Working Paper: CEPR ID: DP127
Authors: Adrian Pagan; Aman Ullah
Abstract: This paper provides a critical survey of the methods employed to model the effects of risk in econometric models. Most of the popular methods are shown to suffer from errors-in-variables bias, and an instrumental variable method is suggested to overcome this problem. The technique exploits the orthogonality conditions existing between the squared unanticipated variables and functions of variables making up the information set defining the anticipations. An alternative procedure used in the paper is to directly estimate the conditional variance (risk) by non-parametric estimators. Applications are made to foreign exchange markets, interest rates and unemployment/inflation risk relations.
Keywords: risk; instrumental variables; errors in variables; ARCH models
JEL Codes: JEL Classification: 131, 211, 212
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| variance of relative price movements (E30) | indicator of risk (D81) |
| changes in risk measures (C22) | influence economic variables (F69) |
| risk measures (G32) | economic decision-making (D87) |
| proxy variables for risk (G17) | underestimation of the effect of risk on economic decisions (D81) |
| errors-in-variables bias (C20) | underestimation of the effect of risk on economic decisions (D81) |
| IV method (C26) | mitigate errors-in-variables bias (C20) |