Risk aversion as a technology factor in the production function

Abstract

We incorporate risk aversion into the technology component of the production function. In a traditional theoretic framework, we show that an increase in risk aversion increases unemployment and reduces potential output. Our out-of-sample forecasting experiments suggest that while interest rates impact the economy through the demand-side. However, an interest rate spread (TED) is used as a measure of risk aversion and is shown to impact output through the economy's supply-side.

Keywords:

• forecasting
• gross domestic product
• interest rates
• risk aversion
• technology
• production function
• time series

JEL Classification:

• E10
• E17
• E32
• E44

Acknowledgements

We are grateful to an anonymous referee for helpful comments and suggestions. All remaining errors are our responsibility.

Notes

¹ A different demand-side cause of liquidity constraint is through the financial accelerator in Bernanke et al. (1999). They argued the changes in asset prices over the business cycle impacts on individuals' and businesses' collateral position.

² Such discussions focused on, for example, the efficiency of a barter economy relative to that of a monetary economy.

³ The LIBOR rate is the interest rate at which banks borrow unsecured funds from other banks in the London interbank market. One explanation for this LB−TB interest rate spread being referred to as the ‘TED’ is from ‘T-bill’ and ‘ED’ – the ticker for the Eurodollar futures contract (Olofsson, 2008, p. 40).

⁴ Data for the treasury-bill rate and the LIBOR were obtained from Economagic.com. All other data were obtained from the St. Louis Federal Reserve Bank. Data for the M1 and M2 money multipliers were constructed by dividing the respective monetary aggregate by the monetary base.

⁵ If the results from model 2C are ignored, model 4C produces MAPEs lower than the 10 remaining models in each of the one-to-four-step ahead forecasts. That is, model 4C is the ‘second best’ model specification. Further, we replicated the analysis described in Equation 16 and Table 1 except that we used M1 as the monetary aggregate. The conclusions in this case are identical with those reported with M2 in Table 2. That is, with M1, model 2C again dominates in the first-, second- and fourth-step ahead forecasts and model 4C again dominates in the third-step. As with M2, using M1 and ignoring results from model 2C, model 4C again produces MAPEs lower than the 10 remaining models in each of the one-to-four-step ahead forecasts. As with M2, model 4C is the second best model specification when using M1.

⁶ We also conducted a series of forecasting experiments for consumer prices (P). That is, we replicated the analysis described by Equation 16 and Table 1 – except that P was substituted for Y in each model specification. In addition, we expanded the analysis to consider M1 as an alternative for M2. We found that neither interest rate measures nor the risk measure serve as information variables for movements in consumer prices. However, our time-moving analysis indicates an intuitively appealing transition of the information content provided by monetary aggregates when forecasting prices: the monetary base provides information content for prices, as does M1 – but not as strong as the monetary base – and M2 does not provide information content for forecasts of prices.