Keynesian economists argues that savings are not necessary for investment. In a modern economy with fiat money and fractional reserve banking, money can be created out of thin air and this can be lent to companies wanting to expand or maintain their capital stock. While Keynesian economists will admit that savings used to be required while on a gold standard and a 100% reserve banking, with these "barbaric relics" gone savings is no longer necessary.
The counterargument to the Keynesian argument is that to make these loans without savings requires expanding the money supply (either through printing or factional reserve banking). This will lower interest rates and increase demand for products across the entire structure of production. This means the demand for lower order products, those products closer to actual consumption, will increase at the same time as demand for higher order products. The eventual result is the demand for raw materials will be bid up and inflation will result. Also, companies will find they do not have enough consumers for their products because consumers don't have the savings to buy the new production. Thus, the Keynesian argument is actually the cause of the business cycle itself according to Austrian economists.
I try to test these two competing philosophies using 2 econometric models. The first is a panel regression with data taken from the IMF. It regresses GDP growth against the savings rate for the years 1981 through 2011 over a cross section of 170 countries. If the Classical school is correct, then there should be a positive correlation between savings and GDP growth. If the Keynesian school is correct, there should be either no correlation, or even a negative correlation.
The results of the regression are below. The data shows coefficient, then p-value.
GDP Growth %
Savings Rating %
The results show that a 10% increase in savings is associated with a 1% increase in GDP growth. The results are significant at the 1% level. While the adjusted r-squared is low, that is understandable considering the many other factors that can influence GDP growth. The results are in line with the Classical and Austrian schools.
The second regression takes 5 year averages of GDP growth and savings. Thus one observation will be from 1982 through 1986, and another from 1987 through 1991. There is not a separate observation from 1983 through 1987. I believe this is a better regression because according to the Classical school, the positive effects of savings on GDP growth should take place in the future. It is impossible to know when precisely in the future the effects will be shown in GDP - 1 month, 6 months, 2 years? Taking 5 year averages will help to address this concern. The averages use the same data as the original regression.
The results of the regression using 5 year averages are:
GDP Growth % - 5 Year Average
Savings Rating % - 5 Year Average
The results now show that a 10% increase in savings results in a 1.8% increase in GDP growth. These results further confirm the Classical argument. Both the economic impact is higher, and the adjusted r-squared increases from .03 to .07.
As a future exercise, I would like to add more independent variables to try to increase the r-squared of the regression and confirm that savings is still significant.