My last post continued thinking about outcomes and governance. I am trying to see if we can tie governance indicators to outcomes, such that we have better construct validity of these indicators.
Basic ideas to date, and questions
I argued that this needs to be done in fields where we can focus on outcomes. Child health is one such field. I argued that we need a measure that captures outcomes citizens care about. Under 5 survival rates is my example. I argued that a second dimension of this outcome, costs, is also important. It is not a great outcome if all children are kept alive but it takes all your resources.
Putting all this together, the idea is that good governance in the child health arena will be reflected in high survival rates at low costs. Such outcome indicates that governments have taken the authority given to them by citizens to set up mechanisms and structures that produce welfare value to those citizens. These mechanisms will differ between places and we are not even getting into the detail of what they are...just whether they have delivered.
I made one additional adjustment. I said that we need to control for big contextual drivers that affect the expectations citizens might have of what outcomes should be--and the realities of delivering such. One could call these relevant indicators. In the health area, we know that levels of per capita income are such a variable. So, we need to compare wealthy countries with wealthy countries, poor countries with poor countries, and so forth.
This is where I got to in the last blog. I showed that when we do this we can see how well countries like the USA are performing relative to other wealthy countries (not well) and how well countries like Pakistan are performing relative to other poor countries (well). So, we have a contextually relevant measure of outcomes that tell us something about governance.
But does it? This is the question Alan Hudson asked and which is pregnant in my mind and I am sure many others. Alan raises two dimensions of the question. First, is the measure fully contextualized? There are contextual factors that matter--other than income--that we have not controlled for. Second, is the indicator capturing variations in 'governance'? If we are not fully controlling for context, and country outcomes vary because of contextual factors, aren't we jumping the gun in saying this is about governance?
Great questions. I think they raise empirical questions about measurement and theoretical and applied questions about what governance is. I don't think I will answer them completely here, but let me start. And let me start by going back to sports...
Boxing and governance: can we learn something
I have likened the indicator of income adjusted, costed survival rates as measures of the ability countries have to box at their income weight. They ask a lightweight boxer what his record is in the lightweight division. This is a good indicator of how well that boxer performs against appropriate opposition.
Such indicator is really different to conventional governance indicators. These indicators identify a set of characteristics in the overall best performers and then say these are the basis of good governance. Think of Doing Business indicators that emphasize deregulated industry, or World Governance Indicators that give countries credit for having things like meritoractic civil service systems and decentralized governments. To me, these indicators are akin to having a measure of 'good boxing' that looks at the weight, height and arm reach of a fighter.
This kind of indicator might suggest that good boxers are all over 200 pounds, 6 foot 2 and with an arm reach of 200 cm. Any boxer that is less than 150 pounds, 5 foot 8 or with a reach less than 180 cm must be bad. Certainly the latter boxer would not succeed against the former, in what the industry calls a 'mis-match'. Like the one shown below.
It is not useful to say that taller, heavier men with longer arm reach will win bouts against significantly shorter, lighter men with less-long arms. I've had some of those scraps on the rugby field and I know they don't work out in my favor! The indicator does not capture how good the fighter is, just how different their basic attributes are--given some intervention in their past (like their birth).
Professional boxing knows this. So it controls for the most defining attribute--weight level--to try and give a better perspective on who is good and who is bad at boxing. Every boxer falls into a weight division and must compete with others in that division. Light middleweights are those weighing between 147 pounds and 154 pounds, for example. They tend to be between 5 foot 6 and 5 foot 11, and their arm reach is between 160 and 190 cm on average.
We know that a fighter is 'good' or 'bad' because of the record they have in their weight division. This is simply the number of wins compared with the number of losses. Floyd Mayweather is 5 foot 8, weighed in the region of 154 pounds in 2007 and 2008, and has a reach of only 183 cm. But he has a record of 42-0. Forty two wins, and no losses (according to wikipedia...I am no boxing genius). Tony Pep, a 6 foot 1.5 inch boxer from Canada, has a record of 43-10-1 (43 wins, 10 losses and 1 draw). One of his losses came against Mayweather.
Mayweather is a good boxer. Pep is less good--but not by any means bad.
Now, if we get to the heavyweights we do the same analysis. Wladimir Klitschko is 6 foot 6, weighs about 210 pounds, and reaches 206 cm with his arm. Wow. He has a fight record of 56 wins and 3 losses and is currently considered world champion. A good boxer. Axel Schulz was also a heavyweight. Standing at 6 foot 3, he also weighed over 200 pounds and had an arm length of 193 cm. His record was 26 wins, 5 losses and a draw. Not a bad boxer, but not as good as Klitschko.
One wonders if Schulz would have beaten Mayweather. I am not sure if boxing aficionados will jump on me in saying I expect he would have. he just had more size. And in boxing size matters. But that does not mean he was a better boxer than Mayweather. Actually many believe that Mayweather is the best of all time, across all weight divisions.
All this to say that we need to compare like with like, controlling for the broadest factors that affect performance. This is something we learn from boxing. And it works in governance indicators as well. The approach I have taken shows this.
- The United States is an income heavyweight that does less-well than the countries we can compare it to in the area of child health--its survival rates are just above average but its costs are way above average.
- Singapore is an income heavyweight that does really well when compared against its comparators (its survival rates are higher than average and its costs are lower than average).
See both in the two dimensional graph below, with the horizontal axis showing how many standard deviations all these rich countries are from the average rich country survival rate; the vertical axis shows how far the countries are from the average cost of producing this outcome.
Now, one may say that this does not account for all of the contextual influences on under 5 mortality rates and thus we cannot say "the USA child health sector is governed less-well than that in Singapore." One would certainly be right in some respects.
Some contextual variables emerge as important.
- First, you will see that bigger countries tend to be located in the upper right quadrant with the USA (relatively high survival rates but relatively high costs as well). The lower right quadrant (with relatively high survival rates at relatively low costs) is in contrast dominated by smaller countries. When I look at a single measure of outcome performance (as per the last blog), the top 25 child health countries average out with populations of 12 million. The bottom 25 have 31 million. The difference in averages is statistically significant (at 5%) suggesting that bigger countries have a harder time in this area.
- Second, the better performing countries tend to have higher degrees of population density. Statistical analysis suggests that ten times more people are located in equivalent land areas in the better performing countries than the poorer performing countries (again, a statistically significant difference in means). So it seems that child health outcomes are easier to provide at lower costs in rich countries with smaller, more densely populated populations. Island states are the place to be!
- Third, you can see that the top right quadrant is dominated by the old guard of the OECD. This shows up in the stats. The average number of years' membership in OECD is 41 for the 25 worse performing countries. It is 9 for the 25 better performing countries. It is a statistically significant difference. I will look to interpret it in future posts but some possibilities are: Rich countries that are more newly rich have better child health performance than others; rich countries with newer child health service delivery mechanisms perform better than others; rich countries with younger populations perform better than others.
- Fourth, if we look at the data we find that countries that perform better than others have lower numbers of doctors. The top 25 performing rich countries have about 2.5 doctors per 1,000 while the bottom 25 performers have 3.2 doctors per 1,000. So there is something here about the size of the medical profession (that could be related to the third point on OECD membership and also to the size and spread of population in the first and second points above).
- Fifth, I looked to see if there is any variation between conditions of children in these countries, and it turns out that the better performing countries have lower levels of undernourished children (5 per 1,000 as compared with 5.3 per thousand).
Now one may say we should control for all these factors--and a bunch of others like gini coefficients, number of years kids are in school, teenage pregnancies, and a range of other factors we think or know impact on a country's performance. One may say that we only get to governance issues once these have been controlled.
So, without taking a firm position, let me say why this may not be a great idea. I will use the boxing metaphor to help me.
- First, boxers in various divisions have something in common--they fall into a particular weight group. Beyond this, they have many things that are not in common. Some heavyweights are taller than others, for instance.
- Second, these variations matter. If we went to do a statistical analysis explaining why some boxers win the heavyweight championship and others do not, we are bound to find that height matters. Apart from Mike Tyson, there has not been a heavyweight champion in the past few decades who is shorter than 6 foot 2 (he stands at 5 foot 10, I believe).
- Third, heavyewights who are shorter than 6 foot 2 still have to compete with other heavyweights. They get no special 'control' to help explain that their losses were against taller opponents. This may have helped David Ruiz improve his record from 44 wins and 9 losses...given 2 of the losses came in decisions against Nikolay Valuev who stands at over 7 ft tall and weighs over 320 pounds (see picture). His record could have been 44 wins, 7 losses, 2 losses against the 7 ft tall guy. No. It was 44 and 9. His context was controlled for in terms of weight, and beyond that he was simply expected to compete. (In fact he did compete, with the losses being controversial...)
- Fourth, one of the reasons why boxing allows someone like Ruiz to be compared and competitive with someone like Valuev is because, when one gets into the ring, factors other then height matter. Ruiz mastered techniques that fighters like Valuev did not. Mike Tyson had aggression, speed and power that many taller fighters lacked. Some fighters have more organized management teams, cut-men in their corners, or preparation approaches.
- Fifth, many of these other factors are variables that the fighters themselves can control. The techniques they adopt, the number of hours they put in at the gym, their selection of manager. They actually need to control and manage these factors in relation to the factors they cannot control. If a tall fighter wants to use his height as an advantage, he should choose a specific technique, for example. If a fighter has natural speed he needs to engineer a fighting technique that works towards this advantage.
- Sixth, good fighters are the ones who put all the pieces together. They build on their different natural strengths by adopting relevant techniques etc.
I would like to suggest that these sixobservations are important for our thinking about governance:
- Put together, they tell me that there are many factors that an econometric analysis will show explains which rich countries are more likely to provide good child health outcomes.
- Some of these are natural or entrenched contextual advantages or disadvantages that countries cannot do that much about. The United States cannot be a small island state like Singapore.
- But there are defining factors that are dynamic and reflect strategic choices countries make to do better. Countries can do something about the medical technologies they adopt, for instance, or the degree to which their children are under nourished.
- The choice or dynamic action is contingent on the natural factors. Countries with large and deconcentrated populations need to adopt different mechanisms to improve their child health performance than Singapore does.
- Governance is about whether governments use their authority to make dynamic adjustments that fit the natural contexts in which they are found. These adjustments should counter the contextual constraints and build on the contextual strengths in place. The USA does not get a pass on poor performance because it is large and dispersed. Good governance is achieved in the USA when it makes the policy, organizational, etc. decisions needed to improve its child health performance given its context.
In this analysis, the quality of governance is indicated--not necessarily measured--by the outcomes in evidence. They tell us if the country is performing at, above or below expectation (given its income level), which is an indication of the degree to which authority is exercised in such a way to build on natural strengths and compensate for natural constraints.
What do you think?