Abstract

Although the empirically tested multifactor models are diverse, they usually
share one common assumption: Agents are sure about the state of the economy.
Thus, researcher abstract from "macroeconomic uncertainty" or "state
uncertainty" . I contribute to the growing literature that addresses the following question:
How important is uncertainty in the context of agents being unsure about the state of the economy in addition to
market risk and factor risk for the pricing of risky assets?

In order to provide an answer to this question, I proceed in three stages.
First, I solve a dynamic portfolio choice problem with robust preferences and
derive a discrete time multifactor asset pricing model. The market return and
the long-term rate of GDP growth are risk factors and state
uncertainty is a separate factor that helps pricing assets.
Second, based on the derived asset pricing model, I introduce a variable
that measures state uncertainty. The variable is based on Hamilton's nonlinear model
for the real GDP with discrete regime switching between periods of expanison and
contraction. Finally, I evaluate the main insight of the
asset pricing model that the new measure of state uncertainty helps to price
assets, and that state uncertainty is priced as a factor. I use the Fama-French
three-factor model, add the constructed state uncertainty factor
and test whether state uncertainty helps to price a set of test assets (Fama-French size and book-to-market portfolios). The empirical analysis is based on the GMM/discount factor method.

I present two main results concerning the asset pricing implications of the
state uncertainty factor. First, my empirical results confirm previous
findings that state uncertainty is a factor that helps to price assets:

• The additional state uncertainty factor does not drive out the
  statistically significant factor loadings.
• The additional factor state uncertainty is statistically significant.
• I apply the methodology of Hansen and Jagannathan
  to evaluate the ability of the additional state uncertainty factor to price
  the test assets. The HJ-distance measure
  improves when the additional factor is added to each model.
  

Second, I find that the factor state uncertainty helps to explain the value
premium:

• The pricing error reduction due to the additional factor state
  uncertainty is largest for small growth stocks. I conclude that investors
  are more concerned about state uncertainty in relation to small growth
  stocks than to value stocks.
• Time series estimates of factor betas reveal that the state
  uncertainty factor beta - although overall quite low - is statistically significant
  for small growth stocks. Further, the cross-sectional estimate of the
  uncertainty factor risk premium is negative. This implies that value stocks
  earn higher average returns than small growth stocks.
  

My second result is consistent with empirical findings that (small) growth stocks
are high-duration assets (more of their cash flow in the future), whereas
value stocks are low-duration assets (more of their cash flow in the near
future).