Effect of Political Instability on Corruption: An inter-state analysis within United States Ahmed Alnefeesi § Shreyo Mallik ‡ Irina Valenzuela † June 2014 Abstract This paper analyzes the effect of political instability on corruption, in particular, we test whether there is a U-shaped relationship between political stability and corruption as proposed by Campante, Chor and Do (2008), providing a within- country evidence. It is expected that at low levels of stability, the incumbent would find optimal to steal today than from the uncertain future (horizon effect), whereas a higher level of stability, private sector is more willing to offer bribe to stable incumbents. Using data from US states and performing a cross- section and panel-data regression, we find that the estimated coefficients show a U-shaped relationship between both variables, however such evidence is not statistical significant. Keywords: Corruption, Political Instability, Incumbent Tenure Research paper for the course of Political Economy, Msc. Economics, Barcelona Graduate School of Economics. § [email protected]‡ [email protected]† [email protected]
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Effect of Political Instability on Corruption: An inter-state analysis
within United States
Ahmed Alnefeesi§ Shreyo Mallik‡ Irina Valenzuela†
June 2014
Abstract
This paper analyzes the effect of political instability on corruption, in particular,
we test whether there is a U-shaped relationship between political stability and
corruption as proposed by Campante, Chor and Do (2008), providing a within-
country evidence. It is expected that at low levels of stability, the incumbent
would find optimal to steal today than from the uncertain future (horizon
effect), whereas a higher level of stability, private sector is more willing to offer
bribe to stable incumbents. Using data from US states and performing a cross-
section and panel-data regression, we find that the estimated coefficients show
a U-shaped relationship between both variables, however such evidence is not
statistical significant.
Keywords: Corruption, Political Instability, Incumbent Tenure
examine the impact of different electoral rules on corruption. They find that larger voting
districts – which entail lower barriers to entry for candidates - are associated with less
corruption, whereas more candidates being elected from party lists – which entails less
individual accountability – is associated with more corruption. Chang (2005) also
investigates the impact of electoral rules on corruption. His paper hypothesizes that under
an open-list proportional representation election system where individual votes are critical
for politicians to win elections, politicians’ uncertainty of election results drives them to
engage in corrupt behavior to finance campaigns. Using individual level data from pre-1994
Italy, Chang finds substantial evidence that electoral uncertainty can lead to corruption.
Ferraz and Finan (2007) examine the impact of an anti-corruption program in Brazil, which
randomly inspects municipal expenditures of federal funds. They find that disclosures of
local government corruption as a result of the program considerably reduced the likelihood
of re-election for corrupt governors, highlighting the impact of asymmetric information on
political accountability.
Numerous authors have also investigated the relationship between political stability and
corruption. For instance, a key insight from Campante et al. (2008) is that political stability
can have contrasting effects on the incentives for corruption. On the one hand, there is a
horizon effect, from which if an incumbent expects a short tenure in public service, then
he will have more incentive to embezzle public funds today rather than wait for funds
accumulate and steal from others at a later date. On the other hand, there is a demand
effect, from which if an incumbent expects a longer time horizon in tenure, then this can
lead to other forms of corruption that require long term relationships between an
incumbent and a third party, for example, when firms bribe public officials to make
concessions for long-term projects. Hence, as stability of an incumbent’s tenure increases,
the incentive to engage in direct embezzlement decreases, but the incentives to engage in
other forms of corruption such as receiving bribes increases. Using indices on corruption
perception and measures of political stability among countries, Campante et al. find a
robust U-shaped relationship between political stability and corruption. Countries with very
unstable or very stable regimes display more corruption than countries that fall in the
middle.
Olson (1991) associates the incentives of an autocrat to steal public assets to the stability of
tenure. Olson suggests that an autocrat with a stable tenure can secure funds in the form of
taxes, and will thus have an incentive to provide peaceful order and other public goods to
retain a stable source of taxes. On the other hand, if an autocrat expects a brief tenure, then
he will have an incentive to steal public assets if their tax yield during his tenure is less than
their total value. This is similar to the horizon effect in Campante et al. in that incumbents
with unstable tenures will have more incentive to embezzle public funds. Acemoglu (2005)
develops a model that also gives rise to a U-shaped relationship between a ruler’s incentives
to act against the public welfare and the strength of the state. In his model when a state is
excessively strong, then the ruler will have an incentive to maximize rent by imposing such
high taxes that economic activity is suppressed. On the other hand, if the state is
excessively weak, then anticipating that he will not be able to extract future rents, the ruler
will under invest in public goods. Gamboa-Cavazos et al. (2007), empirically examine the
impact of different political horizons on corruption. Using variation in gubernatorial office
terms in Mexico, they also find a U-shaped relationship in which corruption is more
intense during long and short political horizons and less intense in intermediate ones. They
associate this relationship to the incentives of incumbents to receive bribes and the
incentives of firms to give bribes over different political horizons. They suggest that during
short political tenures, incumbents prey more intensely on firms to extract bribes, whereas
during long tenures firms tend to bribe incumbents more because of the longer policy
horizons.
3 Theoretical Model
This section is based on the theoretical model presented by Campante et al. (2008), who
show how instability shapes the incentives for corruption. First, the model considers an
infinite-horizon with an initial pool of available resources (K0), whose allocation is
controlled by an incumbent. There is a probability of stability (α) that the incumbent
continues from one period to the next.
At any given point in time t, the resources can be diverted in either of the two following
ways:
(i) Embezzlement (Et): It refers to direct stealing.
(ii) Licensing (Lt): It involves providing the private sector firms the control over some of
the resources in exchange for an upfront bribe payment.
The incumbent chooses to spend some amount (Pt), out of his initial pool of resources in
order to boost his stability. Let π(Lt, α(Pt)) denote the ex-ante expected value of the
profits received by the firm from the license Lt. We assume that π is an increasing function
of stability and that π(Lt, 0) = 0, so that firms have no interest in bribing to an unstable
incumbent who has a zero probability of being in power in the following period. Another
assumption is that the incumbent has the ability to extract a fraction σ of expected profits
as a bribe payment for the license. In each period, the unutilized resources are transformed
into the pool of resources available in the next period, subject to diminishing returns:
Kt+1 = A(Kt − Et − Lt − Pt)γ
The incumbent maximizes his expected income, whose sequence problem is as follows:
α σπ α
γ
Corruption (Γt) is defined in each period as the amount of illicit income that the incumbent
receives, normalized by the resources available at the beginning of the period as follows:
Γt = [Et + σπ(Lt, α(Pt))]/K
Regarding stability, the variable Pt, referred to as the “public goods provision”, captures
the fact that the stability of the incumbent can be affected by the decisions regarding
resource allocation that he makes. Therefore, a higher public good’s expenditure boosts
the incumbent’s stability but also diverts resources away from his own benefit. Based
on this, stability is specified as a function of the incumbent’s choice of Pt, g(Pt), where g(•)
is increasing and concave, with g(0) = 0. Such function illustrates how effective public
goods provision strengthens the incumbent’s position.
Also, it is included a“systemic” level of stability, which reflects an incumbent’s stability
can be affected by factors that are beyond his control, such as ethnic composition of the
population or cultural norms. Such exogenous factors are incorporated by assuming that
overall stability depends on the intrinsic stability of the polity:ζ ∈ [0, 1]. Therefore,
political stability is specified as α = α(ζ , g(P )).
Finally, the analysis of the comparative statics for corruption with respect to stability is
based on the steady state (t ≥ 1). Since Γt is a decreasing function in α, corruption is
decreasing in stability for α< α*. Thus, more unstable incumbents have a greater
incentive to steal resources currently instead of leaving them to future periods when they
might not be in power. While the firms have incentive to bribe the incumbent so long as α >
0, the expected returns from the licenses are small, in order that any bribes offered are
insufficient to persuade the incumbent to substitute away from purloining.
The influence of α on corruption involves a rich interaction between a horizon effect (since
the optimal amount of licensing also takes into care the trade-off with respect to leaving
resources to the future) and a demand effect (by which the firms are willing to pay higher
bribes to more stable incumbents). The latter effect leads to make corruption increase in
stability. We notice that when γ is sufficiently small (γ ≤1), diminishing returns set in
fast enough in the accumulation equation for Kt. As a consequence upon this, the
incumbent may not prefer to set resources aside for the future. Thus, the demand effect
unambiguously prevails over the horizon effect. If γ > 1, the demand effect still prevails
over some horizon of stability. This continues so long as licensing represents a sufficiently
large source of corruption rents for the incumbent. The horizon effect comes into the
scenario at the highest levels of stability because very stable incumbents find it
worthwhile to allow some resources to accumulate into the future instead of disbursing
more licenses. In short, corruption will increase over some range of stability, though not
necessarily decreasing in stability at the highest levels of α.
4 Data and Empirical Strategy
Our sample correspond to information from US states, which includes the period 1977-
2000.
For the variable of corruption, we used the database from Boylan & Long (2003) and
Maxwell & Winters (2006). In the case of the former, the authors constructed a corruption
indicator based on a survey carried out in 1998-1999. In particular, the authors sent a
questionnaire to state-house news reports, who are supposed to be well-informed of the
states’ government corruption. The questions included in such survey are in appendix 1.
Boylan & Long (2003) constructed two measures of state’s corruption, one based on just
the average of question nº 61 and the other one based on normalizing and then averaging
responses to question 3 to 8. In our study, we are using the latter one, since, as Boylan &
Long mention, there are some concern about whether by using the means of ordinal
measure (as by using just question 6) may lead to a loss of information within individual
data. Anyway, the authors find that both measures of corruption are highly and statistically
significant correlated (0.853), proving consistency between them. It is worth to mention
that since they do not receive response from 3 states (Massachusetts, New Hampshire and
New Jersey), they construct their indicator for only 47 states.
In the case of corruption’s index from Maxwell & Winters (2006), they replicate and
continue the work of Meier and Schlesinger (2002), that is constructing corruption
measured based on reports of “criminal abuses of public trust by government officials”.
The data is based primarily on the reports from US Attorney offices (Public Integrity
Section of the US Department of Justice). The authors divided the number of convictions
by the number of officials, from which the exclude Hawaii as an outlier in corruption. The
indicator is measured in logarithm terms.
The correlation between these two measures of corruption is 0.55.
1 They construct their measure of state corruption based on question 6 for two reasons: i) this question explicitly asks reporters to rank their state on overall corruption and ii) there is high level of agreement in the responses among reporters within same state. In particular, an individual score of 1 on question nº6, reflects that the State House reporter perceived government employees in her specific state as the least corrupt among all states, whereas a score of 7 reflects a perception of this state’s government employees to be the most corrupt. The authors, then, estimate the state score as the average among the report’s score within each state, and then rank states based on means scores.
For the variable of political instability, we are going to use to indicator as proposed by
Campante et al. (2008):
(i) Average tenure of incumbent: In our case, we are going to estimate how many years in
average an incumbent stays as governor within 10-years window, which is calculated as 10
divided by the number of different incumbent govern in such period. It is worth to
mention that we are not going to include as a alternative indicator the average tenure of
party, since in the case of US states there is usually two main parties: Democrats and
Republicans, and for the most of the states, they intercalate.
(ii) Strength of incumbent’s position: we create this variable as the difference of the
percentage of votes between the winning candidate and the runner-up in a certain election.
The information about percentage votes and incumbent condition in each election (re-
elected, retired or term-limit) was collected from Wikipedia2. he controls variables include
in the regression are specified in the appendix 2.
Cross –section analysis
For the cross-section regression, it is going to be used two measures of corruption, one
that correspond to Bong and Lang (2004), as it is stated was based on a survey in 1998-
1999, and as a second measure of corruption, Maxwell and Winters’s indicator is used.
Regarding the political instability variable and in the case of cross-section analysis, it is
considered the average tenure of incumbent in the last 10 years 1989-1999 in order to
coincide with the previous 10 years to the data of corruption. The second indicator of
political instability to be considered in the cross-section analysis is the vote margin
corresponding to the previous election held prior to 1998. In the latter case, what we want
to evaluate is whether the level of corruption is associated with the strength a governor has
reflected in the percentage of vote his receives in elections. Finally, as a third measure of
political instability and related to the average tenure of an incumbent, we are considered the
2 Gao, Pengjie and Yaxuan Qi. “Political Uncertainty and Public Financing Costs: Evidence from U.S. Gubernatorial Elections and Municipal Bond Markets”. 2013, also collected data of US gubernatorial elections from Wikipedia.
number of times a re-election has occurred in the last 22 years (1977-1998), this measure
not only captures long-term effect of political stability. In this case, we are not going to
consider New Hampshire, Rhode Island and Vermont since the elections are held every
two years.
Our regression specification for the cross-section stands as follows: