Hi Mehmet,

The WGI comprise 6 dimensions, each based on a number of individual indicators. The WGI are composite governance indicators based on over 30 underlying data sources. These data sources are rescaled and combined to create the six aggregate indicators using a statistical methodology known as an 'unobserved components model'.  A good starting point could be to examine this methodology and see if it can be used or extended so as to form not 6 dimensions but a single composite governance score from the underlying indicators.

See http://info.worldbank.org/governance/wgi/index.aspx#doc

One issue would determining the weights for indicators in a dimension and dimensions, or the effects/consequences of extending the methodology on how indicators weigh in the end result. You would also need to look at what happens in terms of margin of error for an overall governance estimate (i.e. would the estimate still be usefull given its margin of error, which in part depends on your application of the estimate)

http://info.worldbank.org/governance/wgi/index.aspx#doc-methodology

In short, IF you find each dimension equally important, THEN rather than attempting to lump together scores on  6 dimensions, my recommendation would be to go back to the original indicators and form a single estimate from these, in a similar fashion that the dimensions where created.

See http://info.worldbank.org/governance/wgi/index.aspx#doc-methodology

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1682130

Hi Mehmet,

one chance is to give equal weights to all of them... a simple average of all 6 of them. For example the overall index of Doing Buiness by the World Bank was built in this way on about 10 indicators.

Otherwise a Principal Component Analysis could be useful, provided that you end up with one only relevant variable. If you end up with two, then the issue is if they represent some meaningful phenomenon of the starting 6 variables (you've got to see loadings).

Thank you mr bripi. İ tried to use principal component analysis and i got only a single principal component. I named it composite governance index. 

Well, this sounds as an interesting starting point. If you need it for regression analysis, you could try to plug this pca, and then alternatively the 6 original ones, and then check the sources of the different results. Good luckl!

Thank you so much.

Hi Mehmet,

I would like you to help me solve the problem you had to get this clue.

I myself face a similar problem.

thank you

You can look to this paper. Thanks for your interest.

Thanks very much

Hi Mehmet,

I have read your paper and it was great, but your data range was 1996-2014, and when I have downloaded my data from the World Bank Governance Indicators, I have found missing values for years 1997-1999 and 2001

what did you do for these issue?

Thank you

By employing linear interpolation method.

Hi Mehmet,

Thank you very much, I used the linear interpolation method and filled in all missing values of data.

Thank you very much

Is there is any package in python or R that can help replicate the WGI methodology? Something like tensorflow or similar