The Relationship between Consumption and Household Debt at Macro-level : An empirical investigation of US data

Abstract

I decided to carry out a small research aimed at establishing causality between household debt and aggregate consumption. It was important to find out whether household debt affects economy at all.

I started by consideration of the most important macroeconomic theories of consumption. On the basis of those theories I formulated a few empirical relations between aggregate consumption and some macroeconomic variables in log-log form. One of those variables was household debt. Then I wanted to test those relations by the example of data from a country. For that purpose I chose the United States since there are a lot of easily available data on this country. Using the data I estimated the relations.

Ordinary least squares (OLS), and instrumental variables (IV) were two techniques which I applied to the estimation of the relations. First I tried the OLS which gave interesting results about the drop in the savings rate in the United States between the mid -1980s and the recent financial crisis. The IV estimation pointed out that a simultaneity bias was eliminated after introducing a new regressor (lagged disposable income).

Finally, I discovered a positive relationship between accumulation of household debt and aggregate consumption both in the short run and in the long run. This result is preliminary because it relies on the data from only one country (the United States).

The implication of my findings concerns financial sectors in different countries, especially in developing countries. The good functioning of financial institutions requires good working conditions. In the countries with good legislation, and in the absence of different barriers, financial institutions issue credit to both producers and consumers. Consumer credit increases demand for goods which induces output growth.

I used the OxMetrics 6 software with a PcGive 13 Professional package to trace out the graphs, to estimate the models, and to generate the statistics in sections 3 and 4.