First of all apologies to my regular readers as there hasn't been a lot of blogging activity during the past two weeks. The reason for that is primarily my busy schedule as the academic year has started, in addition to a few papers I have in the review process, but mostly it was the media attention that has caught up with me lately.
The reason for that was a research paper I did for the Croatian Banking Association on the topic of corruption in public procurement in Croatia and the relationship between local corruption and re-election (the research paper will be available soon on their website, and I'll post it on the blog as it happens; note: the paper is in Croatian - I do intend to translate it though).
Almost all domestic media outlets (even the local ones) covered the story. After the main results of the study were presented in the monthly financial/business magazine Banka (where I tend to have more or less regular columns), all the major newspapers (Jutarnji list - the biggest domestic newspaper where the story made the cover last week, Vecernji list, Novi list, Slobodna Dalmacija, Poslovni dnevnik), the most popular internet news portals (index, tportal), and even the national television (HRT) reported the story. Apparently corruption in Croatia is always a hot topic.
The theory behind the idea
So what was the research all about? It was an empirical verification of a part of a model from my earlier paper Persistent electoral success with endogenous rents, where I theoretically made the case for long-lasting politicians in power. In addition I used the main findings of the selectorate theory by Bueno De Mesquita, Smith, Siverson and Morrow, to clarify my argument and turn it into a testable hypothesis. The basic idea is that politicians tend to create a stable, small coalition of supporters which enable them easy electoral victories and leave them virtually unaccountable to the general public. In that case the politicians in power no longer use public goods to satisfy the voters that brought them to power, but private goods they distribute to their cronies (coalition members) that keep them in power and who are simultaneously dependent of the politician for their continued rent-seeking. Even though this scenario more likely resembles a dictatorship rather than a democracy (where the small winning coalition of the dictator are usually the army generals), on a local level in democracies politicians can stay in power for decades without ever having any real accountability towards their voters. Bueno De Mesquita and Smith in the popular version of their argument (the book "The Dictator's Handbook") mention numerous examples of both autocracies and democracies where as soon as accountability and scrutiny are missing politicians get complacent and turn dishonest, regardless of the institutional setting surrounding them. This can happen in the United States (recall gerrymandering) as well as in any African dictatorship. After all the logic of political survival is always the same: stay in power!
The methodology and results
Using Croatian data on local level public procurement I was able to generate a testable hypothesis out of the theory, as I created eight proxy measures for corruption based on identifying suspicious and fraudulent procurement cases. I linked them across Croatia's local municipalities and cities and designed a measure of corruption as a ratio of suspicious procurement cases over total procurement cases. The term 'suspicious' was defined in many categories (companies with no employees getting vast sums of money in legal tenders, companies with profits from public procurement 10 times as much as their annual revenues, companies in losses getting indirectly financed by the government, etc.).
The source of data are official government reports on procurement cases and companies involved in them. However they weren't openly available on any of the official government institutions, but had to be collected by independent activists aiming to increase transparency of the public procurement process (some of them received funding from the EU to support this cause). The years observed in the paper were 2009 and 2010, and the sample contained data on around 4600 cases across 300 municipalities and cities. This amounted to a total of 26bn kunas being observed (around $5bn), which is a quarter of all public procurement being done in those two years. There was a problem in creating the sample as a lot of the data wasn't suitable for linking the company with the procurement case they were involved in, since in many cases the key numbers (the financial ID of the company that needs to be reported in the procurement) were missing. This could imply that the sample actually underestimated corruption in public procurement, as the missing numbers could have been removed deliberately.
Even in that case the results were outstanding. They even surprised me. I managed to show that the higher the level of corruption (the proxy for corruption at least), the higher the probability the politician gets re-elected (see figure below). However only up until a certain point where further corruption decreases the probability of winning. From eight measures of corruption that I made, six of them confirmed my hypothesis (the other two showed the same direction but weren't statistically significant).
Basically this implies that voters do punish corruption but only if its exaggerated. According to the graph above (fitted values predicted from a logit model), after around a quarter of public procurement allocated in a suspicious way, the probability of re-election goes down. The politician gets thrown out of office only when he/she allocates more than half of the public funds in a suspicious manner (other things held constant). This suggests that the tolerance level the voters have for corruption is still way too high.
Being able to calculate precisely how much a politician needs to steal in order to stay in power, I was risking of being accused to having provided the politicians a framework upon which they may operate. However they knew this without my research - after all that's how they managed to stay in power for all these years.
Nonetheless I needed to make a series of normative implications and reform proposals aimed at reducing the scope of local corruption in Croatia. My first immediate reform was to introduce a term limit for local politicians (a maximum of two terms in office, i.e. 8 years). This way we can immediately get rid of all those politicians who are in power for over a decade and have created a corrupt, crony system centered around them. Simultaneously with the term limit electoral rule, the number of local municipalities has to be reduced, or more precisely we need to increase the number of voters necessary for electing the politician into office. This goes directly in line with the selectorate theory and increasing the minimum winning coalition. In many local municipalities in Croatia it takes a few hundred voters to get re-elected. By increasing the size of the municipality, we put the politicians in a completely new environment thus destabilizing their currently held concessions.
However implementing the two reforms won't necessarily solve the problem. Eight years can be more than enough to build a corrupt system and a cronyist network of supporters (after all the term limit political economy literature suggests that when politicians are facing their final term in office they tend to steal more; see Besley, 2006 and Ferraz and Finan, 2011), while operating in a new environment is only a temporary setback to which the politicians can quickly adjust. This is why a third reform is crucial, and should also be applied simultaneously. In the next phase of the research I intend to construct an index of local corruption in Croatia ranking the cities and municipalities from the most corrupt to the least corrupt. This way it could be easy to adjust the flow of funds from the central government budget to the local budgets on the basis of their corruption rankings. Those on the bottom of the list get no money from the central budget (these funds contribute to around a fifth of the local budget). This will then send a clear signal to the voters that because of their local politician being corrupt they couldn't afford to solve some of the local problems they tend to encounter on a daily basis (garbage disposal, schools, kindergartens, hospitals, basic infrastructure, etc.). By having this information the voters will react and vote the politician out of office.
All three reforms must be applied simultaneously. Exposing the corrupt practices of current local politicians may not yield them out of office as they still have a strong support group backing them in their municipality. However after throwing them out via a term limit, the new politicians won't have this luxury. They will immediately be prevented from engaging into similar corrupt actions. Such reforms, if implemented correctly, will become irreversible as the voters would possess more knowledge and demand more accountability and transparency. It will be impossible to remove this right from them in a democratic society.