So I find that at municipal level, the larger size of middle class, those who live in urban area or are at least ever educated in high-school, contrary to the idea that they are the strong pressure group in favor to clean governance, is associated with lower, yes, lower bureaucracy's governance quality.
I decided to call Philips, my favorite political scientist (and avid LP collector), to consult this result. After hearing my brief explanation, he just laughed and said, " they are probably the bad guys themselves".
Maybe so. Any other idea?
Sounds really interesting. What dataset(s) are you using?
ReplyDeleteDaniel, I use secondary macro data from various sources including BPS, KPU, Sub National Budget, etc. I am still working into it, though.
ReplyDeleteIt is interesting because when I regress Corruption Perception Index on the left hand side with share of residents with at least 12 years of education as an explanatory variable, I basically get your result too. Mine isn't statistically significant and very small, but the sign is the same as yours.
ReplyDeleteSo, it's showing the same results on different governance measures. Very interesting.
if the governance quality measure is a perception, then it could be because the middle class has a higher standard for quality. it could also indicate a difference in social norms; e.g., the line separating bribes and gifts is very thin and is depend on the prevailing social norms. presumably, the lower class adheres to a more traditional norm where economic transactions are part of larger dynamics e.g., status signaling.
ReplyDeleteRoby, if your argument is true, then why doesn't it seem to hold up for the international cross-country CPI?
ReplyDeleteDaniel and Roby, it's rather statistically significant here, and yes, one thing that I am trying to work on is to build better dependent variable --things like public service efficiency or corruption level.
ReplyDeleteDaniel, i don't work on this area so i can't offer more explanation. i don't even know what "the international cross-country CPI" is, not to mention how it's measured and what it means.
ReplyDeleteit was an "argument" that i made up in about 1 minute after reading rizal's post.
Rizal,
ReplyDeleteIt is interesting finding.
However, I am afraid that the finding may be spurious relationship. Clearly, when the size of middle class in one region is large, I guess economic activity in that area is more lively than one with small size of middle class. Then, demand for clean bureaucracy in the region with more economic activity will be higher than the region with less economic activity. So, you got positive relationship.
The logic above also implies that the region with more economic activity provides more opportunity for corruption. Again, I am afraid if you use corruption level as another dependent variable, you will find similar finding. In other words, if you find positive relationship between size of middle class and level of corruption, it may be caused by the larger opportunity for corruption as an impact of more economic activity in the region with larger size of middle class.
I do not know if you already put regional GDP as a control variable in the model. But, may be the relationship not straight forward.
Dendi, thanks.
ReplyDeleteBut how come higher GDP per capita, you think, can go hand in hand with higher corruption? Let alone,if, as you also said, the higher GDP is associated with the larger size of middle class who demanded more clean bureaucracy?
Rizal,
ReplyDeleteI was thinking that you may need to consider the hypothesis that corruption is like oil in economy. It may help economic activities more fluently. You could test quadratic model to find an optimal level of corruption.
Another twist of Dendi's idea. Large size of middle class (in terms of population)means that public service coverage should be larger too (the Optimal public services would be a function of population, i think) .
ReplyDeleteGiven constrained bureaucratic resources, i'm worried that your corruption perception index may incorporate the effect of poor public services (as i think that corruption is different from poor public services).
Perhaps, it would be..(Ape's favorite techniques...he..he sorry Pe for using/abusing your name)..better if you use instrumental variables. But i'm not sure what sort of variables you can use....still thinking.
Ups..sorry zal i just realized that you did not use corruption perception index. Your result quite sensible since the quality of government may depend on the size of population. But there would be many directions. First is inline with your conclusion about middle class. But it is also possible that the public services are just not optimal given the size of population.
ReplyDeleteYudo , I am still looking for instruments for income per capita (to separate the governance effect) and the size of middle class (to take out GDP effect, hence governance effect).
ReplyDeleteAnyway, I am taking notes for any suggestions here.
rizal, since you mentioned my name, don't forget to send me the data...hehehe..
ReplyDeletephilips
Rizal,
ReplyDeleteI have another thought in measuring middle class variable. Middle class is very independent group in term of economical and political characteristics. It has very strict definition. I am not sure that number of people based on education can precisely be a proxy of middle class. May be number of personal tax payer in each region better variable to measure size of middle class. I think the data should be available in Dirjen Pajak. Just my two cents.
Rizal, I may be able help you think of some instruments if you tell me exactly what your dependent variable is.
ReplyDeleteThe way I understand it, you have some governance measure as your dependent variable, then education and GDP as your independent variables. Is this correct?
By the way, have you read Deininger and Mpuga (2005)?