Wednesday 31 July 2019

5 Fundamental Misperceptions about Transparency International’s Country Corruption “Rankings”

Plenty of Cake to Go Around: Eat Your Fill, Sleep Well

Probably the most well-known source for “rankings” of country corruption is Transparency International.
Well-known as the source, but less so for the contents.
AA believes that most people who use or quote TI’s rankings do not know what they mean and operate with some or all of the following misperceptions.
To be very clear upfront, this post is not arguing that we should not use TI’s CPI.  But rather than we should understand what it is, what are its limitations, and how to use it intelligently.
FIVE COMMON MISPERCEPTIONS ABOUT TI’S CPI
  1. TI’s rankings assess the overall level of corruption in a country.
  2. While not “facts”, the analytic process behind the rankings results in fairly accurate assessments.
  3. TI performs the analysis behind the rankings or at the very least directs it.
  4. Every country is rated using the same common set of standards.
  5. The rankings are sufficiently precise that we can use them to distinguish the level of corruption in one country from the level in another.
TI provides extensive disclosure about the CPI at its Methodologies page.
Those who read this material carefully will not hold any of the first four misconceptions.  The problem is it appears that TI’s disclosures are infrequently read.
TI’s apparently precise ranking system does give the impression that Misconception #5 is correct.  It is not.
MISPERCEPTION #1 – Overall Level of Corruption in a Country
Here’s a quote from a TI FAQ that on its rankings:
“Is the country/territory with the lowest score the world's most corrupt nation?  No. The CPI is an indicator of perceptions  public sector corruption, i.e. administrative and political corruption. It is not a verdict on the levels of corruption of entire nations or societies, or of their policies, or the activities of their private sector.”
What does TI rank then?  What is its definition of “corruption”?
Why should we care?
It’s very important to understand TI’s focus if one is to use their rankings intelligently.
If you read the FAQs in the Methodologies material (page 2), you will find a list of what is included and what is not.
Money laundering, IFFs, informal markets, the private sector are NOT included.
Broadly speaking, TI’s CPI focuses on the public sector only.
TI is very clear on this but AA wonders how many users of TI’s CPI understand this.
What this means then is that a private sector member’s actions do not affect the ranking of its respective country.
This is very important because if one is using TI rankings to construct assessments of money laundering and terrorism finance, one might be mis-specifying the risk, if one assumes that TI rankings assess the overall level of corruption in a country.
Why?
Private sector enterprises are probably the major channels through which ML and TF take place in most jurisdictions.
MISPERCEPTION #2 – Rankings as “Facts”
TI’s annual ranking for 2018 is here.
The first thing to note is that this is described as the “Corruption Perceptions Index”.
The key word here is “perceptions”.    “Opinions” not “facts”.
That makes sense.
There are no formal reports filed on bribes paid or bribes accepted.
One has to infer the extent of corruption in a country from very limited hard data – corruption cases that have come to light—and other indirect indicators.
The first takeaway then is that a ranking for a specific country is an estimate. 

Likely a very rough estimate.
Similar to the 2% to 5% of global GDP (usually mis-stated as amounts from USD 800 billion to USD 2 trillion) estimate bandied about as the annual flows of money laundering, corruption rankings are often treated as scientific fact.  They are not and should not be treated as such. 
MISPERCEPTION #3 - The rankings are based on TI’s research.
TI uses the published assessments of 13 sources.
Each of these sources prepares reports for its own or its clients’ use using its own criteria and methodology.
TI does not do the research itself. It does not set the focus, criteria or methodology for these sources’ studies.
Rather TI repurposes the 13 sources’ reports to create the CPI.  In 2015, one source, IHS Global, stopped providing data to TI.  TI now accesses some IHS data via information published by the World Bank.
MISPERCEPTION #4 – Common Standards and Methodologies
Who are the experts? What are their methodologies?
For a detailed answer click on “Methodologies”. Here you will find a discussion about each expert and its methodology.
Click here to see the sources used in ranking a specific country.
The first thing you will notice is that not every source rates every country.
In a situation where some countries are rated by some experts and other countries are rated by other experts should we automatically assume that all the experts use an identical single common standard and methodology?

Clearly we need to look a bit deeper because if the experts don't have a single common standard, then which experts rate a country will impact that country's rating.
AA has read this material and encourages everyone who uses TI’s CPI to read it as well.
Why?
First, this is quite a heterogeneous group.
It includes multi-lateral institutions (2), NGOs/Foundations (5), companies selling country risk or business information services (4), university affiliated entities (2).
Each of these has a specific purpose for its study motivated by its stated “mission” or, in some cases, perhaps by its ideology.
That is not meant as a pejorative remark. But as a practical one.  We need to be sensitive to conscious and unconscious factors that may influence a rating, particularly in the case where “perceptions” play a key role in determining rankings.
AA argued in another post that the collapse of Abraaj seemed to be treated in some circles as evidencing a more serious failure by regulators and markets than scandals in certain OECD countries that had a much greater impact on the world economydid.
Are there other geographical biases? Is corruption in African Country G more heinous than Baltic Country L?
Without taking a stand on the issue, AA would note that there is some controversy about the independence of Freedom House from US foreign policy. The FH study that TI uses rates former Soviet bloc states.
Second, the experts’ focus is also heterogeneous.
Not all of these sources focus on corruption itself: bribes paid, bribes taken.
Rather a number of them focus on legal/institutional capacity.  Whether the country has an adequate framework to prevent/punish corruption, e.g., legislation, staffing and independence of investigative and legal bodies, administrative practices, e.g., professional independent civil service, open bidding, whether information is available to the public, etc. 
These indicators by themselves are not indicators of corruption but rather perhaps indicators of opportunities for corruption.
Very big difference.
Laws and frameworks are fine but as experience shows repeatedly they do not prevent crime from occurring.
That’s not to say that these elements aren’t important.
They are necessary but not sufficient elements.
The question is how much weight they should be given when assigning corruption perceptions to a particular country.
AA would be in the camp where actual corruption rather than opportunities for corruption would be given more weight in “rankings”.
Third, the experts’ methods are not identical.  Some use in-house experts to make assessments.  Others reach out to local contacts, and other outside experts, e.g., academics, lawyers, accountants, etc.  In some cases like EIU they use in-country free-lancers at least in part.
Some of the experts appear to ask a single or a couple of questions as part of a larger study on more than just corruption.
Others have a more robust set of questions on corruption.  Or survey a wider set of contacts.
For example, in 2018 The World Economic Forum Executive Opinion Survey (WEF-EOS)--one of TI’s sources—received 12,274 responses from executives in 140 countries in 2018 about corruption.
Fourth, some of the experts—primarily the 3 firms that sell political risk and country assessments to businesses -- assess all levels of corruption from the petty to “grand” corruption.  Varieties of Democracy, another of TI's expert sources does as well. 
As a practical matter, their 3 firms' clients (businesses) are likely to be most interested in the need to pay ongoing bribes to ensure their daily operations run unhindered if they invest in Country X.
So smaller recurring cash payments to facilitate clearance through customs of imports and exports, to secure connection to and maintenance of utilities, to deal with tax authorities, to obtain licenses, etc. are of prime concern.
Finding out about them is fairly easy.  One can ask businesses in the country. They will be more likely to report such occurrences because they are imposed on them as opposed to grand corruption where they may be a willing participant.
Because it’s harder to find out the true level of grand corruption, there is a risk that corruption ratings based on petty or moderate corruption may skew the rating for a country.
Fifth, unlike the countries in the CPI, the 13 experts are unranked.  Their perceptions are accorded equal weighting.  Each expert’s score is added and a simple arithmetic mean is calculated.
They are all presumed to be all equally smart and informed and use equally valid methods to evaluate corruption.  It doesn’t matter whether an expert asked a single question or sent a questionnaire and got 12,274 responses.
It doesn’t matter if the expert is expert in a limited geographical area or covers the world.  The Economist Intelligence Unit who use in-country free lancers in part to do their assessments and rated 131 countries in 2018 are presumed to know as much about each of those countries as the African Development Bank which uses in-house economists knows about the 54 African countries it rated. Or PERC which contacts a wide range of potential respondents to ask a single question and rated 15 Asian countries.
As you might expect, not every country is rated by all 13 experts.  Some of this is because of geographical specialty. The experts from the African Development Bank don’t rate Switzerland, the USA, or France. PERC’s focus is a slice of Asia.
It’s not unreasonable to say then that the rating standards across all countries are not uniform given the diversity of focus, methodology, level of detail, etc. of the 13 experts and the fact that the same 13 experts do not rate each country.
The full data set shows the score, the standard error (think standard deviation but for a sample), the Upper CI and Lower CI.
There is a wealth of information here.  If you use the TI CPI, then you should be familiar with this information so you can use it intelligently.
For example, should we treat a rating with only 3 experts (the minimum required for a rating) as being as valid as one with 10?
If the standard error is large, should we assess that the rating is less accurate than one which has a smaller standard error?  For example, the SE for Switzerland is 1.57, Bahrain and the Philippines are at 1.81, Saudi is at 6.34, Qatar at 8.08, and Oman at 9.46.
MISPERCEPTION #5 - Ratings are Precise Measures
TI ranks some 180 countries.  100 is the theoretical “best” score.  0 the worst.
Denmark in the first rank with 88.
New Zealand is at 87.
Then four countries follow at 85.
All the way down to Syria (13) and Somalia (10).
This is some very precise parsing of differences in corruption.
Let’s stop and reflect for a moment.
We started with “perceptions” but we seem to have wound up with “precision”.  AA would argue “false” precision.
On a hundred point scale, NZ would appear to be 1% more corrupt than Denmark.
Can we really parse gradations this fine?
More importantly is there really a practical difference in corruption between Denmark (ranking #1 with 88) and Germany (ranking #11 with a score of 80)?
The answer to both questions is no
TI agrees with this at least in part.
In their FAQs, they answer a hypothetical question from a reader about changes of 1 or 2 points in a specific country’s rating year-on-year with:
“It is unlikely that a one or two point CPI score change would be statistically significant.”
AA would argue that even larger differences among countries are not significant either.
Let’s look at an endeavor that has more data and more rigorous mathematical analysis of the data, though one which is not devoid of opinion:  credit ratings. 
S&P, Moody’s, and Fitch rank issuers.
But they don’t assign them individual ranked ratings.  Rather they group them into categories of similar risk.
Those issuers least likely to default are rated (placed in category) AAA.  If distinctions are made, a “+” or “–“sign is used.
AA doesn’t think it’s a sensible proposition that corruption analysis is more scientific than credit analysis and hopes you do too.
AA suggests that TI adopt a similar approach in an effort to prevent misunderstanding and misuse of its rankings.  That is, divide countries into broad categories of risk of corruption like credit ratings or S&P's BICRA.
This will have the immediate effect of preventing users from plugging the current “precise” ratings into their models and coming up with equally imprecise results in theirs.
Some even more impressive with results to two digits to the right of the decimal point, though admittedly not on a 100 point scale.

No comments: