Is Sunlight Really a Good Disinfectant? The Equivocal Evidence on Freedom of Information Laws and Corruption

Government transparency is widely considered to be one of the most important means for combating public corruption, a sentiment nicely captured by U.S. Supreme Court Justice Louis Brandeis’s famous observation (in a somewhat different context) that “sunlight is said to be the best of disinfectants.” For this reason, many anticorruption activists lobby for the adoption of strong freedom of information (FOI) laws—laws that not only obligate the government to regularly publish certain types of information, but also to respond promptly to citizen requests for a wide range of government records and documents. The thinking is that government corruption is easier to detect when citizens, civil society organizations, and the media can scrutinize information about government operations.

I count myself firmly in the camp of those who tend to believe that FOI laws are useful anticorruption tools, especially given the strong evidence that citizen and media access to government information can indeed help reduce corruption and hold officials accountable (see, for example, here, here, and here). And because of this, I would expect the evidence to indicate that when a country (or sub-national jurisdiction) adopts a stronger FOI law, corruption should decrease afterwards. But I’ve been looking into the research on this recently, and most of the results don’t fit well with my expectations. Long story short, the (admittedly limited) quantitative empirical evidence does not find a strong correlation between the adoption of a strong FOI law and a subsequent decrease in corruption; if anything, the evidence actually seems to suggest that the adoption of a strong FOI law may be followed by an increase in (perceived or detected) corruption.

Does this mean that FOI laws are ineffective or even counterproductive? I don’t think so, for reasons I’ll lay out in a moment. But I do think it’s worthwhile—especially for those of us who are inclined to support broad FOI laws—to consider the evidence carefully and reflect a bit on what it might mean.

So, what does the evidence on the relationship between FOI laws and corruption look like?

  • One line of research looks across countries, classifying countries by the existence and/or strength of their FOI laws, and seeing how that variable relates to (perceived) corruption (typically measured by various international indexes, like the International Country Risk Guide (ICRG), Worldwide Governance Indicators (WGI), or Corruption Perceptions Index (CPI)). Some of this research simply looks at a cross-section of countries to see whether countries with (strong) FOI laws have lower perceived corruption than countries with no (or weak) FOI laws. This is a problematic approach, however, because it’s hard to isolate the causal effect of FOI laws on perceived corruption, especially since other factors (including general political culture and history) may affect both corruption and the likelihood of adopting an FOI law. And even putting that concern to one side, the results of such research are at best mixed. One study did find that countries that had an FOI law in 2002 had higher average ICRG scores over the preceding two decades (1984-1993), but the result did not hold up when using the WGI score (from 1996-2002) as an alternative measure of corruption. Other research, using different control variables, failed to find statistically significant correlations between the strength of countries’ FOI laws and their perceived corruption scores
  • Other cross-country research tries to isolate the causal effect of the adoption of FOI laws by looking across time, rather than simply across countries. This line of research identifies moments when countries adopted or strengthened their FOI laws, and assesses whether such legal changes are followed by notable changes in countries’ perceived corruption scores on various international indexes. (These studies typically use change in the ICRG corruption index to measure change in corruption perceptions over time, though they also sometimes consider the CPI, which I don’t think is a good idea for reasons I’ve discussed at length earlier. There are difficulties with the ICRG as well, though over-time comparisons may be somewhat more defensible with these measures than with the CPI.) One might have expected that the adoption or strengthening of a country’s FOI law would lead to a subsequent improvement in the country’s score on these indexes—that is, a reduction in perceived corruption—but that’s generally not what the research has found. One study, which examined 128 countries from 1984-2003, found no statistically significant overall correlation between the adoption of an FOI law and subsequent changes in perceived corruption. Moreover, when the analysis was restricted to a sub-sample of developing countries, the implementation of an FOI law had a statistically significant correlation with a subsequent increase in perceived corruption. Another study, which used slightly different samples and statistical techniques, similarly found that the adoption of an FOI law was associated with a statistically significant increase in perceived corruption, most notably in the first four years following adoption of the law.
  • This cross-country research has been complemented by some within-country research that takes advantage of variation across jurisdictions. One notable study in this vein looked at U.S. states over time, to see what happened in states that strengthened their state-level FOI laws. Here, the outcome variable of interest was not a perception index, but rather the per capita number of federal corruption convictions of state and local officials. The study found that in states that strengthened their FOI laws, federal corruption convictions of in-state officials went up—in particular, the number of convictions in the period 2-7 years after the legal change appears notably higher than the number of convictions in the period 2-4 years before the change. (It’s worth noting, though, that this result is only weakly statistically significant.)

What should we make of this? These results are certainly contrary to my expectations, and have forced me to consider more seriously the possibility that FOI laws might not be as effective an anticorruption tool as I’d previously imagined. Nonetheless, I can think at least two reasons why the above results might be entirely consistent with the hypothesis that FOI laws do, in fact, decrease corruption.

  • First, the outcome variables in these studies do not actually measure corruption. The various indexes like the ICRG measure perceived corruption, while the law enforcement data measures corruption convictions. The fact that our usual proxies for corruption are not the same as actual corruption is of course well-known. In many research contexts, that’s not too much of a problem, but here it may be, because if FOI laws work, they do so by revealing information about corruption, information that can then be used to hold corrupt officials accountable. One would expect this effect particularly in the short-term, before government officials adjust their behavior (either by refraining from improper activity, or finding ways to better conceal their activities). So, at the cross-country level, the apparent worsening of the perceived corruption score may be due not to an actual increase in corruption, but rather to more discovery of corrupt activity. Put another way, an FOI law might have cross-cutting effects—lower overall corruption but greater awareness of corruption—which might offset. With respect to the conviction data in the U.S. study noted above, we face a similar issue: an FOI law might decrease corruption (which, holding detection probability constant, would reduce corruption convictions rates), but might also increase detection (which, holding corruption levels constant, would increase corruption conviction rates). Indeed, the authors of that study interpreted their results as evidence that FOI laws are effective—these researchers assumed that the higher conviction rates mean better detection. I think they jumped to that conclusion a tad too quickly, but it’s certainly possible. Unfortunately, the research I’ve seen hasn’t come up with a good way to disentangle these cross-cutting effects, making the results difficult to interpret.
  • Second, reverse causality (“endogeneity” in the social science jargon) is a major concern here. Countries or states don’t adopt FOI laws randomly; they do so when there’s sufficient political pressure to do so. And jurisdictions may be particularly likely to adopt such laws in the wake of major corruption scandals (or due to a general sense that corruption is getting worse). If that’s right, then jurisdictions that strengthen their FOI laws will be disproportionately those where we would already expect a significant worsening in their corruption perception (or an increase in corruption convictions of their officials). The fact that that worsening doesn’t show up in the data until after the adoption of the FOI law isn’t all that surprising. It takes time for corruption scandals to show up in international indexes, and there can be a gap of several years between the revelation of a corruption scandal and a conviction. So again, we have potentially cross-cutting effects: It’s possible that a worsening corruption situation increases the likelihood of adopting a FOI law, and it’s also possible that adopting an FOI law reduces corruption, relative to what it would have been in the law’s absence. (I should note that most of the studies I referenced above acknowledge this possibility and try to address it, but—for reasons I won’t go into here—I did not find the proposed solutions convincing.)

Where does that leave us? I fear I can’t really offer a satisfying conclusion here. For the reasons just noted, I don’t think the available empirical evidence offers a convincing refutation of the hypothesis that FOI laws are effective anticorruption tools, but nor does the evidence provide strong support for that hypothesis. I remain sympathetic to the idea that FOI laws are useful measures that more countries should adopt, but I’m less sure about that than I was before I started digging into the research.

6 thoughts on “Is Sunlight Really a Good Disinfectant? The Equivocal Evidence on Freedom of Information Laws and Corruption

  1. Don’t forget that right to information laws can also deter corrupt behavior. If the fee for obtaining a visa on arrival is posted at the border crossing, the employee issuing the visa will be less likely to overcharge, for he or she will know that the visitor knows the correct fee. I don’t believe we have any reliable means for measuring how much crime is prevented by deterrent laws, and if I am correct, this would be another reason why the research fails to show a robust relationship between right to information laws and a reduction in corruption.

    In any event, I don’t think this is the best way to frame the issue when advocating that policymakers enact a law. Apriori, a right to information law has one of three effects. It either increases, decreases or has no impact on corruption. I know of no theory or conjecture suggesting a right to information law would increase corruption so there are only two choices: a reduction or no effect. Assuming the law does no harm, why not pass it? At worst it has no effect and at best it reduces corruption.

    For policymakers considering a right to information law, and the legislation’s advocates, the critical issue is harm or rather balancing any harm against the potential benefits. Those benefits would include not only a possible reduction in corruption. RTI laws force governments to write things down clearly so they can be published. What might be an oral tradition within an agency (how we handle certain requests for a license or exceptions to certain rules) must be now be written down and subject to public scrutiny. Another benefit, at least conjecturally, would be an increase in public trust in government. Hard to put a figure on such benefits, but they are real.

    The most obvious harm is financial. Responding to citizen requests for records can be expensive. An Australian study put the cost to the Australian federal government at A$ 15- 16.5 million over a three year period. Robin Bell and Helen Watchirs, Freedom Of Information: The Commonwealth Experience, Australian Journal of Public Administration 47(4): 296–311, December 1988. The U.S. General Accountability Office found agencies whose annual data it reviewed reported receiving and processing a total of about 2 million FOIA requests at an aggregate cost of over $270 million in 2001. GAO, Update on Implementation of the
    1996 Electronic Freedom of Information Act Amendments, GAO-02-493, August 2002.

    Another harm often proclaimed (and for which there is anecdotal evidence) is the decline in frankness and candor within government. If I know my advice to the prime minister will be subject to public disclosure, I will write it in a way that makes me look good no matter what would best serve the public interest. Disclosure can also skew the ability of a group to reach a welfare maximizing result. In defense of the secret deliberations that produced the U.S. Constitution, James Madison wrote: “[B]y secret discussion no man felt himself obliged to retain his opinions any longer than he was satisfied of their propriety and truth and was open to the force of argument.” Whereas, as Jon Elster notes in the chapter from which I take the Madison quote, had the deliberations been public or the transcript later published, delegates might have felt themselves bound to positions taken beforehand in public no matter how strong the counter-arguments. Elster, “Deliberation and Constitution Making,” in Deliberative Democracy, Cambridge Press, 1998.

    So advocates of right to information laws should focus not on the shaky evidence of how right to information laws in general affect corruption, but how a law can be shaped for a particular country at a particular time that minimizes the harm and maximizes its benefits. For poor countries that might mean stressing affirmative disclosure obligations — the publication of agency rules and procedures — and delaying those sections requiring bureaucrats to hunt through poorly kept records for mentions of the citizen. For all countries it might mean serious consideration of whether all meetings and all advice should always be subject to disclosure.

    • I don’t think there’s anything in this comment that I disagree with. There are good reasons to think that FOI laws will decrease at least some forms of corruption, and little reason to think that they might increase corruption. The real question, then, is how much of a positive impact do they have, relative to the costs of FOI laws, or (perhaps more significantly) relative to the other reform measures that anticorruption advocates might focus on.

      I also agree that the relevance of the available quantitative evidence that I discuss in the post is quite limited. That said, if the evidence had come out the other way–if there was solid evidence that adoption of an FOI law was followed by a notable improvement in corruption perceptions, for example–I’d probably be jumping up and down trumpeting that fact as really good evidence that my pre-existing belief in the efficacy of FOI laws had been validated. So the fact that this is not what we find in the data at least gives me some pause.

  2. Pingback: Is Sunlight Really a Good Disinfectant? The Equivocal Evidence on Freedom of Information Laws and Corruption | Anti Corruption Digest

  3. Well Matthew, my overall comment is that not unlike measuring progress on the rule of law or judicial independence, we won’t have the kind of quantitative evidence one would like to have to fully demonstrate the impact of freedom of information laws for many years to come. One must remember that both the anti-corruption and freedom of information global movements, not to mention related research, are still in their infancy.

    That said, the best empirical research I am aware of was done by NORAD in 2011: “Contextual Choices in Fighting Corruption: Lessons Learned.” Everyone should read it.

    In that well respected global study, which spanned the global experience over a 15 year period, a team of esteemed anti-corruption global experts concluded that the one reform that had the most impact was the freedom of information law, although none alone were so-called silver bullets. These experts also noted that this reform was often followed by important second generation reforms, such as those related to budget & legislative drafting transparency& income & asset disclosure.). Obviously, the latter kind of reforms and more could not successfully implemented without the former.

    So now we all know from experience and research that we should not look at a sensitive governance reform of this nature in isolation and that country context makes all the difference. We also know we can not accurately measure its full impact before it can be effectively implemented. Finally, we now know that the best global evidence is that anti-corruption success is fundamentally rooted in giving civil society the information it needs to promote open government and public accountability.

    In short, I think it’s fine to question the impact of specific reforms but let’s not discourage some of the most important enabling environment ones, like this one, because we can’t prove impact just yet. Indeed, we all need to recognize that unimpeachable quantitative research in the ever-evolving governance arena may never be entirely possible in many cases and countries.

    • Thanks very much for calling my attention to the 2011 NORAD paper, which I had not seen. You’re right that it finds a result much more consistent with my expectations–that subsequent to adopting an FOI law, countries see a significant improvement in their control-of-corruption scores. However, while I admire the study in many ways, I think your praise for its sophistication and rigor may be a tad too effusive. The only control variable included is the Human Development Index score–but there are lots of other factors that could conceivably be correlated both with adoption of an FOI law and improvements in the corruption score. Also, so far as I can tell, there are no country or year fixed effects, or even attempts to control for secular time trends in the key variables.

      As for your last paragraph, well, I don’t know. I agree in part — we shouldn’t wait until we’ve got bulletproof quantitative evidence before acting, especially when we’ve got good reasons to suspect that a particular reform is likely to be helpful. Indeed, I devoted the latter half of the post to explaining why the evidence I note does not refute the case for pushing FOI laws. At the same time, I get a bit uncomfortable when we take a bit too much of a “damn the torpedoes, full speed ahead” approach. Sometimes the things we think are really helpful turn out to be only modestly or trivially helpful, and (as Rick notes above) we always have to be mindful of the fact that our preferred reforms have costs as well as benefits. So while I agree we shouldn’t demand absolute certainty, we should still pay attention to the evidence.

      One other thing, just to clarify: FOI laws are not the only sort of pro-transparency reforms. Public audits, expenditure tracking, and various other disclosure requirements can exist independent of laws requiring the government to provide information in response to public requests. Again, I still support the latter measures too–but it’s important to clarify that the research I discuss above is limited to FOI laws, not transparency generally.

  4. Sunlight if of course a good disinfectant but whether it is sufficient to disinfect an operation theater is a moot question. Given the experience of FoI (provided under the Right to Information Act, 2005) in India, it can easily be said that it has opened another avenue for corruption- by the information commissioners appointed under the Act to enforce the implementation of the law by imposing penalties on those public servants who have failed to provide the information sought. These commissioners are in a position to fail to impose the mandated penalty by taking bribes from the respondents!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.