As a follow up to the story in the Wall Street Journal (paywall), I’ve been delving into the details of the calculations behind the World Bank’s Doing Business rankings for Chile. I thought it would be helpful to illustrate what the rankings would be under an unchanging measure of the business climate.
To be specific, what I decided in advance was to pick all of the underlying variables for Doing Business indicators that are available for all 5 years, DB 2014-2018. Because some of variables used in the past have been discontinued, and some have been added, sticking with a fixed set of variables means that I will not replicate the ranking in the report from any year.
This approach is only one of many different ways to see what the trend would be in Chile’s ranking when we hold the underlying indicators constant. There are other ways to do this that will generate different levels of for the ranking. The only general result that is worth considering from such an exercise is that the change in rank for Chile is smaller under any of these methods that the methods constant.
- – The data year is when the data were collected. The DB year is from the title of the report. For example, DB 2018 was published in October 2017 and has data collected in 2017.
- – Distance to the Frontier (DTF) is defined as a value between 0 and 1 where 1 represents the frontier or best performance. A bigger value for DTF means closer to the frontier. The country that is closest to the frontier in any year has rank 1. A bigger number in the ranking means farther from the frontier relative to others.
- – According to columns that show the DTF and Rank with no changes, between 2016 and 2017, Chile improved according to the DTF but it ended up with a lower ranking. This is because other countries also improved their DTF.
- – The first dashed line shows the break from the Pinera to the Bachelet administrations. The second dashed line notes a change in leadership of DB inside the inside the Bank and a decision not to bring in any new measurements between data years 2016 and 2017. So the published rankings for 2016 and 2017 (or DB 2017 and DB 2018) also use a fixed method, but one that differs from the one I could hold constant for the five years from 2013-2017. So under one fixed way of measuring the business climate Chiles rank fell and under the other it increased. This illustrates the fact that there is underlying randomness, so one should be careful not to read too much into the changes from year to year and look instead for longer trends.
- – The change in Chile’s rank between 2013 and 2017 in the Ranking with no changes column is smaller than the published change during the time when the new measurements were added. In the Ranking with no changes column, Chile’s rank goes down by 5 places; in the published ranking, it goes down by 23-21 places from calendar year 2013 to 2016 or to 2017.
Here are links for two files. The first goes to a github page from which you can view or download a Jupyter notebook that anyone can run if they have a local install of Python 3.6. (I used the Anaconda distribution of Python.)
The second is a PDF printout of the results from the notebook.
The comments in the notebook help illustrate the specifics that one must set to generate a ranking of this type. It shows how I departed from the exact procedures that the Doing Business team uses to calculate the distance to the frontier. So these are not in any sense official calculations. Rather, they are illustrations of how anyone can calculate their own ranking from the underlying data.
In a separate post, I point out that in my attempt at promoting clarity, I failed to live up to the well established rule for an author, “show, don’t tell.”