Show Summary Details

8 How do Countries Shift from High to Low Corruption?free

  • Ray Fisman
  •  and Miriam A. Golden

The world’s wealthy democracies all have relatively honest governments. However, that wasn’t true a hundred or two hundred years ago, when they looked like governments in today’s poor countries. How did they do it? How do countries escape a high-corruption equilibrium?

In this chapter, we look at the three ways that such shifts occur: when voters rise up to demand change; when external actors intervene in the political system and impose change; and when political leaders are motivated to enact change themselves. The three are not mutually exclusive. As we’ll see, voters may rise up when the external environment changes to shake up the status quo. But we can tell separate stories about each type of change. (In the next chapter, we turn to specific policies that may help the process along or lead to more incremental improvements. In this chapter, we provide a more descriptive—rather than prescriptive—account of how countries change.)

We focus primarily on the role of voters, beginning with an exploration of the various reasons why they so often reelect corrupt politicians. We then provide details of a rare instance in which voters managed to coordinate their support behind new politicians, thereby ejecting the old and corrupt ones from office. This case study of the extraordinary collapse of Italy’s corrupt political parties in the early 1990s provides a morep. 204 general set of lessons for the role voters can play in fighting corruption in a democracy.

  • Top of page

8.1 Why do voters reelect corrupt politicians?

There’s a misconception that voters in corrupt countries are happy with the status quo. If they weren’t, wouldn’t they elect different politicians?

That view is almost surely wrong, as survey evidence that we reported in chapter 5 has already suggested. Around the world, most people think that corruption is a big problem.

Yet politicians implicated in political malfeasance nonetheless get elected and then reelected.1 In Italy, in a period spanning more than four decades, voters (97 percent of whom, keep in mind, told Eurobarometer surveyors in 2014 that giving and taking bribes is never acceptable) reelected hundreds and hundreds of legislators who were under investigation for illegal dealings. This continued until the party system collapsed under the weight of political corruption in 1994. (We’ll have more to say on this later in the chapter.) In India, candidates under indictment for criminal wrongdoing are substantially more likely to be elected to state and federal legislative bodies than their unindicted counterparts.

In Japan, a low-corruption country (based on Transparency International’s Corruption Perceptions Index [CPI]), legislators convicted of corruption have also been rewarded with additional votes. How else can one explain the fact that 62 percent of legislators convicted of corruption between 1947 and 1993 were reelected?2 Take the case of Kakuei Tanaka, who eventually rose to the post of prime minister. Early in his career, Tanaka had already been caught taking bribes from coal-mining interests, which landed him in prison in 1947. He nonetheless rose through the ranks of Japan’s dominant Liberal Democratic Party, holding several cabinet positions and then assuming the office of prime minister in 1972. He was forced to resign just two years later, when he was implicated in questionable landp. 205 dealings. In 1976, while a sitting member of parliament, Tanaka was accused of having taken US$1.8 million in bribes during his time as prime minister, in exchange for securing a contract with the country’s airline. Following his criminal conviction, Tanaka filed an appeal and was reelected to parliament in 1983 with an unprecedented margin. Then, perhaps just to thumb his nose at those responsible for the bribery revelations, he had himself appointed to the parliament’s ethics committee.

Italian, Indian, and Japanese voters are not alone in their willingness to support corrupt candidates. U.S. Congressman William Jefferson— infamous for the brick of US$90,000 in cash that the FBI seized from his freezer in 2005—was similarly reelected just a year later. In fact, Washington lawmakers are regularly reelected amid corruption investigations. And we could go on—politicians involved in malfeasance are more likely than not to be reelected in every country of the world on which data have been collected.

Why, if voters don’t like corruption, don’t they throw the bums out?

One obvious explanation is that voters may have political priorities other than corruption—they may vote along partisan, caste, or ethnic lines, for instance, or they may elect corrupt officials they think are wiser or more competent than their less corrupt challengers. (Japan’s Tanaka came into the prime minister’s office in 1972 with the highest popularity rating in that country’s history, so presumably voters thought highly of him for reasons other than his experience with corruption.) Or voters may evaluate how well an incumbent performed while in office and decide to tolerate whatever corruption occurred because policy performance was nonetheless good enough. When the economy is doing well, voters may be more tolerant of the malfeasance and misdeeds of individual politicians.3

From a voter’s perspective, the sense that corrupt politicians are doing a reasonable job is reinforced by their success in directing small bribes (as well as larger chunks of government expenditures) to their constituents. Where clientelism isp. 206 common, corruption is more apt to flourish. And whatever their illicit dealings, these politicians are experienced in the workings of politics.

Below, we’ll focus on two broad classes of additional explanations—let’s call them “information” and “coordination”—for why voters might reelect corrupt politicians despite genuinely preferring honest representatives to dishonest ones. These may be less self-evident than the “different priorities” explanation above but, we would argue, are no less important and are potentially of greater relevance for reformers.

  • Top of page

8.2 Does lack of information lead voters to reelect corrupt politicians?

Voters might be unaware of corruption when casting their ballots, even when a legislator is charged with or convicted of wrongdoing. Perhaps they don’t pay attention to politics because they have other concerns to attend to (which may be compounded by citizens’ disillusionment with corrupt politicians). The average American, for example, doesn’t know which party controls the Senate or the House of Representatives, suggesting that ignorance of political affairs is all too common. Maybe too many stories of corruption get buried in the back pages of the newspaper. And much of the world doesn’t have access to daily news, even if people wanted to know what their political representatives were up to. That’s the situation for the hundreds of millions who can’t read or who have limited access to electricity—and no television to tune in to the news even if they did. Maybe it’s possible to keep up on current events via radio or, today, even by cell phone. But in many poor countries, the government controls the press, so reports of political malfeasance are not broadcast.4 Even where the government itself doesn’t own the media, it’s very often controlled by a handful of family-held companies that enjoy cozy relationships with the political regime.

p. 207 If better information were all that stood between voters’ desires and the reality of corruption-free government, then spoon-feeding them facts about legislators’ records should go some distance toward solving the problem. The evidence for this is surprisingly mixed, however.

One study, by economists Claudio Ferraz and Frederico Finan, illustrates the potential that information may have in shifting voting behavior.5 Ferraz and Finan focus on a program that aimed to expose municipal corruption in Brazil, implemented in 2003 by the government of Luiz Inácio Lula da Silva (known as Lula to the electorate). The program picked cities at random for audits by the national anticorruption agency, the Controladoria-Geral da União (CGU). The CGU took care to ensure that the selection of cities was not itself corrupted. Selection was done through a nationally televised lottery. If you’re picturing a bunch of numbered Ping-Pong balls (with each number corresponding to one of Brazil’s 5,570 municipalities) popping around in a clear plastic sphere until one drops out, you’ve got the right image in mind. Members of the press as well as everyday citizens were on hand to ensure that each municipality’s number was entered into the lottery and that no one doctored the balls to protect any municipality from scrutiny.

Once selected, the municipality received a visit from about a dozen CGU investigators charged with examining whether federally transferred funds had been used appropriately by the local government. The auditors looked primarily for fraud or suspicious procurement practices—over-invoicing on hospital or school supplies; awarding of contracts without open bidding; outright embezzlement. Ferraz and Finan highlight a few egregious examples uncovered by investigators: a company with no construction experience that was paid five times the estimated cost to build a 9-kilometer road (it simply subcontracted the job to a construction firm, netting profits of over 150 percent on the deal); expenditures exceeding US$100,000 on medicines that never arrived; and so on.

p. 208 These reports were then summarized and disseminated via the Internet and through media outlets. In 376 municipalities, the results of these audits were available about a year prior to the local elections that were held across the country in October 2004. In municipalities where the auditors found no improprieties, Ferraz and Finan found that incumbent politicians were elected at higher rates than in places where no audits had been conducted—the investigations had revealed to voters that their leaders were honest. But in places where two or more corruption cases were uncovered, the probability that incumbent mayors were reelected was 17 percent lower than in unaudited municipalities; three or more violations led to a 34 percent lower chance of reelection.

Other evidence indicates that these effects were coming from the audits and the information campaign that followed. First, the effects were stronger in towns that had local radio stations to publicize the audit results. Second, there is a natural “control” group to distinguish the role of dissemination of corruption information from the role of corruption itself. Nearly 300 audits took place in the six months immediately following the 2004 elections. Any corruption revelations coming out of these later audits didn’t become available to voters until after they had cast their ballots. For these postelection audit reports, corruption that investigators uncovered had no effect on how incumbent mayors fared at the polls, as one might expect if the audit reports provided useful information for voters.

It’s a tantalizing result for would-be reformers: information will set you free! But subsequent research on the impact of information on vote choices has produced mixed results. On one hand, there are the results of a 5,000-person survey administered in India that asked voters to evaluate hypothetical political candidates based on characteristics like party affiliation, ethnicity, and criminal history. Perhaps unsurprisingly, voters say they prefer candidates without criminal backgrounds.6 These findings line up with those of Ferraz and Finan, but relyp. 209 on hypothetical choices—they don’t necessarily tell us what voters will do in real elections.

A study of a real election in the state of Jalisco in Mexico produced different, and quite discouraging, results (we also discuss the study in chapter 4).7 There, an information campaign on incumbent corruption led to a large drop in electoral turnout without affecting the incumbent’s vote share. That is, telling voters that their elected officials were crooks apparently just confirmed suspicions that politicians were dishonest. This result generates a Catch-22—giving voters information that they need to thoughtfully cast their ballots about corrupt politicians may just make them more distrustful and apathetic, and lead them to stay away from the polls altogether. Without a clearly honest challenger to endorse, reminding voters that politicians are corrupt is simply demotivating and depressing.

We’re still a long way from reaching any conclusions based on these (and other) findings about how voters respond to information. Furthermore, much probably depends on the source, some of which are more credible to voters than others. A study of an informational campaign in Brazil found that voters were more skeptical of reports accusing a political candidate of corruption when they appeared to come from another political party than when they were attributed to a nonpartisan federal audit.8 Such concerns are well founded: given the dirty nature of modern political campaigns, who’s to know what’s a smear by the opposition and what’s an honest report of misdeeds? Perhaps the Brazilian audits studied by Ferraz and Finan were so successful because the information came from a respected and politically impartial auditing body, and because the revealed transgressions could be clearly tied to a single culpable local politician. Such considerations show that resolving voter ignorance isn’t the cure-all that advocates of the information view had hoped: who provides the information and how voters evaluate electoral alternatives also matter in subtle ways.

  • Top of page

8.3 p. 210Why do voters need to coordinate to get rid of corrupt officials?

In a democracy, any single voter is powerless to do much about corruption in casting her ballot. To eject corrupt incumbents from office, voters need to coordinate their electoral behavior around alternative—and noncorrupt—candidates.

What do we mean by coordination in this context? We’re using the term in a very specific way, to mean a particular action that an individual decides to make because of her knowledge or beliefs about the actions that others will choose (possibly complementary to her own). We introduced the term contingent behavior in the introductory chapter to this volume to describe situations where your decision about what action to take depends on what you think others will do. Here, we go a step further to consider behavior that is contingent on what you know others know. If it seems a confusing or esoteric notion, a concrete example may help. Think back to the Arab Spring of 2011. Put yourself in the shoes of a would-be Egyptian revolutionary deciding whether to go to Tahrir Square to protest against the Mubarak regime. It makes little sense to go if you’ll be the only placard-carrying protester facing down the troops. Nor should you go to Tahrir Square if other protesters are headed to Nahda Square. Your choices of whether to protest and where to go depends on what you think others will do, and that in turn depends on what you know that they know

(i.e., do they all know everyone else is going to Tahrir Square and hence will they show up there as well). During the Arab Spring, protesters, of course, didn’t leave those information flows to chance—they took to the Internet to ensure that everyone knew where and when to show up to take down the government. They used a specific information channel in order to create common knowledge to coordinate their protest activities.

In his book-length treatment of the topic,9 political scientist Michael Chwe argues that the concept of common knowledge explains many otherwise mysterious phenomena. Why, for example, are advertisers willing to pay so much more per viewerp. 211 for ads during the Super Bowl, the championship game of the American football season? It’s the most-watched show on U.S. television. So if you’re selling a product that’s valuable to a customer only if lots of others use it as well, the annual ritual of viewing the Super Bowl—commercials included—is the best way of generating common knowledge that your product may be the next big thing. Chwe gives the example of Apple’s 1984 launch of the Macintosh computer, at the time incompatible with other personal computers on the market. Buying a Mac meant forfeiting use of all the software written for other personal computers on the market. Thus, you’d only buy a Mac if you expected you could trade files with your friends and that there’d be enough sales that programmers would produce Mac-compatible games and utilities. Launching the Mac with a Super Bowl ad guaranteed that everyone who was interested knew that everyone else who might be interested also knew about the Mac. If we go to movies partly so we have something to discuss with our friends, a Super Bowl ad for the next installment of Star Wars can help us coordinate on all seeing the same film.10

Of more direct relevance to our setting, Chwe’s arguments can help to make sense of why pure information interventions may fail to change voter behavior. A flyer that arrives in the mail and decries a politician’s corruption or ineptitude is seen by only a single voter, who has no idea how many others have received the same information or how they have responded to it. A campaign message on a popular TV show—like the Super Bowl— creates common knowledge. You can be sure that any information you receive during the Super Bowl is also known to a very large community of others.

For a voter hoping to kick out a corrupt political incumbent, this type of common knowledge may be critical. That’s because there is probably something a voter needs to give up in order to switch his partisan allegiance. To keep themselves in office, corrupt politicians construct clienteles—groups of voters whose support they buy with payments in gifts orp. 212 money. Even if the bribe isn’t large—a pair of shoes; a toaster; a few dollars—it’s at least certain.

The coordination problem confronting voters depends on their country’s political institutions. In party list electoral systems, where multiple legislative representatives are elected from each constituency, throwing out one rascal may only serve to punish voters from the localities where most of his supporters had been concentrated. Voters need to somehow coordinate to eject all corrupt legislators in a constituency from office at once. Similarly, in single-member districts, throwing out a corrupt but politically experienced representative to replace him with an inexperienced newbie might only generate what is called a “punishment regime” against the constituency as a whole:11 the denial of government benefits either deliberately or simply because of the inability of the new and inexperienced representative to get anything done in the corridors of power.

How much does coordination matter in practice in shifting voter behavior? One recent study, conducted by researchers Kelly Bidwell, Katherine Casey, and Rachel Glennerster in Sierra Leone, explicitly sought to identify the importance of common knowledge. Their findings suggest that common knowledge may be important both for changing how citizens vote and disciplining politicians once elected. The research team partnered with a civil society organization to provide voters in fourteen constituencies access to candidate debates for the 2012 parliamentary elections.12 The debates were screened in two different ways: in some communities, viewing was done at large public gatherings; in others, individual voters were given electronic tablets to watch the debates in private. The difference between the two forms of delivery is subtle but meaningful. Learning in a public setting provides more than just the immediate content of the message—it also conveys the fact that everyone else has received the same information. Furthermore, a communal viewing lets each individual see how others react to the debate performances, which mayp. 213 help them to formulate their expectations of how others will respond to the new information.

Candidates who performed well in the debates received more votes in both groups, relative to a control group of communities where no arrangements were made for residents to watch the debates. So voters assessed the information they received in reasonable ways, and incorporated it into their decisions about how to vote. More strikingly, the connection between debate performance and vote share was much stronger in communities where the debate was viewed publicly than in communities where viewing occurred in private. Equally important, the heightened scrutiny affected the behavior of candidates themselves. Where the debates were public, the candidates spent more time and money on their political campaigns. This shows that they became more attentive to their constituents and more concerned with winning their votes, where they knew that voters possessed common knowledge. Even more intriguing was what Bidwell, Casey, and Glennerster found after the election took place. Elected legislators who had been randomly selected to participate in the debates spent more than twice the amount in discretionary development funds in the following year in the communities where the debates had been screened publicly. (Development funds are government monies that are available to legislators to use at their discretion specifically for projects to enhance economic development, such as public infrastructure.) In other words, generating common knowledge about a candidate’s promises during the election campaign appears to have led to greater responsiveness to voter welfare once elected.

The study highlights the promise of common knowledge in changing voter behavior—and changing the performance of politicians. Voter knowledge about candidates, it seems, matters a great deal more when it is common knowledge.13

We conclude this section with an example that clearly identifies common knowledge as a decisive factor in a major shift in societal norms. While it is unrelated to corruption, itp. 214 underscores the pivotal role that common knowledge can play in driving a rapid shift in social equilibrium. It is the story of the sudden end to the Chinese custom of foot binding at the turn of the nineteenth century. Despite being a source of pain and deformity, foot binding had been a common practice for centuries, and was essential to a young woman’s success in the marriage market. All parents might have preferred not to subject their daughters to such agonizing and dangerous treatment, but none could act alone without endangering their daughters’ marriage prospects. (At the time, remaining unmarried was not an option.) Chinese society was in a bad social equilibrium and needed to coordinate its way out of it. In his game theoretic treatment of foot binding and its demise, political scientist Gerry Mackie emphasizes the very public methods of late-nineteenth-century reformers in bringing about rapid change. Reformers advertised the fact that no society other than China engaged in the practice, indicating that coordinating on a different set of norms was possible and, indeed, normal. Reformers created natural-foot societies, where parents could publicly commit to leaving their daughters’ feet unbound. (They also ran public education campaigns, in line with a role for information as well.) As is the case for a shift in equilibrium, particularly one where once everyone has made the shift they are all better off, once change began it took place very rapidly. A practice that had persisted for a thousand years ended in a single generation.14

In summary, both theory and evidence to date point toward an important role for coordination in bringing about social and political change. Coordinated social change can be rapid and thorough, rather than incremental and piecemeal; it allows an equilibrium to “tip,” thereby escaping the pressures that push outcomes back to their initial state, making these forces push instead toward the new norm. If ordinary people are to succeed in substantially changing the status quo, it seems likely that using common knowledge and coordinating their efforts will be part of the solution. We hope that researchers will, in thep. 215 future, be able to provide reformers with more evidence and better guidance on how best to shift voters’ expectations and beliefs that others also care enough about corruption to do something about it when casting their ballots.

Case Study: How Italian voters threw out a corrupt political class

By far the most dramatic contemporary example of voters successfully retaliating against a whole class of corrupt politicians occurred in Italy in 1994. Almost the entire lower house of representatives—known as the Chamber of Deputies—was thrown out of office in the wake of a gripping corruption scandal that implicated almost the entire political elite of the country. The only established political party to survive the turmoil was (under a new name) the Italian Communist Party (its Italian acronym was PCI), which, as Italy’s long-standing party of the opposition, remained mostly untainted by corruption. This shift in the political landscape was the result of many contributing factors. We believe that common knowledge among voters of a desire for change was a particularly important one.

In this instance, the press—at the time relatively free and independent of government pressure—played a critical role in coordinating voters’ expectations that other voters were also turning against corrupt incumbents.

The voter revolt was set in motion with events that occurred in February 1992, when public prosecutors arrested Mario Chiesa, the head of the Italian Socialist Party (its Italian acronym was PSI) for the province of Milan. Chiesa was charged with accepting a modest bribe of seven million lire (about US$4,000) from the cleaning company that held the contract for a government-run old-age home, where he sat on the board. It might all have stopped there. Except the PSI, hoping to protect its reputation, tried to give Chiesa the cold shoulder. He was not pleased. After seven weeks in jail, Chiesa finally started talking to investigators about corruption among his former colleagues, and within months the judiciary hadp. 216 unraveled a complex hierarchy of bribes and kickbacks that led all the way up to members of the national government. The investigations, known as Mani Pulite, or Clean Hands, implicated—among others—more than a third of the members of the Chamber of Deputies who held seats in the 1992–1994 term as well as six former prime ministers and several thousand local politicians. Within two years, the major political parties that had governed Italy for the preceding forty years had all but vanished, in part because they lacked experienced politicians to run for office who had not been charged with malfeasance and the parties were no longer willing to nominate the same tainted cast of characters.

These were hardly the first corruption accusations to hit Italian politics. The head of the Socialist Party, Bettino Craxi, testified to the Milanese court in 1993 that he had known about bribes and kickbacks to the country’s political parties “since I was in short trousers,” and that the money was essential to the operation of election campaigns. Nor had these earlier cases escaped the notice of Italian enforcement authorities. You can see this in Figure 8.1, which shows the proportion of deputies for whom the judiciary requested a removal of parliamentary immunity, in order to proceed with criminal investigations, for each legislature since World War II. (At the time, legislators were protected from judicial inquiry unless their fellow lawmakers voted to lift their immunity.) For the ten legislatures leading up to the Chiesa affair, the fraction investigated for serious offenses fluctuated between 10 and 20 percent—hardly insubstantial. While there is a notable increase—to about 35 percent—in the Eleventh Legislature, many investigations had occurred in every postwar session of parliament before then.

Figure 8.1 Line Graphs of Percentage of Italian Deputies Investigated by the Judiciary by Type of Allegation for Eleven Postwar Legislatures (1948–1994)

Note: Opinion charges are related to libel, slander, and similar; all others are serious. Data from M.A. Golden, Datasets on charges of malfeasance, preference votes, government portfolios, and characteristics of legislators, Chamber of Deputies, Republic of Italy, Legislatures I–XI (1948–92). Available via Dataverse and retrieved from

It was thus unlikely to have been the direct effect of the increased number of investigations themselves that led to the sea change in attitudes among voters. They’d had the chance to observe hundreds of corruption cases over the preceding decades, and already understood that the political elite wasp. 217 p. 218 involved in illegal activities. Rather, the high number of immunity requests that occurred in 1992 and 1993 may have served to finally catalyze beliefs in a consensus on the need for change. That, in turn, may explain why politicians could no longer circle the wagons to protect their own. Before 1992, judicial requests to lift immunity were routinely denied by the legislature; between 1992 and 1994, even the continued reluctance of a majority of members of the legislature to permit the judiciary to investigate the accused was insufficient to keep the lid on corruption. The public outcry became overwhelming.

Until 1992, even deputies under investigation for suspected wrongdoing would generally seek reelection rather than voluntarily step away from public office. Likewise, voters tended to reelect them, as shown in Figure 8.2, which provides the fraction of deputies reelected, split by whether or not they’d been under investigation in a prior legislature. Through 1987 (when the electoral system was modified), a deputy had better-than-even odds of being reelected, a rate that was unaffected by whether he had been under investigation for corruption.

Figure 8.2 Line Graphs of Percentage of Investigated and Not Investigated Italian Deputies Reelected for Eleven Postwar Legislatures (1948–1994)

Note: Serious charges only. Data from M.A. Golden, Datasets on charges of malfeasance, preference votes, government portfolios, and characteristics of legislators, Chamber of Deputies, Republic of Italy, Legislatures I–XI (1948–92). Available via Dataverse and retrieved from

All this changed in 1994. In elections that year, not only were more lawmakers investigated, but conditional on investigation, they were far less likely to be reelected—all but 15 percent of charged deputies were thrown out of office. Even deputies unsullied by judicial investigation fared poorly at the polls, probably because the public suspected that if the legislature had not been dissolved for early elections, the judiciary would have gathered sufficient evidence to request the removal of immunity for the latter as well.15

Until then, Italian voters had tolerated corrupt politicians for decades. Why did voters suddenly rise up and voice their dissatisfaction with corrupt politicians at the voting booth?

In order to understand the confluence of factors that set these changes in motion, some background on Italian politics circa 1990 is in order. Italian politics at the time were, to a large extent, defined by global geopolitics. The Italian Communistp. 219p. 220 Party was Italy’s main opposition party. Cold War politics being what they were, this meant that many voters were reluctant to support the opposition, thereby locking themselves into voting for the parties of government. The lack of genuine political options for many voters crippled their ability to hold the governing coalition fully accountable. Fear of losing office, which often helps to discipline political parties, was largely absent in Italy. Politicians from the two parties that dominated postwar legislatures, Christian Democracy (DC) and the Italian Socialist Party, became increasingly embroiled in elaborate schemes for bribes and kickbacks in public works contracting. Initially, a lot of this illegal activity occurred in the Italian South, with the active complicity of criminal organizations such as the Mafia. But by the 1980s, kickbacks in public construction had spread northward, which put the squeeze on small construction companies seeking to win government contracts there as well.

So before we even get to coordination among Italian voters, there was a coordination problem to be confronted by Italian businessmen, whose confessions of bribe payments to politicians were essential to exposing corruption to the electorate.

The first businessmen—including Mario Chiesa himself— who were questioned by the Italian judiciary in 1992 were reluctant to incriminate themselves by ratting out government authorities; some of them sat in jail for months before ultimately cutting deals with the judges. (Italy has no habeas corpus, so businessmen could be held in jail until they agreed to cooperate.) These early deals involved charging businessmen with the lesser crime of having been extorted by politicians (concussione, in Italian) rather than the charge of having paid bribes to politicians (corruzione, in Italian).

It’s understandable why those netted in the first wave of arrests were so reluctant to confess—at that point, everyone still believed that the political status quo would be maintained, and that they’d be protected by the political parties whose members had shaken them down for bribes. Breaking the silence thatp. 221 held together this corrupt equilibrium would shut out the firm from receiving government construction contracts once the businessman got out of jail—resulting in financial ruin.

But as the months passed, belief in the stability of this corrupt equilibrium began to fray. As cases piled up, businessmen began to worry that the hold on political power by the DC and PSI might be less secure than in the past. The April 1992 elections, which occurred just at the start of the Mani Pulite investigations, marked the rise of the Northern League—a federalist, antisouthern party that appeared out of nowhere to capture 55 seats out of 630 in Italy’s lower house—and a decline in support for the Christian Democrat and Socialist parties. The election results introduced fear among bribe-paying businesses that their years of political protection might be coming to an end. The concern that surfaced in early 1992 was nothing compared to the panic that ensued among bribe-paying businessmen when, in November 1993, the Christian Democrats saw their vote share cut in half in local elections. In the intervening year and a half, Socialist Party leader Bettino Craxi had his parliamentary immunity stripped so that prosecutors could proceed against him with forty-one separate charges of corruption. No one, it seemed, was safe.

This sequence of events flipped businessmen’s incentives upside down. As more of them began spilling the beans to the investigating judges (providing, in turn, more evidence for further investigations), the more likely an equilibrium shift became. Within six months, businessmen in Milan, where the investigations were centered, became so convinced that the old way of doing business had come to an end—and a shift in equilibrium was underway—that, instead of waiting to be subpoenaed, they voluntarily banged on the doors of the courts of justice to confess to the bribes they had paid.

This kind of snowballing is typical in circumstances that require common knowledge to effect change. Consider the situation of a businessman facing the choice between confessing versus staying silent. The very first confessor expects a steepp. 222 cost to his action, given that a continuation of the corrupt status quo is likely at that point. The next confessor faces a slightly greater chance that the equilibrium will shift, and hence slightly greater incentive to confess. By the time businesses were lining up to sign their confessions, enough had preceded them that the shift had become all but inevitable. A businessman’s cost-benefit trade-off was inverted by this point—staying silent only made sense if his connections in government stayed in office, and with the cascade of confessions flooding in (as well as the change in the electoral fortunes of the governing DC), that had become increasingly unlikely. Failure to confess now would risk killing a businessman’s company, under the increasingly likely scenario that he would get ratted out by another confessor.16

The judges understood the importance of securing these early confessions—that’s why they offered sweetheart deals to the first businessmen to confess. These confessions gave the judges the information they had long been seeking, which set the stage for formal investigations into suspected corruption by seasoned politicians. This in turn set off a further component of Italy’s political transformation. Given their unprecedented scale, the press gave outsized attention to the investigations. Day after day, new confessions and new allegations filled the headlines. In-depth reporting detailed the collusive networks connecting businesses to literally thousands of elected officials. The collapse of the Soviet Union freed voters to switch their loyalties to the communist opposition (by then rechristened the Democratic Party of the Left). Voters could finally see a legitimate opportunity to dump the incumbents. Sensing an opening, new political parties sprang up, taking stronger stands against corruption, the Mafia, and organized crime, and generally against the existing political order. This further served to highlight to voters the shifts and opportunities underway.

There are many elements to this story of transformation: an independent judiciary that was free to pursue leads intop. 223 political corruption; a free press that publicized the results; the confessions of bribe-paying businesses as the costs of continued collusion rose; and finally the common awareness among voters that others were fed up, making it worthwhile to switch their own partisan allegiances. Italian voters had known for decades that corruption was omnipresent at every level of government. The existence of corruption wasn’t new news. The investigations, however, provided a common cause that voters could rally around, leading to an awareness that other voters were equally tired of political corruption and that casting their ballots accordingly could finally result in a change.

In the 1994 elections that followed the Clean Hands revelations, more than 70 percent of the deputies elected had never served in parliament (compared with the previous parliament, in which a third of deputies had already served three or more terms).17 The two largest parties that emerged were new, and by one estimate more than a third of the voters switched their partisan allegiance.18 The parties most implicated in corruption—the Socialists and the Christian Democrats—were wiped out. Change was swift, and devastating to the old political order.

Did the Clean Hands revelations cause an equilibrium shift?

The theory we have advanced to interpret the equilibrium shift in Italian politics focuses heavily on voters. We have argued that voters, who had disliked but tolerated political corruption for many decades, were presented new options at the voting booth and became aware that other voters were likely to shift their support to new parties. The press helped voters to coordinate their rejection of the political parties most implicated in long-standing corruption. When the dust settled, was Italy any less corrupt?

There is reason to be at least somewhat skeptical. In 2014, Transparency International (TI) ranked Italy as corrupt asp. 224 Senegal, Brazil, and Romania—tied with Greece, and well below the rest of Western Europe. That hardly sounds like much of an improvement. That media tycoon Silvio Berlusconi served four times as prime minister after 1994 compounds the suspicion that corruption may have remained systemic in Italy. Berlusconi was repeatedly tried for various illegal activities and eventually convicted of tax fraud in 2013, confirming both the continued prevalence of corruption and the difficulties in legal prosecution. Most of the hundreds of politicians who were implicated in the Clean Hands investigations avoided prosecution entirely. In 2004, a survey of Italian voters asked whether public corruption had risen, fallen, or remained unchanged since 1992. Forty-six percent of respondents felt it had remained unchanged, whereas only a quarter believed that it had fallen.19 So was the entire Clean Hands investigation and the voter uprising it generated in vain?

Let’s keep in mind that surveys tend to overestimate the extent of corruption (a point we investigated in chapter 3), and for similar reasons may also lead respondents to underestimate declines in corruption. Furthermore, because TI’s Corruption Perceptions Index cannot be used to compare corruption in different years (and does not have rankings that predate the Clean Hands investigations in any event), we cannot use it to gauge whether there has been a major shift in the extent of Italian corruption. So the survey data are not very useful for addressing this question.

Other evidence points to a clear reduction in—although by no means the disappearance of—the kind of corruption that the Clean Hands revelations uncovered: the networks that linked politicians, bureaucrats, and businessmen in hierarchies of kickbacks in public procurement and construction. A careful analysis by Italian political scientist Raffaele Asquer found that voters continue to reject politicians who have been exposed by the press for corrupt dealings.20 According to Asquer’s analysis, voters are as skeptical of corrupt candidates in 2013 as they were in 1994: candidates implicated in corruptionp. 225 remain highly unlikely to succeed in getting elected to national office. This supports the view that a fundamental change in expectations has occurred.

Another useful measure consists of reports to the police of crimes committed by civil servants, because these crimes typically involve the misuse of government office for personal gain and are thus direct (if imperfect) measures of bureaucratic corruption. Studies analyzing this data show substantial declines after 1993.21 Alternate measures show comparable trends: the number of corruption-related crimes investigated by the judiciary as well as the number of civil servants convicted of embezzlement both nosedived after 1994.22

One possible reason for the widespread misunderstanding of the extent to which corruption has declined is that the Italian press is now doing a much better job—it continues to provide extensive coverage of politicians charged with corruption, even when the frequency of such cases is tiny. As a result, reading Italian newspapers generates the (mis-)impression that politicians, both left and right, are involved in corruption, when in reality the numbers are small and the ability of voters to keep the corrupt out of national public office is unwavering, as Asquer’s analysis shows.

This suggests that voters in Italy have in fact successfully produced a shift in equilibrium and expectations. Despite electing Silvio Berlusconi to serve repeatedly as prime minister, voters no longer reelect standard-issue corrupt politicians. (The fact that Berlusconi is a billionaire inoculated him against the suspicion of using his office for illegal campaign fund-raising—after all, he already had all the funds he needed to run a campaign. The many accusations of illegal activities against Berlusconi generally involved conflicts of interest in how he ran his business operations, whereas the previous political elite had been heavily implicated in bribery and more direct abuse of office.) The country still has many problems—including the continuing role of organized crime, above all in the South—but this shouldn’t distract us from thep. 226 very real progress that Italian voters have made in curtailing corruption.

  • Top of page

8.4 How do external forces trigger the fight against corruption?

It is very rare for voters to turn against corrupt politicians. Even when voters individually support change, obstacles to coordinating their anticorruption efforts make it difficult for honest challengers to displace corrupt incumbents.

As a result, anticorruption transformations may be triggered by external pressures rather than emerging out of voter protest and electoral retaliation. Often, this external pressure comes from outside the domestic political system altogether.

Guatemala provides a fascinating example. In 2015, President Otto Perez Molina was forced to resign to face legal charges as part of an anticorruption investigation conducted by the International Commission against Impunity in Guatemala (CICIG), a body created by the United Nations to strengthen the country’s rule of law. The CICIG had been deliberately created with the view that only external—and therefore fully independent—investigators would be capable of exposing widespread political corruption. With members of Guatemala’s entire political class suspected of corruption and involvement in related criminal activity during the country’s earlier civil war, it would have been impossible to clean up politics from the inside. Thus, the establishment of an external investigative body was required to make anticorruption progress possible. The public was aware of widespread corruption—although, of course, ordinary citizens did not have all the details that were revealed in the course of the investigations—but was helpless to do anything until the CICIG began its work. Between 2007 and 2015, the CICIG investigated 200 cases and brought charges against approximately 200 former and current government officials, including two former presidents, a handful of ministers, police chiefs, military officers, prosecutors, and judges. Charges ofp. 227 customs fraud, racketeering, and bribery came as a result of 89,000 wiretapped phone calls, which revealed that the president had collected US$3.8 billion in bribes between May 2014 and April 2015 alone.

Guatemala still has a long way to go in overcoming entrenched corruption. Its ranking in TI’s Corruption Perceptions Index has barely budged in the decade since the CICIG’s creation (though the country has very likely been penalized for the heightened perception of corruption that, as we’ve mentioned previously, often accompanies anticorruption crackdowns).

To find an externally triggered crackdown that led to more decisive results, we need to go back a little further in history, to the British decision to clean up corruption in Hong Kong in the 1970s. At the time, Hong Kong residents had to pay “tea money” for just about any government service— tea money for an ambulance to get to the hospital, and tea money to get a glass of water or a bedpan from an orderly once there. There was a thriving trade in gambling and prostitution, protected by a deeply corrupted police force.

The British government was pressed to confront these problems after a high-ranking police officer, Peter Godber, slipped out of the colony while under investigation for accepting hundreds of thousands of dollars in bribes, to the outrage of Hong Kong’s frustrated residents. Colonial authorities responded to the Godber affair with the creation of Hong Kong’s Independent Commission Against Corruption (ICAC) in 1974. The ICAC was endowed with extraordinary powers, including the right to arrest officials suspected of corruption and bribery without a warrant. Their investigative efforts were aided by leads from Hong Kong residents, who were able to report on suspected corruption to the ICAC without fear of reprisal. As a result of these efforts, hundreds of police officers were prosecuted for corruption between 1974 and 1977. (The ICAC also worked to bring about changes in social awareness and attitudes, instituting educational campaigns in schools to increase the moral burden of participation in corrupt activities.)p. 228 The ICAC-led crackdown successfully resulted in a rapid and permanent decline in corruption—Hong Kong is considered one of the world’s least corrupt nations today.23

Even Italy’s Clean Hands investigation involved important external triggers. In addition to the obvious catalyst provided by the collapse of the Soviet Union, which freed anticommunist voters in Italy to switch their allegiance from the ruling Christian Democrats, a less noticed but arguably just as important external factor was judicial. Thanks to 1990 changes to Switzerland’s penal code enacted to prevent money laundering—which were themselves the result of prodding by international judicial authorities—Swiss officials aided the Italian justice system by providing banking records of politicians suspected of corruption. The Clean Hands judges thereby finally obtained crucial evidence from their Swiss counterparts for use in their investigations.

The potential importance of an external actor in inducing a change in equilibrium is inherent to the very concept of equilibrium: in equilibrium, no one wants to change her behavior, given the status quo, so an external jolt may be needed in order to bring about change. In both Guatemala and Hong Kong, outside authorities had the independence and credibility to intervene quickly; when they moved in on corrupt networks, the public was happy to cooperate in exposing and sanctioning corruption.

  • Top of page

8.5 How can political leadership reduce corruption?

Citizens do not always need to wait around for an external force to trigger the fight against corruption. Sometimes government simply moves on its own. This is what happened in the Republic of Georgia, which was, according to perception surveys, the most corrupt post-Soviet republic after the collapse of the Soviet Union in 1991. In addition to suffering from a rotten government, Georgia also served as a locus for organized crime, which, many believed, involved parts of the police and security forces. Popular protest against the regime’sp. 229 growing authoritarianism produced the 2003 Rose Revolution, which installed former justice minister Mikheil Saakashvili as president. Elected in part on an anticorruption platform, Saakashvili began immediately with a radical restructuring of the police. Within the first two years of his presidency, more than 60 percent of the country’s 25,000 police were dismissed. In some units, including the highway patrol, every single officer was fired, and a handful then carefully and selectively rehired. The government was determined to demonstrate its commitment to fighting corruption to a population that had grown accustomed to endemic street-level malfeasance by police officers. Saakashvili provided substantial pay increases to new members of the reorganized police force and improved recruitment, training, and performance evaluation procedures. Suspected corrupt officials elsewhere in government were aggressively prosecuted, the government was restructured (in large part with the aim of bringing the security forces under control), and various governmental responsibilities were reassigned. For instance, a new agency was created to distribute digital identity cards, which could be used in a newly consolidated government office. As a result, instead of dealing with multiple offices, each of which had discretion to extort bribes, citizens dealt with a single office that was centrally monitored for corruption.

The reforms were remarkably successful, leading to a large decline in corruption by low-level government officials, including the police. In 2000, 17 percent of a sample of residents in the country’s capital city, Tbilisi, reported having paid a bribe to a public official; the figure was only 3.8 percent for the country as a whole five years later. Georgians now trust and have confidence in their police. Reforms in other governmental institutions—business licensing and regulation, for instance—while not as dramatic as for the police, have also made the country a poster child of effective post-Soviet anti-corruption efforts. In the TI rankings, Georgia has experienced dramatic improvements, moving from 133rd least corrupt in 2003 to 68th least corrupt in 2010, and then 50th in 2014.24p. 230 Whatever reservations we may have about the validity of TI’s survey data for across-year comparisons, such a dramatic change clearly reflects something real.

Yet this new anticorruption regime proved politically unstable. Reports of human rights abuses and creeping authoritarianism fueled a shift in voter support from the government, and Saakashvili lost his bid for reelection in 2012. He subsequently left the country rather than face what he claimed were politically motivated charges of human rights violations and embezzlement. Regardless of how we parse the story, one moral is that it can be very difficult for democratic regimes to go after corruption effectively, even when they are determined to do so. Vested interests are powerful, and efforts at taking them on may be crippled—unless the regime is prepared to sacrifice civil liberties and the rule of law.

In autocratic regimes, where voters are unable to eject officials from public office even if they want to, there are also cases of leaders who take it upon themselves to clean up the government. This seems to have been the story in Singapore, when Lee Kuan Yew committed himself—and the country—to a zero-tolerance policy. Lee was one of the founders of the People’s Action Party (PAP), which first entered the electoral arena in 1957 on a platform focused on curbing corruption. He became Singapore’s first prime minister when Britain’s colony was granted self-rule in 1959. Upon assuming power, the PAP introduced a comprehensive anticorruption policy, consisting of improved enforcement, harsher penalties, and stronger investigative authority to anticorruption officials. In the 1980s, after the economy had improved, Singapore also raised salaries and improved working conditions for civil servants. The latter was intended to reduce incentives for corruption, whereas the first set of tactics—the harsher judicial penalties and greater monitoring and enforcement—was meant to reduce opportunities.25

Singapore emerged as one of the world’s least corrupt nations, despite a long history of widespread corruptionp. 231 while under British rule. In general, if an autocratic regime is truly committed to reducing corruption, it is more likely than a democracy to have the means to succeed. But as this discussion suggests, major policy shifts by autocrats are, by definition, quite arbitrary, and we have no way to know when or why an unelected leader will decide to spring a genuine anticorruption campaign on his country. Relatively enlightened leaders who decide to commit their countries to strong economic growth and low-corruption are rare; far more common are autocrats who regularly steal public funds and ship them out of the country. Overall, then, autocracies hold few lessons for anticorruption reformers: either autocrats use their authority to steal from the state, or they clean up corruption in ways that are unpredictable and likely hard to replicate in a freer democratic society.

  • Top of page

8.6 What did we learn in chapter 8?

There are two main reasons that voters reelect corrupt politicians:

because they lack information on politicians’ illicit behavior;

because they fail to coordinate their efforts to give up individual vote-buying or patronage benefits for the collective benefit of cleaner government.

As a result of the need to coordinate, common knowledge about candidates (and expectations about how others will vote as a result) is likely important to getting voters to eject corrupt politicians.

Precisely because corruption is a stable equilibrium, external forces may be needed to jump-start anticorruption reform.

Anticorruption reform can come from government leadership itself, but top-down anticorruption regimes are extremely rare and their actions implemented arbitrarily.