See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/263189101

The Shape of Corruption: Colombia as a Case Study

Article  in  Journal of Development Studies · September 2013

DOI: 10.1080/00220388.2013.800858

CITATIONS READS

23 2,074

2 authors:

Laura Langbein Pablo Sanabria-Pulido

American University Washington D.C. Los Andes University (Colombia)
69 PUBLICATIONS   2,157 CITATIONS    60 PUBLICATIONS   195 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Laura Langbein on 25 July 2016.

The user has requested enhancement of the downloaded file.

 PROOF COVER SHEET
Journal acronym: FJDS
Author(s): Laura Langbein and Pablo Sanabria
Article title: The Shape of Corruption: Colombia as a Case Study
Article no: 800858
Enclosures: 1) Query sheet
2) Article proofs

Dear Author,
1. Please check these proofs carefully. It is the responsibility of the corresponding author to check
these and approve or amend them. A second proof is not normally provided. Taylor & Francis cannot
be held responsible for uncorrected errors, even if introduced during the production process. Once
your corrections have been added to the article, it will be considered ready for publication.
Please limit changes at this stage to the correction of errors. You should not make insignificant
changes, improve prose style, add new material, or delete existing material at this stage. Making a
large number of small, non-essential corrections can lead to errors being introduced. We therefore
reserve the right not to make such corrections.
For detailed guidance on how to check your proofs, please see
http://journalauthors.tandf.co.uk/production/checkingproofs.asp.

2. Please review the table of contributors below and confirm that the first and last names are
structured correctly and that the authors are listed in the correct order of contribution. This
check is to ensure that your name will appear correctly online and when the article is indexed.

Sequence Prefix Given name(s) Surname Suffix

1 Laura Langbein
2 Pablo Sanabria
Queries are marked in the margins of the proofs.

AUTHOR QUERIES
General query: You have warranted that you have secured the necessary written permission from the
appropriate copyright owner for the reproduction of any text, illustration, or other material in your
article. (Please see
http://journalauthors.tandf.co.uk/preparation/permission.asp.) Please check that any required
acknowledgements have been included to reflect this.
AQ1 Please define FARC and ELN
AQ2 Please define IRR
AQ3 Please provide the page range for this chapter. Please also note that this reference has not
been cited in the text. Please include this or delete from the reference list.
AQ4 This reference has not been cited in the text. Please include or delete from the reference list.
AQ5 This reference has not been cited in the text. Please include or delete from the reference list.
AQ6 This link is not working. Please check and update this.
AQ7 This link is not working. Please check and update this. Please also note that this reference
has not been cited in the text. Please include or delete from the reference list.
AQ8 Please provide the city of publication, as well as the state.
AQ9 Please provide the location and publisher of this report.
AQ10 Please provide the page range of this chapter, and the publisher of the book.
AQ11 Please provide the publisher for this reference. Please note that this reference has also not
been cited in the text. Please include this or delete from the reference list.
The Journal of Development Studies, 2013
Vol. 00, No. 00, 1–14, http://dx.doi.org/10.1080/00220388.2013.800858

The Shape of Corruption: Colombia as a Case

Study
LAURA LANGBEIN* & PABLO SANABRIA**
*Department of Public Administration and Policy, American University, Washington, DC, USA, **Alberto Lleras Camargo
School of Government, Universidad de Los Andes, Bogota, Colombia 5

Final version received March 2013

ABSTRACT We examine the shape of corruption within a country characterised by stable democracy and
thriving legal and illegal markets: Colombia. We look at whether city officials solicit an extra payment when
citizens seek to use public services. Using individual level data on bribe requests in 55 cities from 2004–2011, we
find that the level of corruption is stable, but varies widely within Colombia. Our results suggest that corruption is 10
stable at different levels in some Colombian cities, along with evidence of recent decreases in many, and increases
in some, cities. Further research is necessary to explain this within-country variation.

Introduction
Corruption, commonly defined as ‘the use of public office for private gains’ (Bardhan, 1997;
Swaleheen, 2011: 23), is widely regarded as a wicked social problem (Roberts, 2000, 2012). It is 15
wicked because it robs countries of economic growth in favour of costly transfers of money, often
from the poor to relatively more powerful gatekeepers or providers of needed public services. It is also
wicked because it is a hard problem to solve. Solutions may require independent, third party
enforcement of ‘rules of law’, a collective good that is likely to be impossible to impose exogenously
where it is most needed (Batory, 2012: 71; Klitgaard, 1988). Societies that are the most corrupt are 20
also the most likely to find it impossible to implement rules to effectively change the incentives that
govern the relation between those in the government and citizens (of any rank), even if they can
change the formal rules (Aziz, 2012; Manzetti & Blake, 1996). However, solutions may not require
external enforcement. For example, if corruption is truly a ‘social’ problem, representing a game that is
more likely to be played when others with whom the player interacts are also playing that game 25
(Fisman & Miguel, 2007; Ostrom, 2000), then solutions may require changes in social norms. It is
widely recognised that the means for changing social norms, whatever they are (Ostrom, 2000), will be
different in different locations, and that success is neither guaranteed nor immediate (Roberts, 2012).
We examine corruption within a single country that is characterised by a stable democracy and
thriving legal and illegal markets: Colombia. Beginning in the early- to mid-1990s, and still continuing 30
today, international organisations, and the Colombian government itself, have actively supported
activities, programmes and laws intended to reduce corruption (Hoggard, 2004; Ministerio del
Interior, 2011). While corruption has dropped in Colombia, it remains a serious problem (Hoggard,
2004). For example, according to one report (Taylor, 2011), some 41 candidates in municipal elections
were murdered in the months before the October 2011 elections, in which allegations of corruption 35

Correspondence Address: Laura Langbein, Department of Public Administration and Policy, American University, Washington,
DC 20016, USA. Email: langbei@american.edu

© 2013 Taylor & Francis

2 L. Langbein & P. Sanabria

were prominent. Accusations of corruption in public services also remain prominent (Gonzalez, 2012).
Equally prominent are allegations of corruption’s entrenched relation with the illegal activities of drug
traffickers, guerrillas, and paramilitary groups, which have shaped strong cooptation processes in the
Colombian state (Garay, Salcedo, De Leon, & Guerrero, 2008).
Nonetheless, the latest national administrations have designed institutional and organisational 40
arrangements in order to establish an anticorruption policy for the country at the national level
(Maldonado, 2011; Ministerio del Interior, 2011). Still, at the local level, several corruption scandals
have emerged in Colombian cities and towns during the last decades (Thoumi, 2011). The emergence
of these high levels of corruption at the local level might be expected in the context of a poorly
centralised state with incomplete authority over its territory (Acemoglu & Robinson, 2012). 45
Empirically, both Garay et al. (2008) and the Colombian government itself (Vicepresidencia de la
República de Colombia, 2006) remark that this appears to be the case in Colombia. Compared to the
central government, Colombian cities and municipalities appear to be less concerned with designing
particular policies and programmes to deter corruption (Maldonado, 2011). Although there have been
some local and regional dialogues about corruption in the last decade (Vicepresidencia de la República 50
de Colombia, 2006), only recently, and particularly in the main capital cities, are local governments
starting to design anticorruption policies. However, these are very recent efforts, with Medellin’s being
the first anticorruption policy at the municipal level in the country.
We focus on local, small-scale corruption, examining whether city officials solicit an extra payment
when citizens seek to use public services. Using individual level data on bribe requests from citizens in 55
55 large and small cities in Colombia, measured from 2004 to 2011, we find that the levels of
corruption are mostly stable and different among Colombian cities. However, within Colombian cities,
there is evidence of recent decreases in many cities, and of stability and increases in street-level
corruption in other cities. Despite efforts in Colombia to modernise its central and, to a lesser degree,
its local governments, corruption appears to persist in some Colombian cities, especially in those 60
places where the conflict between (and among) the State, legal and illegal markets, and armed groups
is present (CERAC, 2011) (for example Monteria, Valledupar, Puerto Asis, Cucuta). Interestingly, the
most notable cities in Colombia (Bogota, Medellin, Cali, and Cartagena) are not the most corrupt.
Thus, we conclude that corruption clearly depends on context: the levels clearly differ within a single
country. It follows that it may be useful to study corruption within countries (for example Azevedo & 65
Colaco, 2010; Ionescu, 2011), supplementing studies that focus on between-country comparisons.
The evidence that corruption is often stable is disturbing. Corrupt exchanges are one form of
cooperative behaviour. If cooperative production is stable, the result is likely to be mutually beneficial
exchanges that lead to economic growth. If cooperative predation is stable, it makes effective political
change difficult, if not impossible, to implement, and represents costly transfers, usually from the less 70
well-off to those who are better off. Corruption is a classic example of cooperative predation. Our
results suggest that it is quite stable at different levels in most Colombian cities. Similarly, they
indicate that the incidence of corruption has decreased within many Colombian cities, but the drop
appears sporadic rather than consistent.

Theoretical Background: Corruption Equilibria 75

Previous research in economics (and even biology) suggests that corruption is a state-dependent game,
with multiple equilibria. Roland and Verdier (2003) develop such an expectation in a formal model of
this process, but they do not test it empirically. They frame the process as the decision by two
randomly matched individuals to produce or ‘prey’, based on the probabilities of getting caught, and
the tax rate from production. The result is a model with a ‘good’, productive equilibrium (‘little’ 80
corruption), an opposite ‘bad’ equilibrium (predation), and no clear equilibrium in between them.
This account of equilibria at relatively ‘high’ or ‘low’ levels of corruption is similar to previous
accounts of the evolution of cooperation as a form of Darwinian selection. In these accounts, the
interaction of two randomly selected cooperative pairs results in a larger payoff than that from any
 The shape of corruption: Colombia as a case study 3

other possible random pair-wise combination, where the probability of cooperation at time t depends 85
on the probability at time t-1. In the model, there are two forms of cooperation. One form is to
cooperate in a corrupt exchange; the other is to cooperate in an incorrupt exchange, under an
externally enforceable contract, which may include future obligations. This implies that cooperation
(no matter whether it is corrupt or not) is a repeated game, and that it is state-dependent, so the play at
time t is dependent on the previous play, at time t-1 (Axelrod & Hamilton, 1981; Boorman & Levitt, 90
1973; Hammond, 2000). Thus, when previous levels of cooperation (corrupt or not) are high, the next
play is more likely to reflect the previous level. An implication is that, with no exogenous shock to the
environment, cooperation (in both of its forms) is likely to be stable from one year to another.
Bicchieri (1990) also develops a theoretical model of norms of cooperation in which the probability
that tit-for-tat rather than ‘always defect’ emerges as a rational strategy depends on the frequency with 95
which others in the group are likely to play tit-for-tat. If the groups or societies in this model have
repeated interactions, so that they are not totally independent or unaware of each other (that is, they are
social), then this model predicts both cooperative and non-cooperative equilibria. Similarly, Callahan
(2005) recognises equilibria of ‘anti-social’ and ‘social’ capital, with high levels of cooperation that
characterise both. 100
Another set of models also predicts multiple equilibria, because the probability of action (for
example smoking, crime, corruption, or even voting) depends on what others in the local group are
doing (Bisin, Moro, & Topa, 2011; Krosnick & Judd, 1982; Norton, Lindrooth, & Ennett, 1998).
Social interactions (that is the correlation between an individual’s actions and the actions of her peers)
may arise either from peer effects or self-selection. In these cases, the equilibrium choice of each 105
individual is dependent on the prevalence of the behaviour in the individual’s social group or network.
Depending on the density of the social behaviour, some equilibria may be more ‘sticky’ or stable than
others. For example, Batory (2012) notes that, because the social costs of corruption are high and
diffuse, and because there is little or no individual utility, except for moral satisfaction (Klitgaard,
1988), to be realised from not being corrupt, corruption equilibria, compared to less corrupt states, are 110
likely to be particularly ‘sticky’, that is stable.
It follows that the formal models generally agree that there should be a predominance of equilibria
of varying levels of both corrupt and non-corrupt cooperation (and few changes over time, especially
at extremely high or low levels of corruption). The models do not define any absolute value of
‘extreme’; nor do they provide systematic empirical evidence regarding the conjecture of stability for 115
understanding corruption. We partially fill this gap, using evidence from a country of many extremes,
Colombia, which has both a thriving market in illegal cocaine trade and a thriving legal market in
petroleum and coffee exports. This is a country characterised by a long standing democracy (that is
consistent elections during the last century), also afflicted by a long lasting guerrilla conflict, which in
the last decades worsened with the presence of paramilitary groups and other armed actors linked to 120
criminal and illegal activities such as kidnapping and drug trafficking. In fact, the country was strongly
identified during the 1990s and early 2000s as one of the failing states in the world, with vast areas of
the country lacking any control from the government (Acemoglu & Robinson, 2012), but also with
modern urban agglomerations with strong public services.
In a country of such extremes it would be likely to find considerable variation in corruption (for 125
example bribe seeking) at the local level. Our focus is on local, small-scale corruption: we look at
whether city officials request an extra payment when citizens seek to use city services. In this context,
it is very likely that local norms and social interaction matters: requesting a bribe payment is hazardous
when the solicited (bribed) is likely to turn the solicitor (briber) in to the authorities; but if requesting a
bribe is likely to go unpunished and may even elicit the bribe, it would be quite rational to demand a 130
bribe payment.
To test whether corruption is characterised by multiple equilibria in different social (that is local)
communities, we use data on corruption over time from 21 department capital cities and 34 villages
and towns surveyed by the Latin American Public Opinion Project (LAPOP). Those 34 towns include
some of the areas in the country most affected by the presence of illegal armed forces. We use a simple 135
linear model: if corruption is characterised by multiple and different ‘sticky’ equilibria in different
4 L. Langbein & P. Sanabria

social communities, then we expect that a regression of corruption on time within each local area will
yield no clear evidence of temporal dependence for each area (because corruption is stable, or
‘sticky’). Instead, we expect significant (and different) intercepts for each local area; we also expect
the indicators of time will show no clear pattern and will be individually and jointly insignificant. 140
Within each city, if corruption is completely time dependent (and unstable), we expect that the time
indicators will be significant (in any direction). It is also possible that corruption is both time
dependent and place specific. It is even possible that corruption is completely random and unstable,
which would predict that neither the temporal indicators nor the place indicators would be significant.

Description of Data and Variables 145

To measure corruption, we use individual level survey responses from 2004–2011 provided by the
Americas Barometer of the Latin American Public Opinion Project (LAPOP) of Vanderbilt University,
which offers data about actual (rather than perceived) corruption. The survey asked respondents how
many times they have been requested to pay a bribe in the last 12 months in their interactions with a
public official, the municipal government, the courts, the police, hospitals and schools, or in the 150
context of their own work. The specific questions are:

(1) Did any public official ask you for a bribe in the last 12 months?
(2) In order to process something in the municipality, a permit for instance, during the last year, has
someone asked you to pay a bribe?
(3) Have you had to pay a bribe at the courts during the past year? 155
(4) Did any police officer ask you for a bribe in the last 12 months?
(5) During the last 12 months, have you had to pay a bribe in order to be assisted in a hospital?
(6) During the last 12 months, have you had to pay a bribe in a school?
(7) At work, have you been requested to pay a bribe during the last 12 months?

We create a summative index of the total of the number of times a respondent has been requested to 160
pay a bribe from each of the categories of bribes for each municipality for each year, which is
subsequently our dependent variable.1
Additionally, we use a number of independent variables to estimate our models. First, if corruption
is stable at different levels in each city, we expect each city indicator to be both statistically and
substantively significant. Thus, to test for an equilibrium (constant) level of corruption in each city, we 165
include a dummy variable for each city (Medellin is the reference city). If there are multiple equilibria,
we expect significant dummies for each individual city, and that most of the dummy values will be
different from each other.
Second, we use each year of the LAPOP survey as a time indicator (dummy) variable to capture the
possibility of temporal dependence, such that the level of bribes in a local area at time t systematically 170
increases or decreases, or fluctuates, depending on its initial level. Many patterns of temporal
dependence could emerge. For example, there could be an upward (or downward) trend, such that
parameter estimates for the time indicators systematically increase or decrease; alternatively, there
could be a clear pattern of instability, so that parameter estimates for each time indicator differ, with no
clear pattern. If there are only corruption equilibria, we expect no evidence of temporal dependence, 175
implying that the time dummies will generally be insignificant. It is also possible that there may be a
mixed pattern of equilibria and temporal dependence.
Finally, not all respondents in our study have the opportunity to be requested to pay a bribe. First, if
the individual did not use any public services he might not be asked to pay a bribe for the service.
Recognising that those who had not chosen to use a particular service (for example a hospital or a 180
court) will not be asked for a bribe, we add a dummy variable to indicate exposure to an opportunity to
be asked to pay a bribe. The variable has a value of one for those individuals who used the service in
the last 12 months and zero otherwise. Next, we add a dummy variable with value one for all those
 The shape of corruption: Colombia as a case study 5

individuals who were asked and actually answered the questions regarding bribes or exposure, and
zero for those who were asked but did not respond. In this way, we make use of statistical controls to 185
mitigate problems of bias due to self-selection (choice to use a service; choice to respond to the
questions).
These two indicator variables also allow us to mitigate some disadvantages of using reported bribe
requests as our measure of corruption. We deliberately avoid using perception to measure corruption.
Perceived levels of corruption are systematically biased in favour of higher levels in countries or areas 190
with a free press, and also measure characteristics that go beyond what is commonly regarded as
corruption (Kurtz & Schrank, 2007; Langbein & Knack, 2010; Thomas, 2010). Further, many of the
perception-based indices of corruption are not comparable over time, a critical limitation for use in our
study (Treisman, 2007). However, actual reports of bribes are also biased: respondents are often
reluctant to report bribes, and the reluctance is unlikely to be random. Hence, we might observe that 195
respondents who used a public service will respond that they were never solicited for a bribe (zero on
the bribe frequency dependent variable) when in fact they were. In this case, the exposure variable will
have a negative value; by contrast, if respondents who use services are more likely to report bribes,
then the exposure variable will have a positive sign. Similarly, respondents who were solicited for a
bribe may choose not to answer any of the questions; they will report zero bribes, even if they used a 200
city service, but simply refuse to answer the question about exposure because of illegal bribe
solicitation.

Method of Estimation of Model Parameters

We estimate two sets of regression models that test for time dependence and stability. Specifically, we
estimate both a model including all 55 cities, and also 55 separate models for each city. The first model 205
examines the entire sample of individual respondents in each city over time. We create a time series,
cross-section design, using individual responses for each type of bribe for each of the 55 municipalities
(21 department capitals and 34 towns) included in the LAPOP sample for each year. Thus we have
repeated measures for independent samples of respondents in each city for each year. To measure bribe
requests, we create a municipality bribe index by summing the seven bribe requests for each city or 210
town for each year. To examine stability and temporal dependence, we include a dummy variable for
each city and each year.
If multiple equilibria and temporal stability characterise the data, then the year indicator variables
will not be significant but the separate city dummies will be significant and different from each other.
The second model examines whether there is a pattern of temporal dependence (or stability) within 215
each city.
We use a zero inflated Poisson regression to estimate the model parameters. We do this for several
reasons. First, our data are count data that are absolute frequencies that cannot take on a value that is
less than zero. Second, it is unlikely that the stochastic term will be normally distributed, since the
probability of observing zero or a low number of bribes (as with any illegal behaviour) is (exponen- 220
tially) much higher than the chance of observing a large number of bribes. Third, we use a Poisson
model because the mean of the dependent variable (0.16) is not significantly different from the
variance (0.24). Fourth, we need to condition the model because there are two types of respondents
who can respond that they had zero requests to pay a bribe for city services: those who used city
services but really had no bribe requests, and those who either used no services or failed to respond to 225
the survey questions. The zero inflated Poisson regression jointly estimates each component of the
model. The first (zero inflation) estimate is from a logit regression, where the dependent variable is a
zero (for any reason, including refusal to respond to the question even if the respondent used city
services) or a one, which means the respondent reported bribes (of any number). The independent
variables are whether the respondent used city services (yes = one, or no = zero), and whether the 230
respondent answered the questions on city services or bribes (yes = one, or no = zero). Thus, the zero
inflation component separates the two sources of ‘fake’ zeros from ‘real’ zeros. Real zeros occur when
6 L. Langbein & P. Sanabria

the respondent used the service, answers the question, and truthfully reports that there was no bribe
request, or when the respondent truthfully reports not using the service. ‘Fake’ zeros occur when the
respondent chooses not to respond to the question (even if a bribe was requested), or when the 235
respondent used the service but reports zero bribes even though a bribe was requested. The second
component of the equation (the Poisson) regresses the zero-inflated bribe index on time and city
dummies. Because of social interaction within local areas, we expect that the variance of the residuals
will be smaller within than between cities; hence we report clustered standard errors. We also report
the results from Poisson estimates of the regression of bribe requests on time dummies with statistical 240
controls for the two exposure variables for each city.2 The estimates we report are incident rate ratios.

Descriptive Statistics
The LAPOP data show similar frequencies for each year (Table 1) of the survey.3
We use seven bribe variables that ask for different chances of corruption. Each of these variables
asks whether the individual has been requested to pay a bribe while using a public service. According 245
to the results in Table 2, the percentage of bribed individuals appears to be relatively low in the
Colombian context. The most common request is from the police or from a city official to process a
request (Table 2a). Most respondents have not been requested to pay a bribe; about 10 per cent have
been requested to pay one or two bribes (Table 2b). However, this situation might conceal local
differences, since some municipalities can be more prone to corruption than others. Thus, our purpose 250
in this paper is to examine bribes at the city level over time across a sample of localities.
Bribes vary among Colombian cities and regions. The overall mean level of the bribe index is 0.16.
The mean level of the bribe index in the city we use as the reference city, Medellin, is 0.18, with a
range from zero to four. Medellin is particularly important as a reference city since it is the main
Colombian city that has adopted a local anticorruption policy. Given the sample size in Medellin, the 255
bribe score of 0.18 implies a rate of bribes requested from three in 10,000 citizens, according to the
proportional sample size. In general, the capital cities appear to show very low rates per 10,000

Table 1. Number of observations by year

Year Frequency Percentage

2004 1,483 12.40

2005 1,487 12.44
2006 1,491 12.47
2007 1,491 12.47
2008 1,503 12.57
2009 1,493 12.49
2010 1,506 12.60
2011 1,503 12.57

Table 2a. Bribes frequencies, by type of bribe

Variable Observations Frequency yes Percentage

Police bribe 11,922 681 5.71

Public official bribe 11,933 347 2.91
Municipality bribe 2,650 135 5.09
Own work bribe 7,004 293 4.18
Courts bribe 1,595 56 3.39
Hospital bribe 5,971 217 3.63
Schools bribe 5,161 126 2.44
 The shape of corruption: Colombia as a case study 7

Table 2b. Number of bribes frequencies, total (zero includes all missing values, which
means no bribe requested, or no use of city services, or no response to either survey
question)

Total number of bribes requested5 Observations Percentage

0 10,597 88.63
1 997 8.34
2 263 2.20
3 75 0.63
4 20 0.17
5 4 0.03
6 0 0
7 1 0.01

inhabitants; however there is still variation among them. The capital of the country, Bogota, reports a
mean of 0.21 (with a range from zero to seven); given the large sample size from this very large city,
the rate is barely one in 10,000. Similarly, Cali, the most important city of Colombia’s south-west, 260
reports a mean of 0.20 and a range of zero to four; the rate is three in 10,000.
However, other capital cities in the Caribbean coast show higher levels of bribe averages.
Barranquilla, the biggest city of the Caribbean region, has a mean of 0.28 with a range of zero to
five, and a rate of bribes requested from eight in 10,000 citizens. Cartagena, another large city on the
same region, reports a mean of 0.19 and a range from zero to three; the rate is also eight in 10,000. The 265
Caribbean region has been traditionally associated with a strong presence of illegal forces (CERAC,
2011: 223) and, according to our evidence, shows high levels of bribe prevalence. This is clearly the
case for another Caribbean capital, Monteria, a city affected by the actions of paramilitary groups, with
the highest mean of 0.36, and a range of just zero to three, but a very high rate of 37.5 in 10,000.
Similarly, Valledupar and Santa Maria have means of 0.23 and 0.18, respectively, which are high for 270
cities that do not exceed 300,000 inhabitants.
Interestingly, the smaller cities show remarkable differences across regions. There is clearly a high
level of bribing in areas affected by the guerrilla conflict, where the State is weak and groups like
AQ1 FARC or ELN have a strong presence (CERAC, 2011). For instance, Rosas (0.25) and Popayan (the
Cauca’s department capital; 0.15) both have relatively high means. Similarly, Puerto Asis in the 275
Putumayo region (0.19), Cucuta, the capital of the border region of Norte de Santander (0.29), and
Villavicencio, the point of entrance to a traditionally afflicted area (0.32), also report relatively high
bribe frequencies.
At the other end of the scale, we found that very small towns appear to exhibit a very low
prevalence of bribes, according to the LAPOP survey. For instance, we find the cities of Quinchia 280
(in Risaralda, a department almost unaffected by the conflict) and San Martin (in Cesar), with means
of 0.04, ranging from zero to two bribes requested, and rates of one in 10,000. Other remarkable cases
in this group of small towns and villages are those in the Departments of Boyaca (Sutamarchan and
Tunja, with means of 0.07 and 0.11, respectively) and Tolima (Coyaima and Icononzo, with means of
0.09). (Table not shown.) 285
Table 3 shows that 21 per cent of the respondents in our sample reported using at least one public
service in the last year, and Table 4 shows that about one-quarter of the respondents in our sample

Table 3. Frequencies, use of public services

Use of public services Observations Percentage

No 9,385 78.49
Yes 2,572 21.51
Total 11,957 100.00
 8 L. Langbein & P. Sanabria

Table 4. Frequencies, response to question about use of services or bribes

Responses to question about use of services Observations Percentage

No 8,949 74.84
Yes 3,008 25.16
Total 11,957 100.00

actually answered the question (one way or the other) about the use of public services and bribe
requests.

Results 290
Table 5 reports the results. The inflation adjustments are significant; however, that is no surprise when
the N is large. Use of public services actually appears to reduce the request for bribes, while failure to
respond to the question appears to have the opposite effect of the same magnitude. The first result,
with a negative coefficient, is consistent with the expected result that respondents who use public
services and are bribed have an incentive to report that they are not bribed. This may be especially
likely to characterise areas where illegal groups have been present: individuals in these areas may be 295
particularly less willing to reveal or report illegal activities, even in an anonymous survey. The second
estimate is consistent with the expectation that those who do not answer either question may have used
the services but were not bribed, they used the services, were bribed, and chose not to answer at all, or
they did not use the service and were consequently not bribed. All of these expectations lead to the
observed positive coefficient. 300
With respect to the main regression, only one of seven time dummies is statistically significant.
Specifically, compared to the reference year, 2004, the incidence of corruption fell by 30 per cent in
2008 (p = 0.02). It is hard to argue that this one ‘blip’ represents substantial temporal dependence.
However, the time dummies are jointly significant.4 Nonetheless, with the exception of one year, the
absence in the overall results of a systematic temporal drop in bribe requests during the years of our 305
study occurs, despite efforts by the Colombian government to modernise its central and local
governments (Sanabria, 2010). In fact, most of the city dummies are significant, implying different
levels of stability in each local area. Of the 55 city dummies, 34 have significantly lower intercepts
AQ2 than the reference city of Medellin (meaning IRRs of 0.90 or less), while 13 have significantly higher
intercepts than Medellin (meaning IRRs of 1.1 or more). Thus, most of the cities have fewer bribes 310
requested than in Medellin. These cities include many of the smaller municipals in Colombia
(Abejorral, Bello, Caldas, San Juan Nepomuceno, Sutamarchán, Manzanares, Padilla, Cerro de San
Antonio, Quinchia, Santuario, Riofrío, and Mitu, for example). This is not surprising given the lower
chance that an individual in a small town with fewer (and poorer) public services is solicited to pay a
bribe, but the zero inflation adjustment adjusts for this. The lower rate of bribe requests might also be 315
due to the existence of social capital networks in small villages and rural areas that rule out the
presence of money bribes in exchange for non-cash ‘payments’.
The cities where more bribes are requested than Medellin include the Caribbean cities of
Barranquilla and Monteria, but also some smaller capital cities like Villavicencio and, arguably,
Cucuta, Armenia, and Pereira. Some of these cities, affected by the presence of both guerrillas and 320
paramilitaries, are vulnerable to their influence over local government activities and citizens’ beha-
viours. The cases of Cucuta, Villavicencio, and Monteria, with particularly high incident rate ratios of
1.4–1.7 compared to Medellin, reflect the persistence of the conflict and the cooptation of State by
illegal armed groups.
It is notable that the largest, most prominent cities in Colombia are not much different than Medellin 325
(that is Bogota, Cali, and perhaps Cartagena). (The incident rate ratio estimates for these cities are 1.1,
which is significantly but arguably not substantively much higher than Medellin.) This reflects a
Table 5. Regression of bribes offered on opportunity for bribe, city and time trend (zero inflated Poisson incident rate ratio estimate, standard errors adjusted for clustering
within city)

Zero Inflated Poisson Zero Inflated Poisson Zero Inflated Poisson

DV: Bribe Index Estimates P>|z| DV: Bribe Index Estimates P>|z| DV: Bribe Index Estimates P>|z |

Dummy 2005 1.063 0.624 Florencia (Caqueta ) 1.077 0.000 Armenia (Quindio) 1.370 0.000
(0.133) (0.018) (0.032)
Dummy 2006 0.789 0.115 Popayan (Cauca) 0.786 0.000 Pereira (Risaralda) 1.191 0.000
(0.118) (0.008) (0.035)
Dummy 2007 0.820 0.099 Padilla (Cauca) 0.429 0.000 Quinchia (Risaralda) 0.236 0.000
(0.098) (0.005) (0.003)
Dummy 2008 0.709 0.018 Rosas (Cauca) 1.263 0.029 Santuario (Risaralda) 0.403 0.000
(0.103) (0.029) (0.004)
Dummy 2009 1.008 0.956 Valledupar (Cesar) 0.896 0.000 Bucaramanga (Santander) 0.836 0.000
(0.139) (0.012) (0.008)
Dummy 2010 0.937 0.635 Pelaya (Cesar) 0.446 0.000 Confines (Santander) 0.548 0.000
(0.129) (0.007) (0.007)
Dummy 2011 0.856 0.239 San Martin (Cesar) 0.217 0.000 Rionegro (Santander) 0.489 0.000
(0.113) (0.002) (0.005)
Abejorral 0.569 0.000 Monteria (Cordoba) 1.745 0.000 Chalan (Sucre) 0.415 0.000
(Antioquia)
(0.010) (0.042) (0.025)
Bello (Anti 0.666 0.000 Ayapel (Cordoba) 0.549 0.000 Majagual (Suicre) 0.766 0.000
oquia)
(0.010) (0.005) (0.010)
Caldas 0.723 0.000 Facatativa 1.175 0.000 Morroa (Sucre) 0.873 0.000
(Antioquia) (Cundinamarca)
(0.011) (0.016) (0.025)
Marinilla 0.975 0.167 Soacha (Cundinamarca) 0.643 0.000 Ibague (Tolima) 1.054 0.001
(Antioquia)
(0.017) (0.011) (0.017)
Barranquilla 1.542 0.000 Neiva (Huila) 0.478 0.000 Coyaima (Tolima) 0.461 0.000
(Atlantico)
(0.026) (0.004) (0.004)
Soledad 1.281 0.000 Mitu (Vaupes) 0.368 0.000 Icononzo (Tolima) 0.500 0.000
(Atlantico)
(0.014) (0.005) (0.005)
Bogota D.C. 1.104 0.000 0.921 0.000 Cali (Valledel Cauca) 1.052 0.000
The shape of corruption: Colombia as a case study 9

(continued )
 Table 5. (Continued)

Zero Inflated Poisson Zero Inflated Poisson Zero Inflated Poisson

DV: Bribe Index Estimates P>|z| DV: Bribe Index Estimates P>|z| DV: Bribe Index Estimates P>|z |

Santa Marta
(Magdalena)
(0.008) (0.013) (0.009)
Cartagena 1.100 0.000 Cerro San Antonio 0.342 0.000 Buga (Valledel Cauca) 0.637 0.000
10 L. Langbein & P. Sanabria

(Bolivar) (Magdalena)
(0.015) (0.003) (0.007)
San Juan N . 0.4 69 0.000 Villavicencio (Meta ) 1.5 30 0.000 Florida (Valle d el Cauca) 0.715 0.000
(Bolivar)
(0.005) (0.035) (0.009)
Tunja ( Boyaca) 0.5 93 0.000 Guamal (Meta ) 0.977 0.066 Riofr io (Valle d el 0.412 0.000
Cauca)
(0.007) (0.012) (0.004)
Muzo (Boyaca) 0.8 07 0.000 Pasto (Nariño) 1.1 31 0.000 Monte rrey (Casanare) 0.675 0.000
(0.011) (0.028) (0.004)
Suta marchan 0.3 78 0.000 Cumbitara ( Nariño) 0.7 44 0.000 Puerto Asis (Putu mayo) 1.001 0.912
(Boyaca)
(0.006) (0.014) (0.012)
Manizales 1.292 0.000 Sandona (Nariño) 0.751 0.000 Inde pendent Variables
(Caldas)
(0.032) (0.005) Dummy Used city −1.1 31 0.000
services
Manzanares 0.336 0.000 Cucuta ( Norte d e Sa- 1.371 0.000 (0.233)
(Caldas) nta nder)
(0.003) (0.023) Dummy Answered bribe 1.095 0.000
questions
(0.284)
Observations 11.957
 The shape of corruption: Colombia as a case study 11

similar institutional framework in these cities, where there is a strong co-existence of modern practices
of government (for example merit based procedures, strict procedures in contracting out) with
entrenched illegal structures and informal rules schemes that favour political patronage, clientelism 330
and cronyism, which finally lead to greater corruption levels.
There is clear evidence of multiple equilibria. In all cases, the estimated value of the intercept
dummies is not ‘large’, ranging from an incidence rate ratio of 1.7:1 and 1.5:1 more bribes requested
(in Monteria and Barranquilla) than in Medellin to about 0.3:1 fewer bribe requests (in San Martin,
Manzanares, Cerro San Antonio, Quinchia, Padilla, Riofrio, and Mitu) than in Medellin. However, 335
while no estimate is especially ‘large’ at the individual level, where the incidence of bribes is low
(ranging from zero to seven, with a mean less than one), an IRR that ranges from 0.3:1 to 1.7:1 implies
30 per cent fewer to 70 per cent more, respectively, bribe requests for each 100 respondents. In
addition, the range of IRR estimates means that the rate of bribed solicitation is nearly six times greater
in some places than others. 340
Five elements might explain the differences between the capital cities and smaller towns. First, the
large capitals have a greater range and variety of public services to offer, which can make bribing a
more frequent outcome. Medellin and the nine other capital cities, as expected, considering their size
and greater range of public services, exhibit more opportunities for corrupt exchanges between citizens
and city workers for city services than smaller towns in Colombia; however, the zero inflation 345
adjustment attempts to control for this effect. Second, it could also be possible that non-monetary
exchanges (for example bartering instead of bribes) take place in towns with small populations. Third,
it makes sense that bribes are less frequent in those areas where the Colombian state has a weak
presence if there are no public officials to bribe. Fourth, in those areas afflicted by the guerrilla or the
paramilitary conflict, citizens may be less willing to report any kind of illegal activities (including 350
bribes), but the zero inflation adjustment also attempts to control for this effect. While we cannot rule
out the possibility that the framing of the question ruled out reporting non-monetary exchanges, we
suspect that the weak Colombian state in some (rural) areas of the country makes bribes less likely,
since there is little public presence. Fifth, the biggest cities are more likely to have an institutional
framework that controls corrupt behaviours more effectively than in peripheral smaller towns. Larger 355
cities may also have a greater concentration of social capital in terms of levels of education and
citizens’ organisations. In fact, the large cities are more likely to have a policy agenda that includes
anticorruption policies, exemplified in the case of Medellin during recent years.
We also estimated Poisson regressions with statistical controls for the two sources of non-response
(non-use of service, and non-response, given use), for each of the 55 cities in our study. (Results 360
available but not shown.) In no case is there evidence of a clear upward or downward pattern in the
time dummies that represent the eight-year duration of our study, relative to the base year (2004).
There are clear equilibria in 14 of the cities, since no time dummy, or only one of the seven time
dummies, is significant. These cities include large cities (for example Cali and Cartagena) as well as
small (for example Guamal, Sandona, and Guadalajara de Buga) and medium (for example Monteria, 365
Cucuta) cities. Some of these cities have been the locus of paramilitary and guerrilla activities, but
some have not; some of the cities are on the coast, and some are inland. Among these cities, the pattern
of reduction often appears in the last two years of our study, representing, in some cases, a 100 per
cent drop in the incidence rate ratio, relative to the base year. Statistically and substantively significant
reductions in corruption during the last two years of our study (2010 and 2011) appear in Barranquilla, 370
Santa Marta, Cerro San Antonio, Padilla, Ayapel, Medellin, Caldas, Bucaramanga, Facatativa,
Villavicencio, San Martin, Monterrey, Neiva, Coyaima, Icononzo, Florida, Riofrio, Popayan, Pasto,
Cumbitara, and Puerto Asis. These cities vary in size, geographical location, and paramilitary and
guerrilla activity. A few cities (seven) show relatively consistent increases in bribe seeking relative to
the base year (Manizales, Quinchia, San Juan Nepomuceno, Muzo, Mitu, and Morroa). The latter five 375
are located in areas of strong internal conflict.
Overall, these results support our conjecture of multiple corruption equilibria within Colombia:
bribe seeking appears less frequently in some cities than others. However, there is also considerable
variation over time within cities. Many cities show that bribe seeking dropped in 2010 and 2011,
12 L. Langbein & P. Sanabria

compared to the base year of 2004. This study provides no explanation about why the decrease 380
occurred, nor can it explain why a decrease (or an increase) occurs in some cities and not others. It is
likely that detailed studies at the micro level, based on theories regarding the social context of
corruption, norms, and cooperation, can provide useful information about why street-level corruption
sometimes persists despite important policy and political intent to reduce its incidence, and why it
sometimes decreases. 385

Discussion and Conclusion

Our general results suggest that corruption, measured as reports of bribe requests, is quite stable
at different levels in most Colombian cities. At the city level, there is evidence of decreases in
the incidence of corruption in most, but not all, cities, and there is evidence of stability in many
cities, and an increase in corruption in a few cities affected by a strong presence of illegal 390
actors.
While we study only one form of corruption, our results in this paper add to existing research on the
presence of street-level corruption in several ways. First, it offers an empirical test of a commonly
expected theoretical model of corruption. We find that street-level corruption (in Colombia) is
characterised by local equilibria; the equilibrium appears to be context dependent, since different 395
cities have different levels of corruption. If corruption is characterised by multiple equilibria, the
policy implications are rather dismal, not only because stable corruption is costly, but also because it is
likely to necessitate multiple solutions. Second, and perhaps most important, we examine street-level
corruption within a country rather than corruption across countries. This type of corruption is
undoubtedly contextual; this implies that it varies within countries. Even relatively ‘clean’ countries 400
have sub-areas where corruption continues to thrive, as it has for centuries. Nonetheless, there is
evidence of decreases in the incidence of corruption in most, but not all, cities. It is also notable that
there are instances of both stability and increases in the incidence of bribe requests within a few
Colombian cities.
Although both economic performance and the quality of national governance have improved 405
in this South American country during the last 10 years, certain local towns and villages remain
affected by the presence of diverse illegal actors (guerrillas, neo-paramilitaries, drug producers
and dealers; Valencia, 2012). While many cities show a recent reduction in corruption, some
show no change, and a few show increases; the level of corruption varies widely from one city
to another. The combination of the strong presence of such actors and the poor presence of the 410
State in marginal areas may determine the existence of multiple corruption equilibria within a
country.
We do not make any strong causal claims about the determinants of this form of corruption.
Rather, we focus attention on its frequently stable, context-contingent appearance. While we do find
evidence of reductions in bribe requests, especially in the last two years of our data, the drops 415
appear to be sporadic. There is also evidence of increases in some places. The predominant evidence
is of generally stable, but different, levels in each city. It is likely that each corruption equilibrium
may have different institutional roots: the institutional conditions that account for corruption in one
place may not work the same way in another context (Bednar & Page, 2005). At the cross-national
level, there are many types of corruption (Johnston, 2005). Cross-national institutional diversity 420
(Ostrom, 2005) makes different forms of corruption inevitable, and implies that multiple causes of
corruption are very likely and are probably specific to each country. Similarly, there is institutional
and economic diversity within countries: legal economic opportunities vary within nations, possibly
in accord with the diversity of the climate and geopolitical history. Our analysis, however, is at the
individual level. We do not know why some citizens chose to pay bribes when given the opportunity 425
while others did not. We suspect that prevailing social norms matter, even in different institutional
settings (Langbein & Jorstad, 2004). Additional research must investigate who chooses to pay and
play, and who does not.
 The shape of corruption: Colombia as a case study 13

Notes
1. The first two items in the set of questions regarding bribes in the LAPOP data ask first about police, then about generic public 430
officials. The subsequent five questions are only asked if the respondent has used the service. The only possible duplication is
from a respondent who says that a generic public official solicited a bribe, and that, for example, she was also asked to pay a
bribe while at work. Among the 1,360 total number of bribes, we count only 57 observations like these, but even these 57
responses may not double-count the same request. Given the contingent nature of the questions, we believe that such a
response reflects two separate instances of bribes, one generic, and one clearly asked only of those who work. 435
2. Because of a relatively small N and numerous zero values for both the dependent and many independent variables, the zero
inflated Poisson did not converge at the city level.
3. Colombia is the only Latin American country that has been consistently surveyed by the LAPOP project every single year
since 2004.
4. The difference between the likelihood estimates from a model that includes the seven time dummies and one that does not 440
include them is 11.28, which is a chi-square of about 22.5 that is significant at p = 0.005 (Wooldridge, 2009: 589). We discuss
the varying temporal patterns of bribe requests when we discuss in results for individual cities below.
5. This shows how many individuals answered yes to each of the seven questions listed above.

References
Acemoglu, Daron, & Robinson, James. (2012). Why nations fail. The origins of power, prosperity, and poverty. New York: 445
Crown Publishing Group.
Axelrod, Robert, & Hamilton, William D. (1981). The evolution of cooperation. Science, New Series, 211, 1390–1396.
Azevedo, A., & Colaco, M. (2010). Relação entre Emendas Parlamentares e Corrupção Municipal no Brasil: Estudo dos
Relatórios do Programa de Fiscalização da Controladoria-Geral da União. RAC - Revista de Administração Contemporânea,
14, 414–433. 450
Aziz, D. (2012). Anticorruption and its discontents: Anti-corruption in post independence colonial bureaucracies. Paper
presented at International Public Management Network Conference on Corruption, Honolulu.
Bardhan, P. (1997). Corruption and development: A review of issues. Journal of Economic Literature, 35, 1320–1346.
Batory, A. (2012). Why do anti-corruption laws fail in central eastern Europe? A target compliance perspective. Regulation and
Governance, 6, 66–82. 455
Bisin, A., Moro, A. & Topa, G. (2011). The empirical content of models with multiple equilibria in economies with social
AQ9 interactions (Report No. 504). Federal Reserve Bank of New York Staff Report.
Bednar, J., & Page, S. (2005). Culture, institutional performance, and path dependence. Unpublished manuscript.
Bicchieri, C. (1990). Norms of cooperation. Ethics, 100, 838–861.
Boorman, S., & Levitt, P. (1973). A frequency-dependent natural selection model for the evolution of social cooperation 460
networks. Proceedings of the National Academy of Sciences of the United States of America, 70, 187–189.
Callahan, W. (2005). Social capital and corruption. Perspectives on Politics, 3, 495–508.
CERAC. (2011). Riesgo por Presencia de Grupos Armados Ilegales en Colombia. In MOE (Misión de Observación Electoral)
AQ10 (Eds.), Mapas y Factores de Riesgo Electoral Elecciones Colombia 2011. Bogota.
Fisman, R., & Miguel, E. (2007). Corruption, norms, and legal enforcement: Evidence from diplomatic parking tickets. Journal 465
of Political Economy, 115, 1020–1048.
Garay, L., Salcedo, E., De Leon, I., & Guerrero, B. (2008). La Captura y Reconfiguración del Estado en Colombia. Bogota:
Método, AVINA & Transparencia por Colombia.
Gonzalez, Jenny. (2012). The corruption at the heart of Colombia’s health care system. America Economía. Retrieved from
http://www.worldcrunch.com/world-affairs/the-corruption-at-the-heart-of-colombia-039-s-health-care-system/colombia- 470
health-care-corruption-privatization/c1s9688/#.UK0ZdY5rW0g
Hammond, R. (2000). Endogenous transition dynamics in corruption: An agent-based computer model. Center on Social and
Economic Dynamics Working Paper 19. Washington, DC: Brookings Institution.
Hoggard, S. (2004). Preventing corruption in Colombia: The need for an enhanced state-level approach. Arizona Journal of
International and Comparative Law, 21, 577–619. 475
Ionescu, L. (2011).Mexico’s pervasive culture of corruption. Economics, Management & Financial Markets, 6, 182–187.
Klitgaard, R. (1988). Controlling corruption. Berkeley: University of California Press.
Klitgaard, R., Fedderke, J., & Akramov, K. (2005). Choosing and using performance criteria. In R. Klitgaard & P. C. Light
AQ3 (Eds.), High performance government: Structure, leadership, incentives. Santa Monica, CA: Rand Corp.
Konrad Adenauer Stiftung. (2011). Alcances y Limitaciones del Estatuto Anticorrupción: Recomendaciones para una 480
AQ4 Reglamentación Efectiva y Eficaz. KAS Papers No. 14. Bogota: KAS – Transparencia por Colombia.
Krosnick, J., & Judd, C. (1982). Transitions in social influence in adolescence: Who induces cigarette smoking? Developmental
Psychology, 81, 359–368.
 14 L. Langbein & P. Sanabria

Kurtz, M. J., & Schrank, A. (2007). Growth and governance: Models, measures, and mechanisms. The Journal of Politics, 69,
538–554. 485
Johnston, M. (2005). Syndromes of corruption: Wealth, power and democracy. New York: Cambridge University Press.
Langbein, L., & Jorstad, C. (2004). Trust and productivity in the workplace: Cops, collusion, communication, and cooperation.
Political Research Quarterly, 57, 65–79.
Langbein, L., & Knack, S. (2010). The worldwide governance indicators: Six, one or none? Journal of Development Studies, 46,
350–370. 490
Maldonado, A. (2011). La lucha contra la corrupción en Colombia: la carencia de una política integral. Bogota DC: Fescol.
Manzetti, Luigi, & Blake, Charles H. (1996). Market reforms and corruption in Latin America: New means for old ways. Review
of International Political Economy, 3, 662–697.
AQ5 Ministerio del Interior y Justicia. (2011). Estatuto Anticorrupción. Bogota DC: Imprenta Nacional de Colombia.
Norton, E., Lindrooth, R. & Ennett, S. (1998). Controlling for the endogeneity of peer substance use on adolescent alcohol and 495
tobacco use. Health Economics, 7, 439–453.
Ostrom, E. (2000). Collective action and the evolution of social norms. Journal of Economic Perspectives, 14, 137–158.
Ostrom, E. (2005). Understanding institutional diversity. Princeton, NJ: Princeton University Press.
Roberts, N. (2000). Wicked problems and network approaches to resolution. International Public Management Review, 1, 1–19.
Roberts, N. (2012). Tacking wicked problems in Indonesia: A bottom-up design approach to reducing crime and corruption. 500
Paper presented at International Public Management Network Conference on Corruption, Honolulu.
Roland, G., & Verdier, T. (2003). Law enforcement and transition. European Economic Review, 47, 669–685.
Sanabria, P. (2010). Dos Pasos Adelante, Uno Hacia Atrás: Colombia y la Configuración de un Servicio Civil Profesional y
Meritocrático. Boletín Política Pública Hoy. Bogota DC: Departamento Nacional de Planeación.
Swaleheen, M. (2011). Economic growth with endogenous corruption: An empirical study. Public Choice, 146, 23–41. 505
Taylor, G. (2011). Municipal elections highlight split among Colombia’s elites. World Politics Review. Retrieved from http://
www.worldpoliticsreview.com/articles/read/ACej3qlY22CyULSDlEC7VNDCsp6h4hT5bKAMwba7LnXI0T-
AQ6 J1xS4K8Tl2HiNbQSQm516xojJePxs9vvnyhXsgUtTBG5-Lm8QLQPFJEkMBZaIZf5JUh8sO2rycSX7QqaX
Thomas, M. (2010). What do the worldwide governance indicators measure? European Journal of Development Research, 22,
31–54. 510
Thoumi, F. (2011). Colombian organized crime: From drug trafficking to parastatal bands and widespread corruption. Traditional
Organized Crime in the Modern World. Studies of Organized Crime, 11, 131–148.
Treisman, D. (2007). What have we learned about the causes of corruption from ten years of cross-national empirical research?
Annual Review of Political Science, 10, 211–244.
Valencia, R. (2012). Colombia and FARC: Will the internal conflict reach and end? Yale Journal of International Affairs. 515
Retrieved from http://yalejournal.org/2012/07/colombia-and-farc-will-the-internal-conflict-reach-an-end/
Vicepresidencia de la República de Colombia. (2006). Propuesta de una Política de Estado para el Control de la Corrupción.
AQ11 Bogota.
Weisstein, E. (1999). Hyperbolic tangent. Wolfram MathWorld. Retrieved from http://mathworld.wolfram.com/
AQ7 HyperbolicTangent.htm 520
AQ8 Wooldridge, J. (2009). Introductory econometrics: A modern approach (5th ed.). OH: South-Western: Cengage Learning.

View publication stats