Firms’ investment decisions in response to demand and price uncertainty

Abstract

We investigate the effect of demand and price uncertainty on firms’ planned and realized investment from a panel of manufacturing firms. Uncertainty measures are derived from firms’ own expectations about demand and prices and firm's sales. We find that demand uncertainty at the time of planning depresses planned and subsequent realized investment. Firms do not revise their plans due to demand uncertainty at the time of spending, suggesting that reducing demand uncertainty will only have lagged effects on investment. We do not find any effect of price uncertainty. Our results are consistent with the behaviour of monopolistic firms with irreversible capital.

Acknowlegements

We would like to thank Paul Butzen, Robert Chirinko, Patrick Sevestre, Anders Warne, Rafael Wouters, participants to the 11th Panel Data conference at Texas A&M and participants at seminars at ECARES (Université Libre de Bruxelles) and the University of Nottingham for useful discussions. We would also like to express our gratitude to two anonymous referees and editor Mark Taylor for useful suggestions. Many thanks are owed as well to Frank Windmeijer for providing us with the codes for the Windmeijer correction in dynamic panels.

Notes

¹An increase in price uncertainty increases the probability of both positive and negative price shocks. However, by convexity of the marginal profitability of capital, increases in profitability due to positive price shocks are larger than reductions in profitability due to negative price shocks.

²This can be attributed to the following fact. With a flat demand curve, the perfectly competitive firm can benefit from both price and output increases, in response to a positive demand shock. With an elastic demand curve, the imperfectly competitive firm can increase output only at the cost of lower prices. Therefore, the profit derived from a positive demand shock is lower.

³For perfectly competitive firms the marginal profitability of capital does not depend on previous investment so that irreversibility does not affect the sign of the investment-uncertainty relationship.

⁴We do not consider risk aversion in the discussion and rather assume that firms are risk-neutral.

⁵To illustrate this, consider the following simplified demand and supply functions for a single firm (1) P_d =C_d·e^εd·Q^ψ with ψ≤0 (ψ=0 for a perfectly competitive firm), and (2) P_s =C_s · e^εs · Q^ϕ with ϕ>0, where εd and εs are, respectively, demand and supply shocks independent of each other and with respective variances σ_d and σ_s . From the equilibrium condition on the goods market it may easily be shown that the variances for (the log of) output, q, and prices, p, depend on the variance of both demand and supply shocks: var[q]=σ_q =(1/(ϕ−ψ))²(σ_d +σ_s ) and var[p]=σ_p =(1/(ϕ−ψ))²(ϕ²·σ_d +ψ²·σ_s ).

⁶Bond and Cummins (2004) use analysts’ quantitative expectations of the firm's future profits. Since expectations are not formed by the firm itself, their uncertainty measure is essentially the market‘s perceived risk about the firm rather than the firm's perceived uncertainty.

⁷The Business Survey reports information by firms’ product and plant rather than by firm. Since the information is qualitative we cannot simply add the information for each product. We select the product that accounts for most of the firm's turnover to approximate the firm's total demand.

⁸See Butzen et al . (2002) for different cash flow sensitivities between small and large Belgian firms and Ghosal and Loungani (2000) for different investment-uncertainty sensitivity of US firms.

⁹We very much rely on this survey because in 85% of the cases reported realized investment in the survey coincides with investment as reported in the annual accounts.

¹⁰Replacing expected future sales growth by its realization introduces a forecast error in the residual of the investment equation. Therefore, the equation must be estimated by instrumental variables. We use the past values of all RHS variables as instruments. So it is equivalent to rational expectations where the information set consists in the past values of all RHS variables.

¹¹However it is significant for realized investment. This result is not surprising since firms have to forecast next year's sales growth at the year of planning and know next year's sales growth at the year of realization. This finding is consistent with the firm's forecast of next years sales growth being largely captured by current sales growth and firms reacting to sales growth surprises in the year of realization (a more formal test is performed below).

¹²The system GMM estimator is necessary because Δy_{t +1} is endogenous since investment revisions in year t+1 will affect the capital stock in year t+1, hence sales over that period. In addition the system GMM estimator allows for fixed effects to differ in the plans and realizations equations. Although this might suggest a systematic bias in the investment plans, this may be the case because firms report investment plans for year t+1 only when these have been approved by the board. So investment decided on in January t+1 and carried out in the same year would not be reported in the plans in year t.

¹³We cannot construct revisions of future sales growth since we have no data on expected sales growth. In the investment equations expected sales growth was estimated indirectly by GMM. In Equation 3 Δy_{t +1} should at least be correlated with revisions in future sales growth. Therefore, a significant coefficient α ₁ may be considered as supportive of the assumption that firms revise their investment decision due to new information regarding their fundamentals.

¹⁴See footnote 13.

¹⁵Strictly speaking, this only applies for uncertainty at the horizon of our measures. Realizations may differ from the plans due to shifts in longer-horizon uncertainty that are not captured by our measure. The same goes for longer-term sales expectations.

¹⁶Since the Business Survey is conducted product by product and the Investment Survey by firm, we consider the product that accounts for the largest part of the firm's turnover as a proxy for the firm's output.

¹⁷The average age of sold and used capital is estimated from the annual accounts information on depreciation. Details will be provided by the authors on request.

¹⁸This is again inferred from annual accounts information on depreciation.