected region by an average of 8.26%, and the positive effect persisted over the years. Additionally, we show that the Tocantins benefited more from the border reform than Goiás, and the increase of fiscal capacity of affected subnational governments may partially explain the positive effects of splitting.

1 Introduction

Interest groups composed of politicians and local elites often aspire greater political and fiscal autonomy in federal systems and, for this reason, claim for border reforms that make it possible to create new first-tier subnational units. In Brazil, a federation comprising 26 states plus a federal district, attempts to split states are common. After the re-democratization and the federative pact of 1988, there were 11 attempts to create new states (Shikida, Faria, and Araújo Jr 2014). However, only the creation of the state of Tocantins, in the Brazilian Midwest, was successful.

The economic effects arising from border reforms and the resulting decentralization are not clearly established. Theoretically, the creation of new jurisdictions can potentially increase the well-being of the local population by reducing the heterogeneities regarding preferences for different levels of public goods (Oates 1999). On the other hand, reforms that generate smaller jurisdictions tend to worsen welfare by reducing government ability to achieve economies of scale in the provision of public goods. This is the essential trade-off in the determination of the optimal size of subnational units (Alesina, Baqir, and Hoxby 2004).

Although there is a wide set of quasi-experiment studies investigating the effects of border reforms in local governments,[1] empirical evidence of first-tier sub-national reforms is scarce. Asher and Novosad (2015) and Shenoy (2018) analyze the economic consequences of the creation of new states in India and Grossman, Pierskalla, and Boswell Dean (2017) and Baskaran and Blesse (2019) investigate the split of first-tier subnational units in African continent. Dalmazzo, de Blasio, and Poy (2018) developed a theoretical model in which homophily driver the secession claim and, to check the model predictions, the authors empirically evaluated the effects of Molise’s split from the Italian region of Abruzzo in 1963. In general, these papers suggest that split first-tier subnational governments promote economic progress and improve public goods supply. The increase in governments’ fiscal capacity, reduced ethnic conflicts, better local social cohesion, and greater political stability are recognized mechanisms that explain the association between state splitting and economic development.

We intend to contribute to this empirical literature by estimating the economic impact of the creation of the Brazilian state of Tocantins, splintered from the state of Goiás in 1988. It is the only first-level subnational unit created in recent Brazilian history, but to the best of our knowledge, its economic effects have never been evaluated. In addition to investigating an unprecedented case study, the paper also sheds light on how the changes in the fiscal capacity of a splintered region can partially explain the economic consequences of a boundary reform.

Applying the synthetic control method Abadie, Diamond, and Hainmueller (2015) and using state-level aggregated data, we show that the creation of the state of Tocantins increased the per capita GDP in the affected region by an average of 8.26%. Using a Differences-in-Differences estimator and disaggregated data, we evidenced that this positive effect was more substantial in the new state. On average, the municipalities in the Tocantins grew twice in comparison to the municipalities located in Goiás after the splitting. Finally, we also present some insights suggesting that the increase in fiscal capacity of the affected sub-national governments can partially explain the positive effects of the border reform.

2 Institutional Background

The Brazilian territory is a federation composed of three self-governing tiers: the central government, 27 state governments plus a federal district, and 5570 municipalities. The country is highly decentralized, with around 47.9% of public spending concentrated on the sub-national governments (BID 2018). States have the autonomy to collect taxes, have their legislation, and enjoy political and administrative independence. This territorial organization was established in 1988 with the creation of a new constitution and the federative pact.

The secession claims of the state of Tocantins were originated since the nineteenth century, and the most formal one comes in 1956, with “Movimento Pró-Criaҫão do Estado de Tocantins” (Gonçalves 2015). These movements were discouraged during the military dictatorship (1964–1985), characterized by strong fiscal and administrative centralization. With the re-democratization and the emergence of the national constituent assembly in 1987–1988, the secession of Tocantins returned to the public agenda, since the policymakers aspire to a more decentralized country. The secession project was unanimity accepted by the National Assembly, and the state effectively emerged with the 1988 constitution, after the split of the northern part of Goiás. This split was contained in a broader territorial reform, which elevated Amapá and Roraima’s federal districts to states’ category and granted political autonomy to the country’s single Federal District (Brasília).

Figure 1 shows the state borders of Brazil before and after the creation of Tocantins. The region affected by the border reform has a broad geographical dimension (7.6% of whole Brazilian territory) and a low population density. Unlike the regions investigated in previous studies, it did not experience any ethnic conflict or political destabilization.

Figure 1: State borders of Brazil before and after the creation of Tocantins.

Figure 1:

State borders of Brazil before and after the creation of Tocantins.

Documentary evidence indicates that the secession was mostly motivated by the under-supply of public goods from the state government to the population located in the former state’s north area. The capital city of Former Goiás was installed in the south part of the territory at about 800 km to the most populous area of Tocantins. In addition to hampering the supply of public goods to the whole region, this great distance also reduced interaction between the northern population and former state government, reducing political efficacy.

During the 1988 Brazilian national constituent assembly, there were five other proposals[2] for the secession of first-tier subnational units rejected by the constituent assembly (Martins 2001). Some failed attempts come from economically dynamic regions experiencing strong growth at the time, which anecdotally suggests that previous positive economic shocks cannot merely explain the secession success. As highlighted by Martins (2001), the process of creating new states in Brazil is a complicated question, involving issues of construction of territorial identities to political and electoral articulations.

3 Methodology and Data

3.1 The SCM for State-Level Aggregated Data

Considering the objective of evaluating the economic impact of the splitting of the state of Tocantins, we need to create an adequate counterfactual for the region affected by the border reform. This paper employs a data-driven approach to construct a synthetic state based on the weighted average of all other Brazilian states that have kept their borders intact.[3] Initially, it is assumed that there is a total of J + 1 states indexed by j where j = 1 is the treated state and j = 2 to J + 1 are the potential control units or the donor pool.

A challenge for studies investigating the effects of border reforms is defining a treated unit that is constant over time. As the state of Tocantins was only established in 1988, we do not have data for pre-intervention periods. We address this issue by defining as the treated unit the sum of the territory of Tocantins and Goiás (i.e., the former Goiás, Panel (a) of Figure 1). This definition makes the geographical size of the region time-constant and allows us to obtain consistent data both in pre-intervention period (data from the former Goiás) and in the post-intervention period (aggregation of data from new Goiás and Tocantins).

The synthetic control method (SCM) is based on the idea that the pre-intervention characteristics of the treated unit can be better approximated by a combination of untreated units than by any single untreated unit. The synthetic control is defined as the weighted average of all states in the donor pool. Formally, the synthetic control can be represented by a vector J × 1 of weights given by W = (W2,…, Wj + 1)′ with 0 w j 1 for j = 2,…,J + 1 and W2 + … + Wj + 1 = 1. Let X 1 be a kx1 vector containing the values of the characteristics of the treated state in the pre-intervention period and X0 be a kxj matrix containing the same characteristics for states of the donor pool. Additionally, let V be a symmetric kxk matrix that allows different weights to the variables included in X0 and X1 depending on their predictive power on the outcome. Abadie, Diamond, and Hainmueller (2015) propose to choose a vector W* that minimizes the root mean square prediction error (RMSPE) given by the following expression:

(1) X 1 X 0 V = ( X 1 X 0 W ) V ( X 1 X 0 W )
Let Y jt be the outcome of state j in period t and let T 0 be the year of intervention. After obtaining the vector W *, the synthetic control estimator of the effect of treatment for a specific post-intervention period, t T 0 , is given by the following expression:
(2) Y 1 t = j = 2 j + 1 w j * Y j t

Our outcome variable is the per capita GDP (deflated to R$ of 2000) at the state level. As control variables, we include the following traditional determinants of regional growth: population, the share of agricultural sector on GDP, the share of the manufacturing sector on GDP, the share of the public sector on GDP, average years of schooling of people over 25 years old (a proxy for human capital) and industrial consumption of electric energy (a proxy for physical capital). All variables were obtained through IPEADATA, a public database collected by the Institute of Applied Economic Research. Our database comprises the period[4] between 1970 and 2000. Only the average years of schooling began to be collected in 1985.

3.2 The Differences-in-Differences for Municipality Data

The two states affected by the border reform may have responded very differently to the decentralization shock. The northern part of the state, splintered and became a new first-tier subnational unit (Tocantins), tends to be more impacted since it has gained a new bureaucratic structure and greater fiscal and administrative autonomy. On the other hand, the state of Goiás did not experience any change in federative status with the reform, but left behind the administration of 45% of its former territory and became more compact. To evaluate the heterogeneity of the effects, it is not possible to adopt an SCM approach, because we do not have information on the state of Tocantins before the split.

To address this issue, we use more disaggregated data to track and reconstruct borders as if they had existed in the past. More specifically, we use municipalities[5] as the observational unit and classified the municipalities of the Tocantins state as if they belonged to the state since 1970. Using this procedure, it is xpossible to analyze how these municipalities behaved after the split. To estimate the heterogeneous effect of the reform, we adopted the following Differences-in-Differences (DiD) specification:

(3) y i t = β 1 ( Tocantins i . Split t ) + β 2 ( Goias i . Split t ) + X i t α + γ i + θ t + ε i t
where, y it is the outcome of municipality i at year t Tocantins i and Goias i are dummies variables that take the value of 1 for municipalities located at the respective state and assume 0, otherwise. Split t is a dummy variable that takes the value of 1 for the periods after the border reform and 0, otherwise. X i t is a vector of time-varying controls, γ i is the municipality fixed-effect and θ t is the time-effect. Coefficient β 1 measures the effect of the border reform on the economic performance of municipalities located in the Tocantins, and the coefficient β 2 measures the effect of the reform on the performance of municipalities located in Goiás. These are the DiD estimates that make it possible to check the heterogeneity of the effects.

In this analysis, we will use data from the Brazilian Demographic Census of 1970, 1980, 1991, and 2000. Thus, we have a balanced panel for 3659 municipalities. Our outcome variable is the per capita GDP at the municipality level. As control variables, we will use the following: population, the share of the agricultural sector on GDP, average years of schooling, and the stock of residential capital.[6]

4 Results

4.1 The Effects of State Splitting

Initially, our donor pool is formed by all Brazilian states, excluding the states of Mato Grosso and Mato Grosso do Sul because they had a border reform in 1979. After minimizing[7] expression (1), our synthetic control was constructed by a combination of the following states: Maranhão (51.75%), Espiríto Santo (27.03%), Paraná (19.44%) and Federal District (1.76%). Table 1 shows the mean of the pre-treatment characteristics for the treated unit (former Goiás), for the synthetic control and the complete set of the donor pool.

Table 1:

Mean of pre-splitting predictors, 1970–1987.

Predictors Treated Synthetic Donor pool
Population (per million) 3.894 3.950 4.906
Log of industrial consumption of energy (MWh) 12.747 13.399 12.703
% Of agriculture on GDP 0.261 0.269 0.177
% Of manufacture on GDP 0.196 0.224 0.300
% Of public sector on GDP 0.094 0.105 0.141
Per capita GDP (per thousand) 9.382 9.368 12.303
Average years of schooling 3.996 3.189 4.227

    Population, industrial consumption of energy, agriculture share, manufacture share, public sector share and are averaged for 1970–1987 period. Average years of schooling are averaged for 1985–1987. Per capita GDP is averaged for 1986–1987 period and is deflated to 2000 Brazilian reais.

Figure 2 shows the trajectory of per capita GDP for the treated unit and the synthetic control in the period between 1970 and 2000, and Figure 3 shows the corresponding gap. It is observed that before the border reform, the trajectories are incredibly similar, and the gap is close to zero. One year after the creation of the state of Tocantins, the per capita GDP of the affected region starts to diverge from the synthetic control. The divergence remained over the years, but at a decreasing rate. This evidence suggests that the border reform has benefited the affected state. More specifically, the average effect of treatment in the period from 1988 to 2000 was 8.26%. In 2000, the per capita GDP of the splintered region was about 4.9% higher than it would have been if the region had not splinted.

Figure 2: Trajectory of per capita GDP: treated region versus synthetic region.The figure plots the trajectory of per capita GDP of the Former Goiás (black) and the corresponding synthetic control (blue). The vertical line indicates the year of the boundary reform (1988). The Synthetic Control was constructed by a combination of the following states: Maranhão (0.517), Espírito Santo (0.27), Paraná (0.194) and Federal District (0.017).

Figure 2:

Trajectory of per capita GDP: treated region versus synthetic region.

The figure plots the trajectory of per capita GDP of the Former Goiás (black) and the corresponding synthetic control (blue). The vertical line indicates the year of the boundary reform (1988). The Synthetic Control was constructed by a combination of the following states: Maranhão (0.517), Espírito Santo (0.27), Paraná (0.194) and Federal District (0.017).

Figure 3: Gap in per capita GDP: treated region versus synthetic region.The figure plots the gap of per capita GDP between the Former Goiás and the corresponding synthetic control. The vertical line indicates the year of the boundary reform (1988).

Figure 3:

Gap in per capita GDP: treated region versus synthetic region.

The figure plots the gap of per capita GDP between the Former Goiás and the corresponding synthetic control. The vertical line indicates the year of the boundary reform (1988).

To check the validity of our results, we followed Abadie, Diamond, and Hainmueller (2015) and conducted placebo studies by artificially reassigning the border reform to states that were not affected by the intervention. This created a distribution of placebo effects that can be compared with our effectively treated unit. Figure 4 displays the in-space placebos tests. The black line represents the gap in the GDP between the treated unit and the synthetic control, and the blue lines represent the corresponding gap for artificially treated units. It is possible to observe that the gap of the former Goiás is unusually larger to the distribution of placebos. Only one state of the donor pool had a more prominent effect, but it only manifested in the mid-1990s.

Figure 4: Outcome gaps in the treated state and placebo gaps in control states.The black line shows the gap of per capita GDP between the Former Goiás and the synthetic control. The blue lines show the correspondent estimated placebo effect for each control state with an RMSPE less than 10 times the RMSPE of treated region.

Figure 4:

Outcome gaps in the treated state and placebo gaps in control states.

The black line shows the gap of per capita GDP between the Former Goiás and the synthetic control. The blue lines show the correspondent estimated placebo effect for each control state with an RMSPE less than 10 times the RMSPE of treated region.

Figure 5 shows the ratios between the post-treatment RMSPE[8] and the pre-treatment RMSPE for each subnational unit. An RMSPE ratio is high when the post-treatment effect is high, and, at the same time, the pre-treatment outcome trajectory fits well with the synthetic counterpart. As can be seen in Figure 5, the treated state has the highest RMSPE ratio. Therefore, if the treatment is artificially reassigned to any state in the donor pool, the probability of finding an RMSPE as high as this would be equivalent to 1 / 24 0.04166 .

Figure 5: Ratio of pos-splitting RMSPE to pre-splitting RMSPE.The figure shows the ratios between the post-treatment RMSPE and the pre-treatment RMSPE for each subnational unit. An RMSPE ratio is high when the post-treatment effect is high, and, at the same time, the pre-treatment outcome trajectory fits well with the synthetic counterpart.

Figure 5:

Ratio of pos-splitting RMSPE to pre-splitting RMSPE.

The figure shows the ratios between the post-treatment RMSPE and the pre-treatment RMSPE for each subnational unit. An RMSPE ratio is high when the post-treatment effect is high, and, at the same time, the pre-treatment outcome trajectory fits well with the synthetic counterpart.

We also check if our SC estimative depends on any specific factors associated with the control group. Firstly, Figure 6A shows the SC trajectory, excluding the main donor pool (the State of Maranhão, 53% of the baseline SC). Figure 6B shows the SC excluding the capital of Brazil (Federal District) and states that had federal districts’ status before the 1988 constitution. Figure 6C shows the SC results when we restricted the donor pool to states that have experienced some secession attempt but have failed. We expect that this group of states have similar (and unobserved) characteristics to the Former Goiás that could influence the split movement. By limiting the control group to states that tried to split but failed, we reduce our concerns about confounders that can potentially threaten to identify a causal nexus. Finally, Figure 6D presents the results of the SC eliminating states that are immediate neighbors of the Former Goiás. This avoids using control units that have suffered spatial spillover from the border reform. In general, the estimates in Figure 6 are very similar to those obtained in Figure 2, suggesting that our baseline results are robust.

Figure 6: Robustness checks: limiting the donor pool.The figure plots the trajectory of per capita GDP of the Former Goiás (black) and the corresponding synthetic control (blue). Figure 6A shows the SC trajectory excluding the main donor pool, and the weights are: Piauí (0.26), Espiríto Santo (0.11), Paraná (0.233), Federal District (0.02) and Pará (0.375). Figure 6B shows the SC excluding actual or past federal districts, and the weights are: Piauí (0.612) and Santa Catarina (0.387). Figure 6C shows the SC restricting the donor pool to only the states that claimed some secession attempt but failed, and the weights are: Piauí (0.264), Paraná (0.164), Pará (0.381), Bahia (0.09), and Santa Catarina (0.18). Figure 6D presents the results of the SC eliminating immediate neighbors of the Former Goiás, and the weights are: Paraíba (0.19), Paraná (0.42) and Acre (0.386). The vertical lines indicate the year of the boundary reform (1988).

Figure 6:

Robustness checks: limiting the donor pool.

The figure plots the trajectory of per capita GDP of the Former Goiás (black) and the corresponding synthetic control (blue). Figure 6A shows the SC trajectory excluding the main donor pool, and the weights are: Piauí (0.26), Espiríto Santo (0.11), Paraná (0.233), Federal District (0.02) and Pará (0.375). Figure 6B shows the SC excluding actual or past federal districts, and the weights are: Piauí (0.612) and Santa Catarina (0.387). Figure 6C shows the SC restricting the donor pool to only the states that claimed some secession attempt but failed, and the weights are: Piauí (0.264), Paraná (0.164), Pará (0.381), Bahia (0.09), and Santa Catarina (0.18). Figure 6D presents the results of the SC eliminating immediate neighbors of the Former Goiás, and the weights are: Paraíba (0.19), Paraná (0.42) and Acre (0.386). The vertical lines indicate the year of the boundary reform (1988).

4.2 The Heterogeneous Effects of State Splitting

Table 2 presents the results obtained by estimating the DiD estimator. Column (1) and Column (2) show the specifications for the model without distinction between two states (equivalent to SCM) and Column (3) and Column (4) show the specifications considering state heterogeneities.

Table 2:

The heterogeneous effect of state splitting.

(1) (2) (3) (4)
Former Goiás * split 0.1107***

(0.018)
0.0769***

(0.015)
Tocantins * split 0.1137***

(0.020)
0.1185***

(0.036)
Goiás * split 0.0968**

(0.047)
0.0682***

(0.016)
Share of agriculture 0.4362***

(0.030)
0.4375***

(0.030)
Log (population) −0.6625***

(0.014)
−0.6632***

(0.014)
Log (residential capital) 0.6758***

(0.012)
0.6761***

(0.012)
Log (av. years of schooling) −0.0321

(0.025)
−0.0329

(0.025)
F-test for β1β2 = 0 0.186 2.885*
Municipality FE Yes Yes Yes Yes
Time FE Yes Yes Yes Yes
Observations 14636 14636 14636 14636

    ***p < 0.01, **p < 0.05, *p < 0.1. Robust standard errors clustered at municipality level are in parentheses. The outcome variable is the log of per capita GDP. The data were obtained at municipality level with time constant borders at four points in time: 1970, 1980, 1991 and 2000.

Initially, the figures presented in column (1) and (2) reinforce our previous estimates. Municipalities located in the region affected by the boundary reform grew an average of 7.8% more than the other Brazilian municipalities. However, column (4) shows that the municipalities located in the new state (Tocantins) grew at a rate almost twice higher (11.85% against 6.82%) than the municipalities located in Goiás. We also performed an F-test to check if β1β2 = 0. The F-test confirms that the DiD coefficients for the different states are statistically different. This evidence is consistent with the fact that the border reform provided more intense fiscal decentralization for the created state.

The key identifying assumption for the DiD is that the outcome trends between the municipalities affected by the reform (treated group) and the municipalities not affected (control group) would be the same in the absence of the border reform. Although it is impossible to evaluate this assumption, we can get an idea of its validity by estimating an event-study DiD design in the spirit of Autor (2003) and investigating whether the anticipatory effects are statistically significant.[9]Table 3 presents the results of our event-study estimation.

Table 3:

Leads and lags specification.

(1) (2)
Tocantins * split (t−2) −0.0033

(0.038)
Goiás * split (t−2) −0.0180

(0.020)
Tocantins * split (t−1) 0.0033

(0.038)
Goiás * split (t−1) 0.0180

(0.020)
Tocantins * split (t) 0.1502***

(0.037)
0.1470***

(0.047)
Goiás * split (t) 0.0772***

(0.020)
0.0592***

(0.021)
Tocantins * split (t+1) 0.0901**

(0.038)
0.0868*

(0.051)
Goiás * split (t+1) 0.0773***

(0.025)
0.0593**

(0.023)
F-test for β1tβ2t = 0 3.171* 2.953*
Municipality FE Yes Yes
Time FE Yes Yes
Controls Yes Yes
Observations 14636 14636

    ***p < 0.01, **p < 0.05, *p < 0.1. Robust standard errors clustered at municipality level are in parentheses. The outcome variable is the log of per capita GDP. The event-study specification is constructed by separating the treatment variable for each specific year. For example, the dummy t−1 for Tocantins assumes one for all municipalities located in the state in the decade immediately prior to the territorial reform. Column (1) shows the specification considering dummy t−1 as the base category and column (2) shows the specification considering dummy t−2 as the base category.

We note that in both specifications, the anticipatory effects are not statically significant. Therefore, in the decades before the boundary reform, there was no difference in the growth trajectory between the two groups of municipalities.

4.3 Insights about the Mechanisms

Considering that our previous results show that the split of Tocantins increased the economic performance of the affected region, it is essential to investigate the mechanism behind this. As the primary motivation for the border reform was the government’s inability to provide an appropriate level of public goods for the entire territory, the increase in the fiscal capacity of the new sub-national governments may partially drive our results.

To check this mechanism, we compared the difference in fiscal capacity outcomes between the affected state (sum of Goiás and Tocantins) and the non-affected states, before and after the border reform. Figure 7 shows this difference for the following outcomes: intergovernmental transfers, total revenues, capital expenditures, and current expenditures. All variables are in per capita terms. The control group used for comparison was formed by the national average (black) or the SCM weighted average (blue). This type of comparison is similar to a graphical Differences-in-Differences analysis[10] (see for example, Barone and Mocetti (2014)).

Figure 7: The effects of state spliting on the fiscal capacity of subnational governments.1986 is used as the pre-treatment period and 1990 is used as the post-treatment period. Public finance data was collected by the National Treasury Secretariat. Each bar shows the difference in outcome between former Goiás and the average of Brazilian states. The values are in per capita terms and are deflated to 2000 Brazilian reais.

Figure 7:

The effects of state spliting on the fiscal capacity of subnational governments.

1986 is used as the pre-treatment period and 1990 is used as the post-treatment period. Public finance data was collected by the National Treasury Secretariat. Each bar shows the difference in outcome between former Goiás and the average of Brazilian states. The values are in per capita terms and are deflated to 2000 Brazilian reais.

Initially, we note that for all outcomes, the pre-splitting difference between the Former Goiás and the SC average is smaller than the correspondent difference using the national average, indicating that the SC group is more reliable for comparisons. From Figure 7A, B, it is possible to observe that before the split, the fiscal capacity of the former Goiás was lower than the synthetic counterpart (negative difference). After the reform, this pattern was completely reversed. The affected states began to receive a higher volume of federal transfers and increased their public budget compared to the control group. Thus, this improvement in the affected state’s fiscal capacity may partially explain the economic consequence of the boundary reform. Previous studies show that intergovernmental transfers have positive effects on the performance of local economies in Brazil (Corbi, Papaioannou, and Surico 2019).

Regarding public spending (Figure 7C, D), we note a similar pattern of change. Both types of spending increase significantly after the split, although capital spending has increased more intensely. This result is quite expected. Immediately after Tocantins’ creation, 44 new municipalities were installed in the territory (an increase of 55%), including the city of Palmas, the current state capital (Gonçalves 2015). The installation of new cities involves expenditures on buildings, transportation infrastructure, and machinery. These investments are directly associated with capital spending. The mobilization of resources for installing new cities also helps to explain the short-term effect of the boundary reform.

5 Conclusions

In this paper, we evaluated the economic effect of the split of Tocantins from the state of Goiás in 1988, the most recent first-tier subnational reform in Brazil. Our case study is unique due to the particularities of the affected region. At the time, the Tocantins was an economically backward region and occupied a large territory with a small population. Additionally, due to the historical underserve of public goods, the region claimed secession from Goiás since the nineteenth century.

Our results show that the state splitting increased the affected region’s per capita GDP by an average of 8.26%. The empirical findings also point that the region’s increased fiscal capacity may explain part of the economic effect of the reform. After the split, the volume of per capita intergovernmental transfers designated to the affected jurisdiction increased by 66.93%. Together with the effort to create new cities, this expressive volume of resources can explain both the short-term effect of the split and the economically large magnitude.

Although we recognize that our case-study has limited external validity, the Tocantins creation’s story indicates that institutional reforms that allow greater government coverage in historically underserved areas can promote long-term economic gains. This type of evidence sheds light on the recent public debate about the feasibility of dividing states with a vast territorial dimension in Brazil (Shikida, Faria, and Araújo 2014).

References

Abadie, A., A. Diamond, and J. Hainmueller. 2015. “Comparative Politics and the Synthetic Control Method.” American Journal of Political Science 59: 495–510, https://doi.org/10.1111/ajps.12116. Search in Google Scholar

Alesina, A., R. Baqir, and C. Hoxby. 2004. “Political Jurisdictions in Heterogeneous Communities.” Journal of Political Economy 112: 348–96, https://doi.org/10.1086/381474. Search in Google Scholar

Asher, S., and P. Novosad. 2015. The Impacts of Local Control over Political Institutions: Evidence from State Splitting in India. Unpublished manuscript. Search in Google Scholar

Autor, D. H. 2003. “Outsourcing at Will: The Contribution of Unjust Dismissal Doctrine to the Growth of Employment Outsourcing.” Journal of Labor Economics 21: 1–42, https://doi.org/10.1086/344122. Search in Google Scholar

Barone, G., and S. Mocetti. 2014. “Natural Disasters, Growth and Institutions: a Tale of Two Earthquakes.” Journal of Urban Economics 84: 52–66, https://doi.org/10.1016/j.jue.2014.09.002. Search in Google Scholar

Baskaran, T., and S. Blesse. 2019. Subnational Border Reforms and Economic Development in Africa: ZEW-Centre for European Economic Research Discussion Paper. Working Paper. Search in Google Scholar

BID. 2018. Melhores gastos para melhores vidas: como a américa latina e o caribe podem fazer mais com menos: Relatório das Américas - Banco Interamericano de Desenvolvimento. Policy Report, 1–485. Search in Google Scholar

Corbi, R., E. Papaioannou, and P. Surico. 2019. “Regional Transfer Multipliers.” The Review of Economic Studies 86: 1901–34, https://doi.org/10.1093/restud/rdy069. Search in Google Scholar

Dalmazzo, A., G. de Blasio, and S. Poy. 2018. “Local Secessions, Homophily, and Growth. A Model with Some Evidence from the Regions of Abruzzo and Molise (Italy, 1963).” Journal of Economic Behavior and Organization 151: 284–306, https://doi.org/10.1016/j.jebo.2018.04.018. Search in Google Scholar

Gendźwiłł, A., A. Kurniewicz, and P. Swianiewicz. 2020. “The Impact of Municipal Territorial Reforms on the Economic Performance of Local Governments. A Systematic Review of Quasi-Experimental Studies.” Space and Polity: 1–20. https://doi.org/10.1080/13562576.2020.1747420. Search in Google Scholar

Gonçalves, S. R. 2015. O movimento separatista do tocantins: análise das reportagens dos jornais a folha de são paulo e o popular na década de 1980: Brasília: Trabalho de Conclusão de Curso (Licenciatura em História)—Universidade de Brasília. Search in Google Scholar

Grossman, G., J. H. Pierskalla, and E. Boswell Dean. 2017. “Government Fragmentation and Public Goods Provision.” The Journal of Politics 79: 823–40, https://doi.org/10.1086/690305. Search in Google Scholar

Hall, J. C., J. Matti, and Y. Zhou. 2020. “The Economic Impact of City–County Consolidations: A Synthetic Control Approach.” Public Choice 184: 43–77. https://doi.org/10.1007/11127-019-00699-z. Search in Google Scholar

Martins, H. T. 2001. “A fragmentação do território Brasileiro: a criação de novos estados no Brasil.” Caderno CRH 14: 263–88 https://doi.org/10.9771/ccrh.v14i35.18599. Search in Google Scholar

Oates, W. E. 1999. “An Essay on Fiscal Federalism.” Journal of Economic Literature 37: 1120–49, https://doi.org/10.1257/jel.37.3.1120. Search in Google Scholar

Roesel, F. 2017. “Do Mergers of Large Local Governments Reduce Expenditures?–Evidence from germany Using the Synthetic Control Method.” European Journal of Political Economy 50: 22–36, https://doi.org/10.1016/j.ejpoleco.2017.10.002. Search in Google Scholar

Shenoy, A. 2018. “Regional Development through Place-Based Policies: Evidence from a Spatial Discontinuity.” Journal of Development Economics 130: 173–89, https://doi.org/10.1016/j.jdeveco.2017.10.001. Search in Google Scholar

Shikida, C. D., F. R. F. Faria, and A. F. AraújoJr. 2014. “Referendum on the Creation of New States: The Pará Case.” Economic Analysis of Law Review 5: 1–21, https://doi.org/10.18836/2178-0587/ealr.v5n1p1-21. Search in Google Scholar

Received: 2020-06-16
Accepted: 2020-10-01
Published Online: 2020-10-16

© 2020 Ricardo Carvalho de Andrade Lima, published by De Gruyter, Berlin/Boston

This work is licensed under the Creative Commons Attribution 4.0 International License.