Simon Domberger, Steve Easton, Dennis Maki, Richard Murnane and Zane .... actions costs (see Stigler, 1976; Becker, 1983; Becker, 1985; Toumanoff, 1984). ... tive competition among subordinates for jobs and competition among bureau-.
Public Choice 73: 205-239, 1992. © 1992Kluwer Academic Publishers. Printed in the Netherlands.
Ownership versus competition: Efficiency in public enterprise* AIDAN R. VINING Faculty of Business Administration, Simon Fraser University, Burnaby, B.C., Canada V5A 1S6
A N T H O N Y E. B O A R D M A N Faculty of Commerce, University of British Columbia, 2053 Main Mall, Vancouver, B.C., Canada V6T 1 Y8
Received 11 June 1990; accepted 27 September 1990
1. Introduction Recently, several scholars have argued that competition in product markets is a more important determinant of allocative efficiency than whether a firm is publicly owned or privately owned. An extreme version o f this "primacy of competition versus ownership" argument is that even where competitive product markets are normatively appropriate (i.e., there are no significant market failures in the product market) ownership does not matter: there is " n o difference in allocative efficiency" between public and private firms in competitive markets. There are two theoretical ways of reaching this conclusion. The first emphasizes inefficiencies in private corporate (PC) ownership monitoring (hereafter, IPCM) while the second emphasizes efficiency in state-owned enterprise (SOE) political monitoring (hereafter ESOEM). Proponents of either explanation claim that the aggregate empirical evidence supports their view. For example, in an influential review, Borcherding et al. (1982: 136) conclude that "given sufficient competition between public and private producers (and no discriminatory regulations and subsidies), the differences in unit costs turn out to be insignificant" (see also Millward, 1982). Important policy consequences could flow from the widespread acceptance of the " n o difference in efficiency" conclusion, whatever its genesis. In partic* For a change the authors are listed in reverse alphabetical order. This research was supported by a grant from the Social Sciencesand Humanities Council (Canada). We wish to thank Joan Caravan and Odette Vaughan for research assistance. We have benefitted from discussions with Simon Domberger, Steve Easton, Dennis Maki, Richard Murnane and Zane Spindler. Usual caveat.
206 ular, it argues that introduction of product market competition in a given public enterprise market should largely eliminate the need for privatization.1 In this paper we will demonstrate that neither theoretical explanation of the " n o difference in efficiency" in competitive markets argument is convincing. Further, neither is supported by the relevant empirical literature. Specifically, this article will show that: (1) ownership is both theoretically and empirically important; (2) that most of the evidence that purports to show the "primacy of competition versus ownership" or " n o difference in efficiency" does not, and indeed cannot, convincingly do so; and (3) new empirical evidence, using Canadian data, confirms the importance of ownership. In short, ownership does matter for both technical and allocative efficiency.
2. Why the IPCM explanation is wrong The most common theoretical argument in support of the " n o difference" position is based on the claim that principal-agent problems in PCs are no worse than in SOEs such that the lack of a "market for corporate control" for SOEs generates no comparative disadvantage. The logical extension of this position is, as Whitehead (1988: 9) argues, "there are no inherent reasons why enterprises in private ownership should operate more efficiently than those in public ownership." In essence the PCs' market for corporate control doesn't work well. This argument implicitly assumes firms do not differ in their production of socio-political (distributional) outputs (in fact, most authors probably assume these outputs are zero, although if this is the case its hard to see why proponents would consider an SOE to be a relevant policy tool in the first place). 2 IPCM therefore concerns technical efficiency - movements of the production possibility frontier. Whitehead (1988: 7 - 8 ) states this case in detail: In practice, while by definition property rights do not control public sector operations, it is not clear that the comparison is so straightforward. Management in the private sector is in the hands of executives whose own objectives are not the same as those of the owners; owners do not always have adequate information to exercise control, and anyway ownership is diffuse, so that individual owners have little incentive to exercise control. Thus in both public and private sectors what we observe are agency relationships. There is a large literature on the nature of the market for corporate control and how it impacts on firm efficiency. The dominant research strategy to get at this question distinguishes between owner-controlled private firms (OCPC) and
207 managerially-controlled private firms (MCPC), which reflects differences in the effectiveness of owner property rights and the ability of managers to pursue their objectives (for a recent symposium review see the Journal of Law and Economics, June 1983). The IPCM argument ignores this distinction. Implicitly, IPCM argues that either: OCPC(efficiency) = MCPC(efficiency) = SOE(efficiency), or there are no OCPC firms and: MCPC(efficiency) = SOE(efficiency). Clearly, there are OCPC firms (see below). Therefore, to establish that private firms in aggregate (i.e., OCPCs and MCPCs combined) are more efficient than SOEs, viz.: PC(efficiency) > SOE(efficiency), we need only to establish that: OCPC(efficiency) > MCPC(efficiency).3 In this section we will discuss the theoretical and empirical evidence that supports this latter contention. Further, we maintain that "technological" innovations are making the market for corporate control more efficient so that MCPC(efficiency) is likely increasing, thus widening the gap between PC(efficiency) and SOE(efficiency), ceteris paribus. Property rights theory suggests that firms in which managers dominate (i.e., where there is a serious principal-agent problem) are likely to be less efficient and consequently less profitable than OCPC firms (Furubotn and Pejovich, 1972). Firms may differ in terms of economic profitability, X-efficiency (Leibenstein, 1966, 1975), strategy or other manifestations of inefficient managerial behavior. For example, one specific agency loss hypothesis is that MCPC firms engage in more conglomerate diversification than OCPC firms as managers cannot adequately diversify their human capital risk (because there is no market for it) except by diversifying the firm. This represents agency loss because shareholders could more directly and efficiently diversify their own risk (Larcker, 1988). While the exact specification of the managers' objectives has varied (see Baumol, 1967; Williamson, 1967; Fisher and Hall, 1969; Marris, 1974; Amihud and Lev, 1981), the resulting outcome is that MCPCs are likely to be less efficient than OCPCs. Empirical tests of property rights theory attempt to determine indirectly
208
. ~
0
~.~'~ ~:
0"."
~r
~
°~ r-, o0 ~
oo
°~
~
0
~
~o_oo
_ e-,
~ ~
-,..~ °~
..o
e~ 0
o
o 0
o
209 whether managers achieve "managerialist" goals (at least partially). It is hypothesized that such attainment necessarily occurs at the expense of profitability or some other efficiency outcome. Thus the test become one of determining whether OCs are more or less efficient than MCs. Operationally, these studies generally use share dispersal - concentrated or dispersed - as the indicator of OCPC and MCPC, respectively.4 The empirical evidence for a range of dependent variables, including profitability, is presented in Table 1. The weight of the evidence frequently indicates that O C P C firms are generally more profitable than MCPCs. We found no study where MCPCs were more profitable, fourteen studies where there was no significant difference in profitability and seventeen studies (plus one simulation) where OCPC firms were more profitable. 5 The evidence is broadly consistent for other dependent variables. There is thus some " g o o d news": since OCPC firms are more efficient than MCPCs then, in aggregate, PCs are more efficient than SOEs. But it also contains " b a d news" in that MCPCs continue to exist as a category, indicating the market for corporate control is not perfect. There are plausible theoretical reasons why M C P C inefficiency might persist because of specific takeover market failure, even given broadly efficient capital markets (see, for example, Grossman and Hart, 1980). However, recent financial innovations appear to have reduced takeover market inefficiencies and the viability of inefficient MCPC behavior. The rise of institutional investment has decreased the aggregate, marketrelevant dispersal o f shareholdings (Aoki, 1983). Additionally, capital market innovations such as " j u n k b o n d s " and leveraged buy-outs put tremendous pressure on under-performing firms. Takeover markets in both North America and Europe have been especially active over the last decade (Jarrell, Brickley and Netter, 1988; Dodd, 1986). One view is that "the market for corporate control is now largely complete. Few, if any American corporations are today beyond the reach of this market because of their size or scale" (Coffee, 1986: 5). In short, overt managerialist behavior in MCPCs is more risky than previously. Consequently, firms that can be classified as MCPCs are likely to act more like OCPCs thus increasing the difference between PC(efficiency) and SOE(inefficiency) in aggregate.
3. Why ESOEM is wrong The only other convincing way to reach the " n o difference in efficiency" conclusion is to argue that SOEs are technically and allocatively efficient (i.e., operate on the possibility production frontier) but they produce socio-political outputs (necessarily at the expense of profitability) which are not included in
210 standard efficiency studies. Further, SOEs produce these (multiple) outputs efficiently because the political market is competitive and generates results equivalent to those obtained by the market for corporate control (Wintrobe, 1987). In short, there is "efficient SOE political monitoring" (ESOEM). The gist of the first part of the argument is that SOEs "are often constrained to pursue employment and regional development goals; their contribution to these goals presumably aids the re-election prospects of government, but neither the social benefits nor the increased probability of re-election are listed on the firm's balance sheet" (Wintrobe, 1987: 437; see also Borcherding, 1988). This argument can be rebutted quickly. To the extent that the SOE that the SOE raises employment or wages or produces any other socio-political output as well as its "core" output(s), there must be a net deadweight loss (Boardman and Vining, 1989: 9-10). The gist of the second part of the argument is that the political market for SOE control is a close substitute for the market for PC corporate control. Specifically, Wintrobe (1987: 439) argues: The maj or limitations on management discretion [in private firms] are (1) the possibility of a takeover bid, and (2) competition from other m a n a g e r s . . . each of these mechanisms has its counterpart in the public sector. Elections may be the process by which opposition parties' 'bid" for control of the public sector. And there is competition among public sector managers as well as private ones. In other words, elections generate contestability at two levels. As Wintrobe makes his case at two levels we amplify and respond accordingly. But first we will make some general comments concerning the efficiency of political competition.
3.1. The efficiency of political competition Wintrobe draws parallels between political competition and economic competition (the market for corporate control) to claim that political competition necessarily results in socially efficient outcomes. This position has been maintained in the probabilistic voting literature. Mueller (1989: 214), for example, summarizes this literature thus: •.. if voters reward a candidate who promises them a higher utility, by increasing the likelihood of voting for the candidate, then competition for votes between candidates leads them 'as if by an invisible hand' to platforms that maximize social welfare. The analogy between market competition and
211 political competition does exist. Both result in Pareto-optimal allocations of resources. Peltzman (1990:61) also contends "(t)he voting market emerges as a strikingly efficient aggregator of economic information." Implicitly, Wintrobe's argument is a variant of the Panglossian result (dilemma) that "what is, is efficient." (see Dahlman, 1979: 150-156.) According to this view any apparent inefficiency is the result of ignored outputs or transactions costs (see Stigler, 1976; Becker, 1983; Becker, 1985; Toumanoff, 1984). The "what is, is efficient" proponents suggest that SOE inefficiency or "waste" is a puzzle, which political market competition should inevitably eliminate. These arguments, however, ignore rent-seeking and the associated deadweight losses (Tullock, 1967; Krueger, 1974; Posner, 1975; Tollison, 1982). Crew and Rowley (1988) have brilliantly summarized the public choice refutation of the "what is, is efficient" position (see also Mitchell, 1989). Following Buchanan (1969) they argue: The notion that what exists is efficient, given institutional constraints, is a logical implication of the axioms of rational b e h a v i o r . . . Institutional constraints, however, may well impede the attainment of efficiency gains potentially available within a different set of rules or constitutional conditions. In such circumstances, what is is not efficient if meta-level rules changes are feasible. (p. 54). Constitutional rules in the political market, while allowing competition, are not as efficiency-inducing as those of the market for corporate control. Rules that permit pressure group activity may generate a stable "competitive" equilibrium, but it is not efficient in the usual sense of the word. 6 There will be deadweight losses associated with rent-seeking. In short, "aiding the reelection prospects of the government" may represent "efficient" rent-seeking but this is not the same as an efficient social outcome. After an extensive review of the literature Tollison (1982: 594) concludes: (P)olitical competition under one man-one vote conditions does not lead to efficient outcomes in the same sense that such outcomes are produced by competition in private markets. One cannot therefore rely on supply-side forces in politics to generate efficient outcomes. There is, of course, a large literature on the failures of political competition. Two general examples will suffice. The political business cycle literature suggests that interactions between government and voters is not always efficient
212 (Mueller, 1989: 285-286). Second, most political market competition is nonprice competition, not price competition, which drives the market for corporate control - poor management capitalizes into a single variable, stock price (Carroll, 1990).
3.2. The analogy to the takeover market
As the above quote by Wintrobe clarifies, the specific variant of the efficient public market that applies to SOEs is that "contestability" generates efficient behavior. But contestability requires credibility. The only way the political market could directly replicate the market for corporate control is by imitating the takeover process, i.e., by threatening privatization. But such a threat must be credible. It can only be partially so, for several reasons. Governments are only in power for limited periods, therefore the time-frame of such a threat is limited. Arguably, the only credible threat to the marginal SOE is to have privatized some other SOE. Ironically, the most credible type of government is a left-ofcenter government with previous privatization experience (also, there is little likelihood of effective political opposition in this case). The New Zealand Labour government is one major example. Even here, however, there is a caveat. The emerging conventional wisdom is that privatization is most likely to succeed if employees are "bribed" to ensure minimum opposition (Burton, 1987). This may well provide incentives for inefficiency. X-inefficient SOEs are the ones government would most like to threaten with privatization ("takeover"). Government, therefore, will tend to offer bigger "bribes" to these employees. Rational government enterprise employees will, in such circumstances, continue to perform inefficiently (retain as much organizational slack as possible). They cannot lose. If they are not privatized, they continue to "consume" the benefits (e.g., "the quiet life"). If they are privatized, they are likely to receive the capitalized value of the slack. Thus, the threat is only credible if exercised. There is no credible threat that will provide efficiency to the SOE qua SOE.
3.3. The analogy to the market f or managers
Wintrobe (1987: 443) also argues there is another form of "market for corporate control" because public firm managers can be fired: " a government bureaucracy appears to be a competitive labor market in which there is effective competition among subordinates for jobs and competition among bureaucratic superiors for subordinates.., public managers who do not perform well
213 will be replaced." Again, he equates competition with efficient social outcomes. It should be obvious from our above discussion why we believe the market for SOE managers will be not be driven by allocative efficiency considerations. It is more likely to be driven by politician's desires to increase their re-election prospects. In contrast, the private sector market for corporate managers is largely driven by efficiency considerations (see discussion above in Section 2 and Coughlan and Schmidt, 1985; Weisbach, 1988; and Warner, Watts and Wruck, 1988). Notice that inefficiency arises even in the absence of principal-agent problems. Wintrobe apparently believes that these problems are less severe in SOEs than in PCs (see also Breton and Wintrobe, 1982). However, a large literature on public sector managerial rent-seeking suggests otherwise, implying further efficiency losses (Tullock, 1965; Migue and Belanger, 1974; Niskanen, 1975).
3.4. The analogy to contracting out Another version of the "contestability" argument, although not directly made by Wintrobe, is that when faced with "contracting out", SOEs can be as efficient as PCs. There is evidence that competitive tendering lowers costs and that public units can compete successfully for many of these contracts (see, e.g., Hartley and Hubey, 1985; Domberger, Meadowcroft and Thompson, 1986). The conventional interpretation of this literature is that competition induces efficiency improvements. Our interpretation is that, in practice, the form of ownership has changed from "public" to "private". If government is actually prepared to award contracts to private firms, it demonstrates that "public" providers have no greater (property) right to perform the service than a "private" firm. This is equivalent to saying that the "public" unit is essentially "private" (Vining and Weimer, 1990). This specific circumstance is indeed almost equivalent to a fully functional takeover market - ownership of the public firm is perfectly contestable. However, in practice, competitive contracting with no implicit or explicit property rights for the incumbent public providers is rare. It is interesting to note the semantics here. The main problem this paper seeks to address concerns the relative importance of ownership versus competition. In the extreme (perfectly contestable ownership) case, public ownership can be thought of as either ownership or competition. Many proponents of the "ownership is not important" position start out defending public ownership given product market competition but then slide into defending public ownership because of ownership competition (i.e., contestable ownership). After the slide they end up concluding that competition is central - possibly implying that product market competition is more important than ownership. In con-
214
g, ,~- t-.
~.
~
,_,
o
~,.~ OX ~
oo
c~
oo
~,
~
z 0
0 ¢.a t~ e~
0
~
8 ~Z~
~
=~
0 ~'x:l
-r, ca
ca
g,
0
,'v.. a:
CO .~0 ~ 0
0
I:~
215
O~
o ~
~
~
~
~_~
N~
2-
s ._= ,b.., 0'~
-: ~ N ,.~
0
0
lb, 0
0 0
,