Per-unit bidding rules and buyer under-performance in natural resource sales1

¹ The views expressed in this article are not to be attributed to the authors’ employers.

Abstract

In this article, we examine the role of per-unit bidding rules on firm-level contractual performance. In particular, we test the hypothesis that buyers will act on incentives to under-perform when bids are accepted in per-unit form. The empirical application uses data from per-unit auction sales of US Forest Service timber. The statistical analysis indicates that buyers systematically undercut when per-unit bids exceed the value of individual units of timber.

¹ The views expressed in this article are not to be attributed to the authors’ employers.

Notes

¹ The views expressed in this article are not to be attributed to the authors’ employers.

² Such moral hazard problems have been identified in other specific cases, such as health insurance (Aoki, 2005), and in the more general context of contract design (Nygaard and Myrtveit, 2000).

³ Forest Service managers are aware that undercutting occurs and they expend significant resources to monitor cutting behaviour.

⁴ For these and other technical reasons sellers often place and enforce restrictions on the timing and rate at which each well in a field operates.

⁵ We have no independent measure of the resources spent cruising timber tracts prior to sales. Baldwin et al. (1997) found that bidders invest significant resources in such cruises. However, Haile (2001) and Leffler et al. (2000) offer testimonial and empirical evidence indicating that buyer cruises of tracts sold under per-unit bids are less common and/or less extensive than under lump-sum bid rules.

⁶ As a matter of policy the Forest Service does not allow bids on species which comprise a relatively small volume on the tract. Instead, these species are sold at the reserve price set by the Forest Service.

⁷ The regression reduces the weight of influential observations using the weighting algorithms of Huber (1964) and Beaton and Tukey (1974) using the implementation programmed into Stata 7 for Unix.

⁸ In particular, the estimates allow for measurement error in the overbid variable such that the signal-to-noise ratio is three-fourths.

⁹ In this sample, nonindexed sales exist primarily as a result of differences in the timing of adopting bid indexing in the various forests and also as a result of the different starting dates of the sales. The exception to this are few in number and are the very small volume sales for which the short contract length renders indexing irrelevant. This rules out the possibility that some rule for choosing whether or not to index has an effect on revenue that has been omitted from the model specification

¹⁰ For a parallel discussion with respect to the economics of institutions, property rights and transactions cost, see Barzel (1982) and Eggertsson (1990).