Doctoral Dissertations

Date of Award

8-1987

Degree Type

Dissertation

Degree Name

Doctor of Philosophy

Major

Economics

Major Professor

George C. Philippatos

Abstract

Although it has become one of the central models of modern financial theory, the Arbitrage Pricing Theory (APT) remains controversial. The present study tests the theory in the markets for spot foreign exchange, which have received relatively little attention in the literature of APT testing.

This research extends the analyses of earlier studies in several respects; It employs a larger set of currencies, a shorter data interval, a longer sample period, and multiple subperiods. Most importantly, the tests implemented here have greater statistical power, since they explicitly allow for the possibility that idiosyncratic risk may be priced.

The implications of the APT tested here are that: (a) at least one factor is priced, (b) the relation is linear, and (c) no other source of risk is priced. Essentially, the present test consists in regressing estimates of expected returns on estimated systematic and non-systematic risk.

The raw data, provided on tape by the Federal Reserve, are daily noon interbank bid rates (in $US) from January 2, 1974 through July 12, 1985, for the currencies of 18 industrialized nations. Rates of return are calculated as the first differences of successive natural logarithms. There are 2,889 observations in all. In order to conduct several independent tests, the data are split into four periods of approximately equal length.

Expected returns are proxied by mean realized returns, while systematic risk is estimated by means of maximum likelihood factor analysis. Non-systematic risk is measured by both standard deviation and the residual from the factor model.

Although its results are generally consistent with previous tests using stock market data, the present study augments earlier work using currency data and concludes that: (1) just one factor is priced; (2) there is no convincing evidence that mean returns depend in a nonlinear fashion on factor risk; (3) there is not much evidence that non-factor risk is priced; and (4) there is little evidence that total risk either contributes explanatory power or is priced independently of factor risk. Therefore, the testable implications of the Arbitrage Pricing Theory cannot be rejected by data from the foreign exchange markets.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS