Economic Structure & GDP/Capita

“The first problem for the government in carrying out an industrial policy is that we actually know precious little about identifying, before the fact, a ‘winning’ industrial structure. There is not a set of economic criteria that determine what gives different countries preeminence in particular lines of business. Nor is it at all clear what the substantive criteria would be for deciding which older industries to protect or restructure.” Charles Schultze, chairman of the Council of Economic Advisors under US President Jimmy Carter

This essay is going to present some research I’ve done recently on industrial structure or what I like to call ‘economic structure.’ A country’s economic structure describes what industries or economic activities it is involved in. We are used to classifying a country’s economic activity as either agricultural, manufacturing or services. Knowing the percentage of economic output that is either agricultural, manufacturing or services we then know the overall economic structure of the country. Of course this is only to a ‘3-industry’ level of detail. You could classify a country’s economic activities much more precisely and in fact in future essays I hope to look at data where a country’s economic activity has been classified to a ’12-industry’ level of detail. The aim of this essay though is to investigate the relationship between a country’s ‘3-industry’ economic structure and its GDP/capita. If there is a strong correlation this would suggest that the path to economic development is largely the same for all countries because for every level of development (i.e. GDP/capita) there is a specific economic structure. If there is a weak correlation this suggests that there are many paths to economic development and that a country’s success is not just a function of what it does but also how well it does it.

3-industry analysis

I did a 3-industry analysis of all the countries in the world I could find data on excluding countries that have populations of less than a million people. The reasoning for excluding them was partly because I didn’t have any data for them but also because intuitively I felt like small countries could have bizarre economic structures which might be very misleading. For example a country of just a few hundred thousand people might have an economy that is largely based on tourism which clearly for much larger countries, with tens of millions of people, is not scalable.

What we find is that the R², a measure of how well the data fits our statistical model, is 0.35 which is not bad but not great. Removing the ‘oil countries’ though we find the R² improves significantly to 0.47.

One puzzling aspect of the results is why the p-values are so low. Even after removing agriculture, especially as for many of the developed countries agriculture is just a few % of GDP, multicollinearity should be a big problem. This is because lowering manufacturing’s % of GDP should (in some cases almost 1-1) increase services’ % of GDP. However regressing manufacturing on services I find an R² of just 0.09 suggesting that there is no multicollinearity problem.

However, further investigation suggests that the t-tests are not valid because the errors are not normally distributed. Eyeballing the pnorm and qnorm graphs we can see the errors are clearly not normal.

Taking a look at a 3-D graph though it becomes clear that it is actually agriculture that does most of the explanatory work. Below you can see a top-down view of the 3-D graph where the red line represents the 1-1 trade-off between services and manufacturing. The closer a data point is to the red line the smaller agriculture contribution to GDP is and the larger the combined contribution of services and manufacturing is.

Taking a side-on view we can see that when agriculture is a high percentage, more than 10% of GDP, i.e. left of the blue line GDP/capita is very low. If agriculture is less than 10% of GDP suddenly GDP/capita is much higher. In fact if you compare the average GDP/capita of countries with agriculture that is more than 10% of GDP/capita to countries where agriculture is less than 10% of GDP/capita you find a stark difference. The more agricultural economies have an average GDP/capita of just $4294.8 whereas the less agricultural economies have an average GDP/capita of $26261.7. And in fact a regression on just % of GDP that is agriculture has an R² of 0.43 only slightly lower than the manufacturing and services regression R² of 0.47. As you can see below, as GDP/capita rises countries tend to have a larger Services % as compared to Manufacturing % but the effect is not as extreme as one might expect.

I think the real puzzle to me is why there is this turning point at around 10% agriculture. It cannot just be that the less productive agriculture is being phased out in favour of the more productive manufacturing and services. There must be something more fundamental going on but what exactly I’m not sure. Nonetheless I think you could make a good argument that the countries in the left column which have relatively low GDP/capita given their low % of GDP that is agriculture are perhaps on the verge of a massive economic growth as they race up the curve and join the countries on the right column.

10% Agriculture

I want now to try and investigate what is happening when countries agriculture drops below 10% and explain why GDP/capita rockets up so much once this threshold is breached. Unfortunately the Serbian and Argentina websites are not in English and so I was unable to find the data I wanted. Thailand, Macedonia and Malaysia on the other hand have only very limited data sets available. This leaves us with Turkey, Belarus, Tunisia and China.

Part of the problem is I’m not sure what I’m looking for. I suppose it seems like manufacturing and services suddenly become much more productive when agriculture dips below 10%. So why is this? Is it coincidence? Is that at that point enough labour transfers across? Is it that old industries suddenly become more productive or new industries become less productive?

Next steps a) Investigate countries around the 10% boundary. Why the sudden increase in GDP/capita b) Investigate whether countries historic development in terms of the turning point around 10% agriculture matches the current distribution. c) Investigate 12-industry data to see whether there is one path to economic development or multiple. Does economic structure determine GDP/capita?