I did not find the time to do a blog post yesterday, because I had the Macroeconomics class test coming up (which went quite well). So I sat down to do some exercises for the blog today instead and again ended up having a look at South Africa’s economy over the time period from 1961-2014 for which data is available from the World Governance Indicators by the World Bank.
In particular, my idea for today was to quantify the economy’s structural transformation since the 1960s. The approach I have chosen for this is to use the growth accounting equation in a modified version including the three sectors agriculture, industry and services and regress the change in log GDP per capita on changes in log value added in agriculture (agr), industry (ind) and services (srv):
In case you want to check on the data I have used, it comes from the World Governance Indicators for South Africa 1961-2014 and are the following series:
- GDP per capita (constant 2005 US$)
- Agriculture, value added (constant 2005 US$)
- Industry, value added (constant 2005 US$)
- Services, etc., value added (constant 2005 US$)
As you can see from the model specification above I use a difference-on-difference model which has the advantage that it turns non-stationary series into stationary series and helps to avoid spurious regression (Gu, 2013). Personally, I am not sure whether my model really suffers from this problem but I turned to this specification in the first place because there was evidence for model misspecification when only using a log-log model or a difference-on-difference level-data model (Ramsey RESET test). Furthermore, I deploy heteroskedastic robust errors as the Breusch-Pagan/ Cook-Weisberg test for heteroskedasticity in STATA provides evidence for such.
The coefficients can then be interpreted as follows: they measure how much the GDP per capita growth rate changes in response to a 1 percent change in the value added of a sector. This makes the regression results convenient in their interpretation – as far as I understand – showing the sectors’ impacts on the economy given their relative weights.
The results of the propsed OLS regressions are summarized in the table below where one star, two stars and three stars reflect the significance levels of 0.10, 0.05 and 0.01 respectively. I estimate the coefficients for each decade from 1960s, 70s, 80s, 90s to the most recent somewhat longer period as well as one regression for the whole period looked at.
The key findings are that over the complete period from 1961 to 2014 all three sectors are statistically significant but agriculture has become “insignificant” in the most recent period (as well as in the 1980s). Industry is significant at 1 percent level except from the 1960s but its impact is steadily declining ever since. Services are on the rise with a sharp increase in their impact from the 1960s to the 1970s. After a short-term decline in 1990s they have now become more important than ever.
These are not particularly surprising finding; they are evidence for the structural change that has happened in South Africa over time. It proofs that South Africa has turned into a services economy in its transition to an emerging market and rise to one of the BRICS. Value added of services now has the largest impact on the country’s growth rate and in the most recent period, a 1% increase in the growth rate of the services sector has translated into a 0.78% increase in the growth rate of GDP per capita (correct me if I am wrong in my interpretation, please).
A cautionary note at the end: One should not forget that there are inter-industrial links that tie agriculture, services and industry together. It would therefore be foolish to focus solely on growth in the services sector even given its importance. What is more, growth of the tertiary sector and structural change must be accompanied by a stable agricultural sector to ultimately stabilize the food supply – assuming that a country is not relying completely on food imports – as admitted by Lewis (1954; Lewis labour surplus model) and formalized by Bruce Johnston and John Mellor in 1961. Another takeaway from this exercise is that – as services are on the rise – improvements in the sector’s overall productivity become crucial to the country’s future economic development and long-run growth.
Thanks for reading!
Gu, S. (2013). Are Mortality Rates Random Walk? Panel Unit-root Tests with Evidences from Micro Data. University of Notre Dame, April 2013.
Johnston, B., and Mellor, J. (1961). The Role of Agriculture in Economic Development. American Economic Review, 51(4), pp.566-593.
Lewis, W.A. (1954). Economic Growth with unlimited Supplies of Labour, The Manchester School of Economic and Social Studies, May 1954.
World Bank (2016). World Development Indicators South Africa [Data]. Retrieved April 9, 2016, from World Development Indicators (WDI) database: http://data.worldbank.org/country/south-africa.