Creditor financing and overbidding in bankruptcy auctions: Theory and tests

Abstract

We present unique empirical tests for auction overbidding using data from Sweden's auction bankruptcy system. The main creditor (a bank) can neither bid in the auction nor refuse to sell in order to support a minimum price. However, we argue that the bank may increase its expected revenue by financing a bidder in return for a joint bid strategy, and we show that the optimal coalition bid exceeds the bidder's private valuation (overbidding) by an amount that is increasing in the bank's ex ante debt impairment. We find that bank–bidder financing arrangements are common, and our cross-sectional regressions show that winning bids are increasing in the bank-debt impairment as predicted. While, in theory, overbidding may result in the coalition winning against a more efficient rival bidder, our evidence on post-bankruptcy operating performance fails to support such allocative inefficiency effects. We also find that restructurings by bank-financed bidders are relatively risky as they have greater bankruptcy refiling rates, irrespective of the coalition's overbidding incentive.

Introduction

In Sweden, a firm filing for bankruptcy is turned over to a court-appointed trustee who puts the firm up for sale in an auction. This mandatory auction system has an attractive simplicity. All debt claims are stayed during the auction period and the bids determine whether the firm will be continued as a going concern or liquidated piecemeal. Payment must be in cash, allowing the auction proceeds to be distributed to creditors strictly according to absolute priority. Moreover, the auctions are quick (lasting an average of 2 months) and relatively cost-efficient, and as much as three-quarters of the filing firms survive the auction as a going concern (Thorburn, 2000). A going-concern sale takes place by merging the assets and operations of the auctioned firm into the bidder firm, or into an empty corporate shell—much like a leveraged buyout transaction.

There is an ongoing debate over the relative efficiency of auction bankruptcy versus Chapter 11 in the U.S., where firms are reorganized in bankruptcy. Proponents of a more market-oriented auction system point to costs associated with conflicts of interests and excessive continuation of operations due to managerial control over the restructuring process in Chapter 11.¹ Perhaps as a result, there is a trend towards increased use of market-based mechanisms in the U.S., as evidenced by prepackaged bankruptcies (Betker, 1995, Lease et al., 1996), participation by distressed investors (Hotchkiss and Mooradian, 1997), and sales in Chapter 11 (Hotchkiss and Mooradian, 1998, Maksimovic and Phillips, 1998). In fact, Baird and Rasmussen (2003) report that more than three-quarters of all large Chapter 11 cases resolved in 2002 involved the sale of company assets or a prepackaged bankruptcy procedure.

On the other hand, proponents of Chapter 11 argue that the time pressure of an auction system is costly as it possibly causes excessive liquidation of economically viable firms when potential bidders in the auction are themselves financially constrained. However, Eckbo and Thorburn (in press) fail to find auction fire-sale discounts in going-concern sales, and the three-quarters survival rate reported for the Swedish auction system is similar to that of Chapter 11.² Also, in Eckbo and Thorburn (2003) we show that firms purchased as going concerns in the auction tend to perform at par with non-bankrupt industry rivals.

In this paper, we study an empirical issue of importance for the workings of the auction system: the incentives and opportunities of the bankrupt firm's major creditor to influence the auction outcome. We show that creditor incentives have the potential for affecting both the liquidity and allocative efficiency of bankruptcy auctions. Our empirical evidence is based on 260 private Swedish firms auctioned in bankruptcy. The average firm has $5 million in sales and assets of $2 million ($8 million and $4 million, respectively, in 2007 dollars), and it has an average of forty-three employees. This small-firm evidence is of interest to the U.S. debate as a majority of Chapter 11 filings are also by small private firms: Chang and Schoar (2007) report average sales of $2 million and twenty-two employees in a large and representative sample of Chapter 11 filings between 1989 and 2003. Bris et al. (2006) report that the median firm filing for Chapter 11 has assets of $1 million.

In our sample, as is common for small firms, the bankrupt firm's secured creditor is a bank. Upon filing, whenever the bank's debt claim is impaired, the bank becomes the firm's residual claimant and therefore effectively the seller in the auction. Auction rules and banking regulations prevent the bank from openly enforcing a seller reserve price (it can neither bid directly nor refuse to sell). However, we observe that the bank often finances the winning bidder in the auction. We postulate that the bank uses this bid financing as a vehicle for engineering a (bank–bidder) coalition to get around the auction bidding constraint. Under certain conditions, the coalition optimally overbids, i.e., bids more than the private valuation of the bank's coalition partner. We show that the optimal coalition bid mimics the sales price of a monopolist seller when the bank's debt is greatly impaired. In this situation, the bank effectively enforces a reserve price and raises its expected debt recovery. This also raises the possibility that the bankrupt firm ends up being acquired by a relatively inefficient bidder. Allocative inefficiency occurs when the bank–bidder coalition wins against a rival bidder that has a higher private valuation of the target.³

Our empirical analysis addresses both the overbidding- and efficiency aspects of the auctions. First, we present a unique test for the existence of overbidding based directly on auction premiums. To characterize the test approach, it is instructive to think of the bankruptcy event as creating an instant “bank toehold” equal to one (100% bank ownership of the auctioned firm) when bank debt is impaired. As with toehold bidding in general, toeholds raise the bidder's reserve price above and beyond its private valuation—referred to as “overbidding” (Burkart, 1995, Bulow et al., 1999). In our bankruptcy auction setting, bank–bidder coalition overbidding increases the expected auction revenue flowing to the bank.⁴ Hotchkiss and Mooradian (2003) also examine creditor overbidding incentives, but in the context of voluntary sales in Chapter 11 of the U.S. bankruptcy code.

Importantly, the coalition optimally overbids only if the bank's debt claim is impaired at the outset of the auction (further overbidding just benefits junior creditors). It is this cross-sectional restriction that makes overbidding testable in our setting. This restriction does not exist for takeovers outside of bankruptcy. Extant empirical evidence on toehold-induced overbidding is therefore indirect. For example, theory implies that overbidding increases the probability of winning, which is supported by studies of corporate takeover bids with equity toeholds (Betton and Eckbo, 2000).

Our tests require a measure of the degree of bank-debt impairment at the beginning of the auction, capturing the bank's overbidding incentive. We use the ex ante liquidation recovery rate, defined as the firm's expected piecemeal liquidation value scaled by the face value of the bank's debt. At the beginning of every auction, the bankruptcy trustee publishes an industry estimate of the piecemeal liquidation value, which we use directly. Bidders appear to rely on this estimate as well: when the auction does lead to piecemeal liquidation, the average price paid by the winning bidder is close to (on average 8% above) the trustee's estimate. In contrast, when the bankrupt firm is purchased as a going concern, the average auction premium more than doubles the trustee's piecemeal liquidation value estimate.

In theory, the greater the liquidation recovery rate, the lower is the incentive to overbid and, in turn, the lower is the expected premium paid by the winning bidder. We find that when the firm is sold as a going concern, auction premiums are decreasing in the liquidation recovery rate, as predicted. Equally important, there is no evidence of overbidding in auction premiums for subsamples where the theory implies zero overbidding incentive. There are two such subsamples, representing 40% of the total sample. The first subsample consists of auctions that result in a going-concern sale, but where the bank's debt is not impaired at the ex ante liquidation value. In this case, the bank has no incentive to push for a coalition-bidding strategy since any revenue increase flows to junior creditors. Second, overbidding for targets that are expected to be liquidated piecemeal is suboptimal because these targets have only negligible going-concern value. When no bidder values the target as a going concern, there are no rents to be extracted from rival bidders through overbidding. We detect no evidence of overbidding in either subsample. Overall, our test results provide surprisingly robust support for our coalition overbidding theory.

Our main overbidding theory presumes that that bank managers act in the interest of the bank's shareholders by attempting to maximize auction revenue. We also consider an alternative coalition-bidding strategy, dubbed “debt rollover”. Here, entrenched bank managers use the first-price auction to roll over impaired bank debt at face value in order to hide non-performing loans from their superiors. Under the rollover hypothesis, the coalition's bid equals the face value of the bank debt (plus any debt senior to the bank), effectively establishing a price floor in the auction. While the rollover hypothesis also implies overbidding, it predicts an auction revenue that is approximately equal to the face value of the bank's debt when the coalition wins. The data rejects this prediction.

Overbidding results in allocative inefficiency whenever the bank–bidder coalition wins against a higher-valuation bidder. Thus, a second objective of our empirical analysis is to shed light on whether the auction system tends to produce inefficient outcomes due to overbidding. Previous work shows that firms restructured under Swedish bankruptcy auctions typically perform at par with non-bankrupt industry rivals (Eckbo and Thorburn, 2003). However, the previous evidence does not condition on bank financing and overbidding incentives. The intersection of bank financing of the winning bidder and high overbidding incentives is the ideal place to look for ex post allocative inefficiency. We find post-bankruptcy operating performance to be at par with industry rivals for these subgroups of auctions as well. If anything, the operating performance is higher in the subsample where the bank–bidder coalition wins and overbidding incentives are high ex ante.

Finally, we show that the probability of bankruptcy refiling over the 2 years following the auction is significantly greater when the bank finances the winning bidder. This refiling probability is, however, independent of overbidding incentives and therefore cannot be attributed to inefficiencies in the auction per se. A consistent interpretation is that banks tend to finance relatively high-risk restructuring prospects. In sum, bank financing of bidders increases auction liquidity, and seem to result in overbidding without significantly distorting the restructuring process.

The paper is organized as follows. Section 2 develops the coalition-bidding model and the model's central empirical implications. Section 3 describes the data selection and sample characteristics. Section 4 performs the main cross-sectional tests for overbidding. Section 5 shows post-bankruptcy performance statistics, while Section 6 concludes the paper.