Monday, February 22, 2016

Economic failures

I read recently a critique of the economics field, by Timothy Garton Ash, based on the supposed failure of the field to achieve the basic principles of science. The author stresses the need, in economics, of more humble and modest assertions giving the uncertainty of the field.

I went to Wikipedia to find the definition of science and it says “Science is a systematic enterprise that creates, builds and organizes knowledge in the form of testable explanations and predictions about the universe.”

And goes on saying:

“Popper proposed replacing verifiability with falsifiability as the landmark of scientific theories, and replacing induction with falsification as the empirical method. Popper further claimed that there is actually only one universal method, not specific to science: the negative method of criticism, trial and error. It covers all products of the human mind, including science, mathematics, philosophy, and art. […] A scientific theory is empirical, and is always open to falsification if new evidence is presented. That is, no theory is ever considered strictly certain as science accepts the concept of fallibilism.”

This is probably one of the main deficiencies of the economic science, the lack of replicability. Not that it can’t be done, but that there is no incentive whatsoever to the economic researcher for doing so. That causes a big black hole, without new evidence there is no falsification.

So, when Andrew C. Chang and Phillip Li decided to start digging on replicability and falsifiability in economics, this is what happened: 

Sunday, February 14, 2016

Cities' households geographic distributions

David Cuberes and Jennifer Roberts published this paper in October last year. It deals with the geographic distribution of households’ income within the main British cities (excluding London).

They study the extent to which distance of residence from the city centre is a function of income. This is, apparently, the first study of its kind for British cities. They take into account the main potential factors influencing location, such as household characteristics and access to transport, as well as city and time effects, and taking account of both spatial and serial correlation.

They state four main findings: i) there is a strong positive association between household’s income and distance from the city centre. ii) there is no evidence that richer households locate further from the city centre mainly because they prefer larger dwellings. iii) poorer households that live closer to the city centre experienced increasing real incomes over the period relative to those who live further away. This supports the view that cities in Britain attract poor people rather than generate poverty. iv) public transport availability can’t explain the spatial distribution of income.

These results are very similar to those found for the US, and, at least for the case of Britain, they contrast with those who argued that in Europe richer households tend to live in the city centres where amenities are concentrated.

To analyze the relationship between household income and distance from the CBD they used the following model:

where i is household, t is wave (British demographic survey waves), D is distance (km) from the CBD (Central Business District), Y is the household’s real equivalent net income and X is a set of control variables.

This is the regression results:

Interesting to see that children and age have a very important and significant push, while higher education population attract people to the CBD. The negative sign of higher education looks counterintuitive but the authors say "this may reflect the fact that household’s income and education are highly correlated."

Sunday, February 7, 2016

Data languages 2016

A year ago I wrote about statistical software, data analysis languages and their impact on salaries. I decided to give an update today. 

Using LinkedIn US, I searched for jobs that had any of those languages in their description in addition to the words ‘data’ and ‘economics’. Then I worked out the average salary for those jobs where wages were available. 

Surprisingly, Eviews came on top followed by R. Salaries for those two are, on average, slightly above 63,000 $US. Yet, a big differerence remained; R had more than 3,000 opening positions across US while Eviews had only dozens. Excel ranked, not surprisingly, the lowest; the average salary happened to be 58,000 $US.

It was strange to see Stata performing so badly. Yet that is something we already saw on last’s year analysis. The reason could well be that Stata is the data language of the academia where salaries tend to be lower than the financial/consultancy sectors, for a given set of skills.