Corruption “Tells” — An Overlooked Factor in Determining Corruption Perceptions

Last month, the European Commission released a comprehensive report on corruption in the EU, based on two perception surveys (one of the general population and one of businesspeople) as well as existing public data. One of the report’s most striking findings was the prevalence of perceived corruption among the general public: over 75% of Europeans surveyed thought corruption was “widespread” in their country–even in countries where very few respondents had personally experienced or witnessed corruption.

The EU Report is not the first study to find a sizeable gap between people’s perception of corruption’s prevalence and their reported personal experience with corruption.  What explains this gap?  The two most common explanations are: (1) perceptions of corruption overstate true corruption (as perceptions may be swayed by sensationalistic media reports, and perhaps skewed by factors like ethnic heterogeneity and low social engagement, or because of different understandings of what “corruption” means); (2) self-reported experiences with corruption understate true corruption, because people do not respond truthfully to questions about their personal experience even when anonymity is guaranteed.

But there is another possibility, which highlights a limitation of studies that compare only general perceptions of corruption with direct, personal experience with corruption: These surveys typically fail to account for “tells” – observable indications of potential corruption.

Tells are the day-to-day visual images or interactions that drive peoples’ perceptions of corruption: a luxury car driven by a public official, absenteeism, bottles of high-end cognac on store shelves, a corporate social responsibility project in the neighboring town, or even an oil company’s promotional calendar sitting on a politician’s desk. They’re little signs of graft that lead to inferences of corruption. Today, most corruption indices get at perception by asking questions about (1) personal experiences with actual corrupt acts and/or (2) the impact of mass media. But they don’t ask about observable environmental cues.

I don’t mean to imply corruption “tells” completely explain the gap between experience and perception–just that they may be playing some role, which most existing corruption surveys don’t get at. It might therefore be helpful to expand corruption surveys to include follow-up questions that encourage participants to articulate what “tells”, if any, are driving their perception. For example, the EU Report asked respondents how widespread they think corruption is in their country–very widespread, fairly widespread, fairly rare, very rare, no corruption, or don’t know. For those respondents who reply “very widespread” – as many in the EU survey did – the survey could and should have asked the follow-up question: “What observations or interactions led you to perceive corruption as very widespread?”

To be sure, it’s not entirely clear what re-structured surveys would show. Knowing tells could make survey results more accurate by reflecting overlooked metrics of corruption and teasing out some cultural aspects of perception. On the other hand, they could make results less accurate in that they are anecdotal and centered solely on the perceiver’s experience. After all, tells can mislead: the politician with the promotional calendar on his desk is not necessarily more corrupt than the politician with an unadorned desk. That said, a better understanding of corruption’s tells would add important nuance to corruption survey results, and make country-to-country comparisons like the EU report’s more meaningful.

9 thoughts on “Corruption “Tells” — An Overlooked Factor in Determining Corruption Perceptions

  1. Addar,

    This is a very nice observation. Discussions about corruption perceptions sometimes do assume that perceptions must either be based on direct personal experience (like being asked for a bribe) or very general sources like media reports. But people might be able to directly observe things that indicate corruption is occuring, even if the people themselves are not directly involved. And the idea of getting more information about where, exaclty, perceptions are coming from seems sound (though perhaps practically challenging). I did have a few questions, though:

    First, as you note, the “tells” may be accurate or inaccurate. So suppose we got the information you’d like to be gathered. People mention a bunch of things, like officials with fancy watches, or that some people seem to be jumping the queue at the government office, or what have you. What do we do with this information? If we had that info for the EU study that you mention, what would we do with it? Would it lead us to adjust our assessments of “true” corruption up or down? I’d love to know more about what you have in mind in terms of what kind of informaiton we might get, and how we might then proceed.

    Second, I wonder about how well people can identify the “tells” that they’re actually using. Your proposed survey questions, while surely worth asking, do require that people are sufficiently self-conscious about how they’re drawing inferences about the extent of corruption.

    Third, you frame your comment by noting the very high percentage of EU citizens who think corruption is widespread (even though very few have direct experience with corruption). Is it your sense that they’re overestimating corruption because they’re relying on misleading “tells”? Or that corruption is actually much more widespread than many experts think, and these survey results indicate that EU citizens, using relatively accurate “tells”, are picking up on this widespread corruption?

    • I agree, this was an interesting perspective; there have been a few posts about perceptions on GAB so far, but I especially like the way that this one addresses some of the potential biases inherent in the very method of aggregating perceptions.
      This may be a little off topic, but I particularly like the distinction drawn by Abramo between richer and poorer countries, in this regard. In the EU, my feeling is that the “actual” prevalence of corruption is probably closer to the lower level suggested by personal experiences, rather than the higher one based on perceptions of systemic corruption. The systemic perception may be biased higher by some of the factors you cited, or perhaps just by flawed reasoning based on an instinctive mistrust of those in power – I’ve always found this type of study to be interesting, from a decision research perspective (http://www.gallup.com/poll/142658/americans-views-public-schools-far-worse-parents.aspx). In poorer countries, though, it would make sense that so-called “grand corruption” is more likely to be hidden from the average person, and that their personal experience therefore understates the true prevalence of corruption. In this context, “tells” would actually help to equalize the disparities between these two types of indicators, I think.

    • Michael, I’m curious what you’re defining as the “tells” in this index? Do you mean, being a billionaire is itself a tell?

      One difference with this crony-capitalism index is that I don’t think it’s based on public perceptions at all, although certainly being a billionaire in a rent-heavy industry (which I think is what The Economist was measuring) might cause people to assume corruption.

      • My interpretation: The Economist, instead of using perceptions as the basis of its index (e.g., perceptions of crony-ism among business people, or the public), itself designated a “tell”–the number of billionaires in particular industries. I took the index to suggest that a high percentage of billionaire wealth in crony industries signals corruption. This practice would be consistent with Addar’s suggestion to focus on explicitly measuring things that make people think there’s corruption, instead of only measuring the perceptions themselves. (Putting aside whether The Economist’s chosen “tells” are accurate/relevant or not.)

        • I think you are both right, although Eden’s interpretation
          is probably closer to how I originally conceived of “tells.”

          Certainly, physically seeing billionaires in a crony industry could trigger peoples’ inferences of corruption. In that way, they are “tells.”

          But, as you point out Michael, the crony index doesn’t actually measure perception but rather, actual levels of corruption. My original point was that tells are a useful guiding question in perception surveys because they capture something important about the way people form perceptions. Thanks for helping me make that point more clearly.

  2. Matthew,

    These are all great questions. I’ll try to respond to them in turn:

    In regards to your first point, I’m actually very curious to see what “tells” people would point to. Part of me wonders if the question itself could cause people to alter their response as to how corrupt they perceive their country to be. Once the information is collected, I can imagine three main areas where “tell” results add value. First, I think there is a purely expressive value to publishing the results that is independent of any pragmatic use. This type of information adds depth to the concept of “perception” and could potentially shed a more focused light on problem areas. It’s worth noting that publishing these results could also be useful information for perpetrators of corruption and others who care about keeping corruption out of the limelight.

    The second use for results is in improving analysis of survey results. Let’s assume we find out that the French see absenteeism as a significant “tell,” but Nigerians don’t. Let’s also assume that absenteeism is at about the same rate in both countries. Having this data could help begin the process of teasing out culture- and country-specific differences in corruption perception. I think that using “tells” to adjust corruption ratings is a different, more complicated proposition. But one can imagine a scenario in which many people note lobbying efforts as their “tell” and surveyors who don’t want to adopt that broad a definition of corruption could adjust the totals downward. The same goes for a situation where many people note a “tell” that likely has a more innocuous explanation (ie. a large CSR project or an unsubstantiated rumor of political scandal).

    Lastly, this information could be useful in the policy arena. If a statistically significant population does note absenteeism as a “tell,” governments or NGO’s who care about corruption (or at least perceptions of corruption) could better target their work. If, on the other hand, many people note a large CSR project in the region as a “tell,” community education about CSR and increased transparency in the CSR process would be a better prescription. If the data was fine-grained enough, the information could even help investigatory arms of anti-corruption agencies hone in on targets.

    Re: your second point, I agree that this type of information is difficult to gather, although the same argument can be made about “perceptions of corruption.” It’s a topic that requires a certain amount of self-reflection, and the question about “tells” can be seen as a helpful guide to the reflection. As I note above, because “tells” require concrete observations (rather than completely subjective qualifications of perception), just thinking about how to answer the “tell” question could have a positive impact on the “perception” question. Giving a few carefully chosen, concrete examples of tells in the survey instrument could also be helpful.

    Finally, regarding the results of the EU survey, my gut sense (and it’s nothing more than that) is that two things are going on. First, the average European is relying on a broader definition of corruption than the survey is capturing. Reading articles about special interests, continually voting for a losing political party, or personal experience with tax evasion: these are the “tells” of the developed world that may not fall within the Commission’s more traditional definition of corruption. Does this mean Europeans are overestimating corruption or that the survey is under-capturing it? You decide. Second, I think perception is all relative, and that the average European is probably using the Europe of his or her childhood as a (potentially rose-colored) baseline. This suggests inflated results although, assuming the EC continues to administer the survey in its current form, it will be interesting to compare this year’s data to future years in order to see if the 76% figure is actually a useful baseline or more of an aberration.

    • Addar,

      Thanks for such a rich and thoughtful reply. I think I’m convinced that it would be really useful for those conducting corruption surveys to pursue your suggestion, at least on a trial basis. One of the practical difficulties is that usually, due to cost and administrability constraints, these large-scale surveys ask multiple choice questions, rather than open-ended questions; but the “tells” you describe don’t lend themselves as easily to that format. Perhaps it would be possible to conduct targeted trial surveys with more open-ended questions, as a way of identifying perception sources that could then be addressed in standardized questions in larger surveys.

      On the definition of corruption issue, one thing that researchers sometimes do is to use so-called “anchoring vignettes” — the survey will describe a series of hypothetical scenarios and then ask the respondent to assess them (in this case, the vignettes might describe a range of interactions between a government official and private party, and the respondents could be asked whether the interaction is “corrupt”, or how corrupt it is). This techniques allows the researchers to get a sense of how people are using/interpreting terms, and facilitates cross-country comparisons.

      Finally, an aside on absenteeism: I’m not sure whether this is merely a corruption “tell”, or rather a form of corruption itself — if, on a consistent basis, schoolteachers aren’t in school, or health workers aren’t at the clinic, that’s effectively a form of embezzlement of government resources from social welfare programs. And absenteeism is something that people have in fact studied quantitatively – but generally not via survey instruments, which typically ask only about direct experience with bribery, not about direct experience with these other forms of corruption.

  3. Phil, I’m so glad you brought up the rich/poor country divide in corruption perception! And I agree with you that the prevalence of corruption in the EU (at least as defined by the survey) is probably closer to the levels suggested by personal experience. At the same time, it seems like your points regarding the openness of corruption and mistrust of government could easily cut the other way.

    The robust civil society and rule of law enjoyed by many countries in the developed world could mean that corruption perpetrated there would actually have to be more well-hidden than in the developing world. And wouldn’t growing up in a country where you have no paved roads and no electricity but your parliamentarian drives a Range Rover lead to more mistrust of government, not less?

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s