A taxonomy of contract enforcement mechanisms

Ellickson (Reference Ellickson1991) provides a useful taxonomy of mechanisms of ‘social control’, distinguishing them on the basis of who applies the sanction for breaking rules. His framework can be applied to contract enforcement.

1. (1) Morality provides first-party enforcement: a party in breach of contract sanctions himself by developing a bad conscience. This has a bilateral aspect since the other party, too, must recognise and acknowledge the moral rules and the first party's adherence to them. Since a truly internal sanction excludes hypocrisy, parties need to detect each other's actual moral character, which typically requires personal experience of cooperation.

2. (2) Threatening to discontinue cooperation is the most important form of sanction applied by the contracting party who suffers from a breach of contract.Footnote ¹ Such second-party enforcement is the basis for a self-enforcing contract (Baker et al., Reference Baker, Gibbons and Murphy2002; Telser, Reference Telser1981). The sanction is particularly strong if future transactions in the same relationship are highly valuable and the parties plan for the long term. Morality and self-enforcement can function without direct help from the social environment. Parties get to know each other and reveal their moral qualities. By doing so, they invest in increasing the value of cooperation, which they will not want to lose later on. Thus, productive relationships are built and sustained.

In any well-functioning economy, these two basic enforcement mechanisms are, to some extent, complemented or replaced by informal mechanisms that rely on third-party enforcement.

1. (3) Community norms imply sanctions imposed by members of a community to which one or both parties belong. Sanctions may take the forms of disapproval, unfriending or ostracism for dishonesty or unreliability. While community norms have a moral aspect and may be partially internalised (Cooter, Reference Cooter1994), they assume external application by others. For this reason, they only contribute to enforcement if a community exists and if it matters to the sanctioned party.

2. (4) Reputation is another informal third-party mechanism.Footnote ² Here, the sanction is the loss of valuable future transactions, and it is applied by potential future contracting parties (Milgrom et al., Reference Milgrom, North and Weingast1990). In some cases, reputation may overlap empirically with community norms if future business partners are also community members (e.g. Bernstein, Reference Bernstein1992; Landa, Reference Landa1981) but it is a distinct mechanism in principle and often in reality (Klein, Reference Klein1997). Reputation has an enforcement effect especially if a party is widely known, plans to remain in business, and credible information about his contractual behaviour spreads quickly. Reputation may be personal, belonging to the entrepreneur, or impersonal, belonging to the firm. It is worth distinguishing between the two because transferring reputation from individuals to impersonal entities is a crucial aspect of modern economies, which enables transactions beyond personal ties (Greif, Reference Greif2006).

3. (5) In any advanced economy, formal legal rules for contractual breach exist and are applied by the courts as specialised third parties. Besides the courts and the general order of private law, government agencies could also be considered as enforcers of contracts (Hendley et al., Reference Hendley, Murrell and Ryterman2000; Milhaupt and Pistor, Reference Milhaupt and Pistor2008). However, they tend to be sector-specific and our focus here is on general patterns in an economy.

Although the list could be extended, we focus on these as the most fundamental means of contract enforcement in any advanced economy (cf. Brousseau, Reference Brousseau, Brousseau and Glachant2008; Fafchamps, Reference Fafchamps2004; Greif, Reference Greif, Ménard and Shirley2008; Hendley and Murrell, Reference Hendley and Murrell2002). Private-order organisations such as business clubs, professional associations or chambers may also provide rules and sanctions for contract enforcement (e.g. Greif, Reference Greif, Ménard and Shirley2008; Prüfer, Reference Prüfer2015; Pyle, Reference Pyle2005). A more detailed analysis could take them into account explicitly. Here, we presume that, as part of the firms’ institutional environment, they affect business relationships through one or more of the basic enforcement mechanisms above.

Empirical evidence on contract enforcement

Within the large literature on the economics of contract enforcement, three strands have particular relevance to our questions. First, a burgeoning literature discusses the interactions between different contract enforcement mechanisms. While insightful, it has two important limitations: (1) it focuses on interactions between just two or three mechanisms and (2) explores their marginal effects on each other. By far the greatest attention has been devoted to the relationship between legal enforcement and the use of morality and self-enforcing contracts that constitute ‘relational governance’ (Poppo and Zenger, Reference Poppo and Zenger2002). One argument, corroborated by some evidence in experimental settings (Gächter and Falk, Reference Gächter, Falk, Zwick and Rapoport2002; Malhotra and Murnighan, Reference Malhotra and Murnighan2002), is that external sanctions may ‘crowd out’ internal motivations to cooperate. Another line of reasoning is that contract-specific investments tend to make the termination of a contract very costly and the threat to go to court non-credible. In such cases, relational or self-enforcing contracts (Macneil, Reference MacNeil1978; Telser, Reference Telser1981) and bilateral governance (Williamson, Reference Williamson1979) may be preferred. By contrast, some economic models (Baker et al., Reference Baker, Gibbons and Murphy2002; Crocker and Masten, Reference Crocker and Masten1991) and recent management scholarship (Lazzarini et al., Reference Lazzarini, Miller and Zenger2004; Poppo and Zenger, Reference Poppo and Zenger2002) suggest that the threat of judicial enforcement is in fact important in securing complex, uncertain, long-term transactions, formerly considered the domain of relational governance. Rather than undermining trust, it may support its creation or even perpetuation by providing clear ‘threat points’ and ‘last resort’ sanctions. Overall, most empirical studies to date support the thesis of marginal complementarity between legal enforcement and relational governance but the question is far from settled (Cao and Lumineau, Reference Cao and Lumineau2015).

We know even less about links between the uses of other contract enforcement mechanisms, which few works discuss. Courts can support the mechanisms of reputation by providing reliable information about business conduct (Milgrom et al., Reference Milgrom, North and Weingast1990). Vice versa, effective reputational mechanisms may make up for the inefficiencies of a legal system and make reliance on the latter more likely (Woodruff, Reference Woodruff, Kornai, Rothstein and Rose-Ackerman2004). In other cases, the availability of legal sanctions can make reliance on reputation less necessary (Johnson et al., Reference Johnson, McMillan and Woodruff2002). Community norms may be enhanced (Cooter, Reference Cooter1994) or replaced (Johnson et al., Reference Johnson, McMillan and Woodruff2002) by legal enforceability.

A second strand of literature focuses on contract enforcement in specific segments of the economy. An important genre is the case study that shows how a market is characterised by a peculiar institutional configuration. Some of the best-known studies highlight special industries that rely heavily on informal mechanisms of norms and personal reputation as opposed to contract law, which is thought to govern the broader economy (e.g. Bernstein, Reference Bernstein1992, Reference Bernstein2001; Landa, Reference Landa1981; Woodruff, Reference Woodruff1998). This view is reinforced by a theoretical argument: communities or industries may have to choose between formal (legal) and impersonal enforcement and informal enforcement (Dixit, Reference Dixit2003; Greif, Reference Greif1993; Kranton, Reference Kranton1996).

Such reasoning relies, often implicitly, on what we might call the classical view of development: while traditional societies tend to conduct economic transactions in closely knit communities, developed economies are dominated by impersonal market relations (Greif, Reference Greif2006; Hayek, Reference Hayek2013[1982]; Tönnies, Reference Tönnies1957[1887]). Underdeveloped countries are often characterised by a dysfunctional public order, which forces entrepreneurs to rely on informal substitutes for law (Macmillan and Woodruff, Reference Macmillan and Woodruff2001; Wank, Reference Wank1996). As an economy develops, while informal ‘pockets’ may survive, there is a broad institutional shift from informal and personal solutions towards formal law and impersonal corporate reputation (North, Reference North1990; Peng, Reference Peng2003; Weber, Reference Weber1927). Accordingly, the creation of an effective formal legal system is seen as a crucial factor of economic development (Djankov et al., Reference Djankov, La Porta, Lopez-de-Silanes and Shleifer2003; La Porta et al., Reference La Porta, Lopez-de-Silanes, Shleifer and Vishny1998; Levine et al., Reference Levine, Loayza and Beck2000; Pistor et al., Reference Pistor, Raiser and Gelfer2000). However, many scholars since Macaulay's seminal work (Reference Macaulay1963) have stressed that informal contracting is present in a broad range of business settings even in the most developed economies (Arrighetti et al., Reference Arrighetti, Bachmann and Deakin1997; see Hadfield and Bozovic, Reference Hadfield and Bozovic2016 for an overview). Moreover, even in sectors where informal institutions dominate, they may be operating in the shadow of law (Deakin et al., Reference Deakin, Gindis, Hodgson, Kainan and Pistor2017).

A third, rather thin, line of studies is most relevant to our endeavour. They seek to provide a comprehensive view of the most important contract enforcement mechanisms in a country's entrepreneurial economy. Fafchamps (Reference Fafchamps1996) and Kähkönen and Meagher (Reference Kähkönen and Meagher2001) investigated African countries where formal institutions were basically absent. A few studies conducted during the postcommunist transition in the 1990s in Eastern Europe are of more immediate interest. Hendley et al. (Reference Hendley, Murrell and Ryterman2000) asked managers of Russian manufacturing firms to evaluate the importance of various methods for enforcing the contractual promises of suppliers and buyers. Correlations between mechanisms within relationships showed that personal trust (morality) and self-enforcement were closely linked but independent of other mechanisms. Formal institutions (e.g. courts, governments) and third-party informal mechanisms (e.g. personal ties and business reputation) were used together in different combinations that reflected the firm's relationship with the former Soviet state sector, corresponding with the still transitional state of the Russian economy in the late 1990s. A survey using the same method was conducted in Romania in 2001, and its results were analysed by Hendley and Murrell (Reference Hendley and Murrell2002) and Murrell (Reference Murrell2003). Extracting the principal components of the institutional variables revealed three independent aspects of strategic choice among contract enforcement institutions: (1) bilateralism, i.e. the joint use of personal trust and self-enforcing contracts; (2) reliance on the legal system; and (3) a decision about the aggregate use of contract-supporting mechanisms in general. Regression analysis suggested that, as in Russia, the use of mechanisms was driven by the closeness of the firm to the former state sector.

The results of Hendley et al. (Reference Hendley, Murrell and Ryterman2000), Hendley and Murrell (Reference Hendley and Murrell2002) and Murrell (Reference Murrell2003) are highly specific to the transitional turmoil after the collapse of communism. A more general caveat is that an analysis of correlations and principal components assumes that the relationships among enforcement mechanisms are linear (and, thus, monotonic). The theoretical ambiguity of interactions between any two mechanisms questions the validity of this assumption. Indeed, the only clear finding in the transitional context was the joint use of morality (or personal trust) and self-enforcing contracts, forming two aspects of bilateral governance. We follow these authors’ work in identifying the patterns of reliance on a comprehensive set of enforcement mechanisms but move beyond their analysis by applying the more suitable technique of latent class analysis, which posits no specific functional form for the relationships among enforcement variables.

The economic and legal environment

Situated in Eastern Central Europe, Hungary has a highly developed legal system. It ranked eighth in the world in the category of ‘enforcing contracts’ by the World Bank's Doing Business Survey in 2016.Footnote ³ An OECD member, Hungary is one of the less well-off countries in the European Union, with per capita GDP at 65% of the EU average (in 2014, at ppp). Although the country was occupied by the Soviet Union after World War II and had to endure communist rule for decades, it belongs historically to Central Europe in terms of religion, culture and economy. Like other countries in the region, Hungary saw a functioning institutional order of markets re-emerge roughly by the turn of the millennium (Beck and Leaven, Reference Beck and Laeven2006; Campos, Reference Campos1999; Crafts and Kaiser, Reference Crafts and Kaiser2004; Murrell, Reference Murrell2001, Reference Murrell, Ménard and Shirley2008). This was reflected in the relatively high expert scores on the rule of law and the effectiveness of corporate and bankruptcy law as well as firms’ perceptions that the legal system would protect their property rights and enforce their contracts (Pistor et al., Reference Pistor, Raiser and Gelfer2000). While these measures are much higher than in most post-Soviet countries further east, general societal perceptions about the rule of law tend to be weaker than in Western Europe (Kaufmann et al., Reference Kaufmann, Kraay and Mastruzzi2009). As for the general level of trust, 25% of respondents to the World Values Survey claimed that ‘most people can be trusted’ – less than in Western Europe (37%) but more than in the former socialist countries on average (19%) (Tóth, Reference Tóth2010).

Firms and business relationships in the sample

A nationwide survey was conducted among 300 privately owned small and medium-sized enterprises (with 5–49 employees) in Hungary.Footnote ⁴ In April and May 2011, personal interviews with executive managers were carried out in seven Hungarian cities, including its capital city (Budapest), three mid-sized cities in East Hungary and another three cities in West Hungary. Locations were chosen to cover all major regions of the country and all major sectors, except agriculture,Footnote ⁵ roughly in proportion to their contribution to national income (29% manufacturing, 32% commerce and 39% services). Within cities and sectors, companies were chosen randomly from the database of the official firm registry. Overall, the survey sample can be considered as fairly representative of the country's population of non-agricultural small and medium enterprises. As such, it belongs to a very limited number of surveys about contract enforcement mechanisms, with at least some claim to national representation (Hendley et al., Reference Hendley, Murrell and Ryterman2000; Johnson et al., Reference Johnson, McMillan and Woodruff2002; Lu and Tao, Reference Lu and Tao2009; McMillan and Woodruff, Reference Macmillan and Woodruff2001; Murrell, Reference Murrell2003; Steer and Sen, Reference Steer and Sen2010).

Managers were asked questions about their company's experience with two firms: one that they considered a ‘typical supplier’ and another, considered a ‘typical buyer’. The questionnaire focused on the transactional characteristics of their relationships and their reliance on various enforcement mechanisms to safeguard their contracts. The responses covered a diverse array of suppliers and buyers. While the majority were other Hungarian small and medium enterprises, many relationships with large Hungarian-owned enterprises, multinationals and foreign firms were mentioned.Footnote ⁶ This diversity was also reflected in the geographical distance between the interviewed firm and its business partner. Hence, our sample allows us to examine contracts both within and beyond the local environments of small and medium-sized firms. Most managers equated ‘typical’ business partners with long-standing ties. Only 10% of the relationships were two years old or younger. On the one hand, it is an important finding in itself that long-term business ties are typical for firms throughout the economy. On the other hand, such responses limit our sample to relatively long-term contractual relationships and exclude novel and ephemeral dealings. Extending survey data to newly founded and occasional relationships may provide additional insights in the future.

The business relationships were very diverse in terms of transactional characteristics, including the recurrence of transactions, transactional uncertainty and asset specificity (cf. Masten and Saussier, Reference Masten and Saussier2000; Murrell, Reference Murrell2003; Williamson, Reference Williamson1979). Although most relationships had a long history, the parties’ dependence on each other and the degree of exchange hazards varied considerably. Thus, the sample includes a mixed array of relationships along the continuum between the extremes of easy-to-replace market transactions and virtual bilateral monopolies (Williamson, Reference Williamson1979). Despite its limitations, our data set covers a broad range of contractual governance forms in Hungary's entrepreneurial economy.

Contract enforcement mechanisms

Managers were asked to rate the importance of six contract enforcement mechanisms for safeguarding their two typical contractual relationships: morality, self-enforcing contracts, community norms, personal reputation, impersonal market reputation and the law. Table 1 shows the questions used to identify the mechanisms and summarises the distributions and averages of evaluations. Although our data cover all major mechanisms, a caveat concerning social norms is in order. A businessperson may belong to several (overlapping) communities, in which contract-supporting norms may develop. Our focus is on the closest and most fundamental social community in which businesspeople are embedded: friends and the extended family. Nonetheless, the norms of other communities (e.g. neighbourly, religious, or professional) may well matter but are not examined here due to lack of data.

Table 1. The importance of contract enforcement mechanisms: distributions and average values

* The variables are measured on an ordinal scale and have no arithmetic means in the strict sense. Nonetheless, the average of the response numbers denoting the ordinal categories is suggestive of the average importance of an enforcement mechanism.

All mechanisms were used by a substantial number of firms. Morality and self-enforcement stand out: they were deemed important or very important in approximately 90% of relationships. Personal reputation, the legal order and impersonal market reputation were less widely relied upon: they were perceived important or very important in 40–50% of all relationships. Norms based on personal (friendly or familial) ties were used least (by less than 20%).

The findings reflect the relatively highly developed character of the country's legal system and the lack of importance of kinship and community, compared with East Asia's economies (McMillan and Woodruff, Reference Macmillan and Woodruff2001; Steer and Sen, Reference Steer and Sen2010; Upham, Reference Upham2002). It also reinforces the view that informal, highly personal contract enforcement mechanisms are widespread and important not only in transitional environments with a relatively weak rule of law (Hendley et al., Reference Hendley, Murrell and Ryterman2000; Johnson et al., Reference Johnson, McMillan and Woodruff2002). but also in countries with highly developed legal systems (Arrighetti et al., Reference Arrighetti, Bachmann and Deakin1997; Macaulay, Reference Macaulay1963).

Correlations between pairs of mechanisms show that the survey questions indeed grasp distinct mechanisms (Table 2). The strongest correlation is between morality and self-enforcing contracting. This is consistent with earlier findings that parties must rely on these two mechanisms if they are to manage their relationship by bilateral governance (Hendley et al., Reference Hendley, Murrell and Ryterman2000; Murrell, Reference Murrell2003), without third-party support.

Table 2. Correlations between contract enforcement mechanisms

* Significant correlations at 5%, with Bonferroni-correction.

Ideal-types and the method of latent class analysis

Firms engage in diverse strategies to govern their relationships. Governance always has a structure, i.e. an ‘institutional matrix within which transactions are negotiated and executed’ (Williamson, Reference Williamson1979: 239). Although governance may not take the form of ‘discrete structural alternatives’ (Williamson, Reference Williamson1991), we should still be able to identify ideal-types of governance structures with which real-life relationships bear a more or less close resemblance (Ebers and Oerlemans, Reference Ebers and Oerlemans2016; Nee, Reference Nee1992). We therefore assume that it is possible to demarcate distinct types of governance in the population of business relationships, and for each relationship to find the type of governance to which it is closest. To put it differently, contractual relationships can be clustered into classes, each class corresponding to an ideal-type of governance structure. Following legal theory and transaction cost economics, we also assume that each governance structure relies on certain combinations of enforcement mechanisms rather than others (Hendley et al., Reference Hendley, Murrell and Ryterman2000; Macneil, Reference MacNeil1978; Williamson, Reference Williamson1979).

The analytical challenge is to identify the unobserved (i.e. latent) ideal-types of governance, given the observations about the use of contract enforcement mechanisms. Statistically, we need to group contractual relationships into classes, each of which is characterised by a distinct configuration of enforcement mechanisms. We employ the method of latent class analysis (LCA) to identify the number of classes of governance structure, class sizes and the patterns of enforcement mechanisms typical of each class.

LCA is a statistical method especially developed for categorical data and used widely in the social sciences (Hagenaars and McCutcheon, Reference Hagenaars and McCutcheon2002). It can be thought of as an analogue of the factor analysis model for categorical observed and latent variables. While factor analysis models the observed variables as linear combinations of hidden factors, a latent class model uses a mixture distribution that is qualitative rather than linear (Agresti, Reference Agresti2002). It does not presume that enforcement mechanisms have monotonic relationships, which is a crucial advantage for our analysis.

A latent class model of governance strategies

In each business relationship, we measure the levels of reliance on the six enforcement mechanisms introduced above (morality, self-enforcing contract, etc.), j = 1, …, 6. The level of reliance is given by a categorical answer to the question if a given mechanism is important: r = {1 = ‘No’, 2 = ‘Rather no’, 3 = ‘Rather yes’ or 4 = ‘Yes, very much’} (see Table 1). Let R_ij = [R _ij1, R _ij2, R _ij3, R _ij4] be a vector of binary variables that represents the level of reliance on mechanism j in business relationship i, in which R _ijr = 1 if the level of reliance is r, and 0 otherwise. Thus, each business relationship is described by a set of six vectors: R_ij, j = 1, …, 6, with multinomial distributions.

There is covariation among these observed vectors if the levels of reliance on enforcement mechanisms are not independent. We assume that this non-independence is due to a latent categorical variable that captures the underlying governance structure for each transaction. If there are M classes of governance structure, they can be represented by a vector G_i = [G _i1, G _i2, … G _iM], in which G _im = 1 if the structure of relationship i belongs to class m∈M, and 0 otherwise.

As a crucial feature of latent class modelling, the theoretical assumption that the latent variable explains the relationships among the observed variables is translated into the formal assumption that the observed levels of reliance on enforcement mechanisms are conditionally independent of each other, given the class of the latent governance strategy (Agresti, Reference Agresti2002). In other words, the levels of reliance on different enforcement mechanisms are mutually independent within each class of governance. Formally, it holds for each possible (r_i1, …, r_i6) outcome of the vectors of enforcement mechanisms (R_i1, …, R_i6) and each governance class g_i∈G_i that:

(1)$$P\lpar {{\bi R}_{i1} = {\bi r}_{i1},\; \ldots \;, {\bi R}_{i6} = {\bi r}_{i6}{\rm \vert} {\bi G}_{\bi i} = {\bi g}_{\bi i}} \rpar = \prod \nolimits_{\,j = 1}^6 P({\bi R}_{ij} = {\bi r}_{ij} \vert {\bi G}_{\bi i} = {\bi g}_{\bi i}).$$

The class-conditional probability that relationship i relies at level r on enforcement mechanism j is π _jrm = P(R _ijr = 1|G _im = 1). The prior probability that the relationship belongs to class m is p _m = P(G _im = 1). By conditional independence, the probability that relationship i produces a particular set of R_i = [R_i1, …, R_ij, …, R_i6] observations is given by the following probability mass function:

(2)$$\; f\lpar {{\bi R}_i} \rpar = \sum \nolimits_{m = 1}^M p_m \prod \nolimits_{\,j = 1}^6 \prod \nolimits_{r = 1}^4 \pi _{\,jrm}^{R_{ijr}} $$

The model is estimated by maximising the log-likelihood for N observations over the parameters π _jrm and p _m:

(3)$$L = \sum \nolimits_{i = 1}^N ln\left( {\sum \nolimits_{m = 1}^M p_m \prod \nolimits_{\,j = 1}^6 \prod \nolimits_{r = 1}^4 \pi_{\,jrm}^{R_{ijr}}} \right)$$

The analysis is performed using the poLCA package for R developed by Linzer and Lewis (Reference Linzer and Lewis2011, Reference Linzer and Lewis2013), which applies the expectation maximisation (EM) algorithm (Dempster et al., Reference Dempster, Laird and Rubin1977) to maximise the log-likelihood function. The model's estimations can be used to characterise the typical governance structure of each latent class as well as the shares of the latent classes in the population. The governance structure is described by the estimated conditional probabilities of the levels of reliance on all six enforcement mechanisms in that class (π _jrm). There are two ways to gauge class sizes. First, the model estimates the unconditional probabilities of class memberships (p _m), i.e. the probabilities that any business relationship in the population belongs to different latent classes. Second, the model's estimations allow us to predict the class membership of every relationship in the sample. Once the model parameters are estimated, the conditional probability that relationship i belongs to class l∈M, given its observed levels of reliance on different enforcement mechanisms (R_i), can be obtained by applying Bayes’ formula:

(4)$$P(G_{il} = 1 \vert {\bi R}_i) = \displaystyle{{\,p_l\mathop \prod \nolimits_{\,j = 1}^6 \mathop \prod \nolimits_{r = 1}^4 \pi _{\,jrl}^{R_{ijr}}} \over {\mathop \sum \nolimits_{m = 1}^M p_m\mathop \prod \nolimits_{\,j = 1}^6 \mathop \prod \nolimits_{r = 1}^4 \pi _{\,jrm}^{R_{ijr}}}} $$

Each relationship can then be assigned to the class in which it belongs with the greatest Bayesian probability. We also estimate the sizes of the latent classes by this method.

The model does not estimate the number of classes: this is fixed before the estimation. How to decide their appropriate number? For LCA, a standard method is to use an information evaluation criterion (Linzer and Lewis, Reference Linzer and Lewis2011; McCutcheon, Reference McCutcheon, Hagenaars and McCutcheon2002). Such a criterion jointly evaluates how well the model estimations fit the data and the number of estimated parameters. More classes (i.e. more parameters) tend to provide a better fit to the data but at the risk of fitting to noise. Therefore, an appropriate evaluation criterion will penalise for a larger number of parameters.

There are several criteria in use. The best known is the Akaike Information Criterion: AIC = −2L + 2Φ, where L is the log-likelihood and Φ is the number of estimated parameters. A drawback of AIC is that it is not consistent as the sample size grows (Bozdogan, Reference Bozdogan1987). Simulation results also cast some doubt on its accuracy, which tends to worsen with sample size, and reveal its tendency to overestimate the number of classes (Nylund et al., Reference Nylund, Asparouhov and Muthén2007; Yang, Reference Yang2006). Two Bayesian information criteria achieve consistency by including a logarithmic function of the sample size. These are Schwartz's Bayesian Criterion (SBC = −2L + ΦlnN)Footnote ⁷ and Bozdogan's Consistent AIC (CAIC = −2L + Φ(1 + lnN) (Bozdogan, Reference Bozdogan1987; Linzer and Lewis, Reference Linzer and Lewis2011). While no criterion is superior in all contexts, simulation results suggest that the Bayesian criteria are more accurate for relatively simple latent class models like ours (Lin and Dayton, Reference Lin and Dayton1997; Nylund et al., Reference Nylund, Asparouhov and Muthén2007). Therefore, we prioritise them over AIC in cases of conflicting results. SBC is to be preferred to CAIC when class sizes are unequal or there are relatively few cases per class (Nylund et al., Reference Nylund, Asparouhov and Muthén2007). We must add that while formal evaluation criteria are helpful in deciding the reasonable number of latent classes, they cannot entirely replace the researchers’ judgement. The estimated classes must be sufficiently different and economically meaningful.

Types of governance structures

We first estimated a benchmark case with the number of latent classes fixed at one, i.e. assuming that the observed enforcement variables are independent in the whole population. Then we began increasing the number of classes, one at a time. We stopped at five classes, when the smallest class shrank to a very small size (7%). This choice was also reinforced by the substantive analysis of the model estimations. Table 3 summarises the comparative evaluation of the estimated models with between one and five latent classes.Footnote ⁸ As expected, the model with one latent class fares worst. The Akaike criterion points to five classes. However, the two more reliable Bayesian criteria, SBC and CAIC, both select the model with three classes. Therefore, we focus on the estimates of the three-class model. Our main findings would not change substantially if we chose the estimates with four and five classes. They reveal similar general patterns among governance structures but fragment them into more classes.Footnote ⁹

Table 3. Evaluation of latent class models with between one and five classes

Note: The models chosen by the information criteria are highlighted.

Recall that managers were asked to evaluate the importance of six contract enforcement mechanisms on a scale from 1 (unimportant) to 4 (very important). Each latent class is characterised by a distinct mixture of the conditional probability distributions of responses for all mechanisms. Table 4 summarises the results for three latent classes. For relationships in class 1, the probability that a manager answered ‘no’ to the question whether morality was an important enforcement mechanism was 27%; the probabilities of his answer being ‘rather no’, ‘rather yes’ and ‘yes, very much’ were 0, 21 and 52%, respectively. The same probabilities are given for each mechanism in each latent class. The rightmost column contains the ‘average’ response (on the scale 1 to 4), which facilitates the comparison of the relative importance of mechanisms. A sum of these scores signals the overall reliance on all enforcement mechanisms. One could think of this as a rough measure of enforcement intensity. Greater intensity means that either more mechanisms are relied upon or those that are used are relied upon more heavily. Differences in enforcement intensity are to be expected since some transactions pose a greater threat of opportunism than others (Williamson, Reference Williamson1979).

Table 4. Conditional probability distributions of contract enforcement mechanisms for three latent classes

Note: To the question ‘Is the mechanism important?’, replies were: 1 = ‘No’, 2 = ‘Rather no’, 3 = ‘Rather yes’ or 4 = ‘Yes, very much’; modal probabilities are highlighted.

Business relationships that fall into class 1 are characterised overall by a relatively low level of reliance on contract enforcement mechanisms. Among the mechanisms used, morality and self-enforcing contracting clearly stand out, while third-party informal or formal methods have no or little weight. Therefore, it seems apposite to describe the relationships in this class as having a bilateral governance structure.

Class 2 is characterised by the greatest reliance on enforcement mechanisms overall. Bilateral means (morality and self-enforcement) are even more important than in class 1 but are complemented by increased reliance on all third-party mechanisms: personal reputation, community norms, impersonal reputation and law. This suggests the existence of a comprehensive governance structure, relying relatively heavily on a broad range of enforcement mechanisms.

Relationships in class 3 are less likely to have a strongly bilateral character than those in class 1. Reliance on morality and self-enforcing contracts is less pronounced. Instead, both personal and impersonal third-party mechanisms are used to a greater extent than in class 1 (but less than in class 3, with the exception of community norms, which are, however, rarely important). This implies a third distinct structure of third-party governance. We should bear in mind, however, that bilateral aspects are also present in these relationships, only somewhat less so than in the other two classes.

Population shares of governance structures

How large are the classes identified? The estimated unconditional probabilities of class membership (p _m) are given in the upper part of Table 5, together with standard errors. The numbers show that a randomly chosen relationship uses a bilateral governance structure with a probability of 16%. The same probabilities for comprehensive and third-party governance are 59.5% and 24.5%, respectively.

Table 5. Population shares of governance classes

As we noted, an alternative way to estimate the population shares of governance structures is to assign every relationship in the sample to the class in which it belongs with the highest (Bayesian) probability, given its observed reliance on different enforcement mechanisms (P(G _il = 1|R_i)). The average of this probability for the sample is 90%, which signals that the model allocates the business relationships among the three classes with a high degree of certainty.Footnote ¹⁰ The class shares calculated in this fashion are given in the lower part of Table 5. They are close to the model's estimations of class sizes. Both calculations show that comprehensive governance dominates, followed by third-party and then bilateral governance.

What do our results reveal about the patterns of contract enforcement mechanisms among Hungarian enterprises? The first thing to note is that morality and self-enforcing contracts contribute to enforcement in virtually all business relationships in our sample. It seems impossible to do without them. They comprise the two fundamental mechanisms that business parties can create and manage bilaterally, without the direct involvement of other social actors. For less than a sixth of business relations, they are by and large sufficient on their own.

The rest of business relationships rely on a mixture of morality, self-enforcement and both informal and formal third-party mechanisms. That is, when the parties move beyond bilateral solutions and turn to their social environment for assistance, they tend to rely on several mechanisms rather than just one or two. In roughly a quarter of all relationships, morality and self-enforcement become somewhat subdued and are partially substituted by informal third-party mechanisms (community norms, personal reputation and impersonal market reputation) and law. In the majority of relationships, however, there is no such substitution effect. To the contrary, greater reliance on bilateral methods is complemented by greater reliance on third-party mechanisms. That is, firms choose to safeguard their contracts with a broad range of enforcement mechanisms.

Law is perceived by the firms as having no or little importance in roughly half of all relationships. Where it is considered important, it is always used in conjunction with several informal mechanisms, and never on its own. The decision whether to use it or not is most closely and positively associated with the use of informal third-party methods. It functions as an element of either third-party governance or comprehensive governance.