Diversification Strategies and Firm Performance: A Sample Selection Approach

Santarelli, Enrico; Tran, Hien Thu Working Paper Diversification Strategies and Firm Performance: A Sample Selection Approach Quaderni - Working Paper DSE, No. 896 Provided in Cooperation with: University of Bologna, Department of Economics Suggested Citation: Santarelli, Enrico; Tran, Hien Thu (2013) : Diversification Strategies and Firm Performance: A Sample Selection Approach, Quaderni - Working Paper DSE, No. 896, Alma Mater Studiorum - Università di Bologna, Dipartimento di Scienze Economiche (DSE), Bologna, https://doi.org/10.6092/unibo/amsacta/3845 This Version is available at: https://hdl.handle.net/10419/159735 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc/3.0/ ISSN 2282-6483 Diversification Strategies and Firm Performance: A Sample Selection Approach Enrico Santarelli Hien Thu Tran Quaderni - Working Paper DSE N°896 1 Diversification Strategies and Firm Performance: A Sample Selection Approach Enrico Santarelli University of Bologna, Department of Economics; Piazza Scaravilli, 2 – 40126 Bologna, Italy; tel. +39 051 2098487; e-mail: [email protected] Hien Thu Tran University of Bologna, Department of Economics Piazza Scaravilli, 2 – 40126 Bologna, Italy; e-mail: [email protected] Abstract: This paper is based upon the assumption that firm profitability is determined by its degree of diversification which in turn is strongly related to the antecedent decision to carry out diversification activities. This calls for an empirical approach that permits the joint analysis of the three interrelated and consecutive stages of the overall diversification process: diversification decision, degree of diversification, and outcome of diversification. We apply parametric and semiparametric approaches to control for sample selection and endogeneity of diversification decision in both static and dynamic models. After controlling for industry fixed-effects, empirical evidence from firm-level data shows that diversification has a curvilinear effect on profitability: it improves firms’ profit up to a point, after which a further increase in diversification is associated with declining performance. This implies that firms should consider optimal levels of product diversification when they expand product offerings beyond their core business. Other worth-noting findings include: (i) factors stimulating firms to diversify do not necessarily encourage them to extend their diversification strategy; (ii) firms which are endowed with highly skilled human capital are likely to successfully exploit diversification as an engine of growth; (iii) while industry performance does not influence profitability of firms, it impacts their diversification decision and degree. Keywords: Diversification; Firm performance; panel data; sample selection; parametric and semiparametric models. JEL Codes: L21; L25; C14; C23. 2 1. Introduction What determines the optimal boundaries of the firm across industries? How does a firm expand from its core business into other product markets? These questions have raised substantial research interests from the initial landmark article The Nature of the Firm by Coase (1937) and book The Theory of the Growth of the Firm by Penrose (1959). Since then, different theories (resource-based view, transaction cost, agency theory, etc.) have been proposed to explain firm diversification (Chandler, 1962, Berry, 1975, Rumelt, 1974, Andrews, 1980). The early industrial organization literature has argued that no significant relationship exists between diversification and performance, meaning that, when entering new markets, existing firms have no special advantages (see, e.g., Gort, 1962, Arnould, 1969, Markham, 1973). More recently, researchers have shown that diversification generates multiple outcome directions depending on the degree of relatedness of a firm’s diversification activities (Palich et al., 2000; Qian, 2002). These studies share one common finding that the diversification/performance relationship follows a non-linear pattern: they are positively related up to a point, after which a further increase in diversification is associated with declining performance. Notwithstanding this change of perspective, scholars have so far mostly focused on the single causal effect of degree of diversification relatedness on firms’ subsequent performance, neglecting the whole diversification process that firms involve in until the final diversification outcome is recognized. The important question left unanswered is therefore: why not all firms engage in diversification activities or receive equally positive outcomes from their diversification strategies? Exploration of antecedent factors determining a firm’s likelihood to diversify as well as how much it can diversify upon its green-light decision might lead to an answer. First, we argue that it may not be appropriate to analyse the diversification/performance relationship in a single-equation framework, since it is strongly related to the pre-determined factors that induce firms to engage in diversification. Thus, we investigate the whole diversification process in three interrelated and consecutive equations: diversification decision (what determines a firm to diversify?); diversification degree (once a firm decides to diversify, what determines the degree of its diversification relatedness?); and diversification outcome (how does a firm’s diversification degree influence its profitability?). Since the three equations are by nature interrelated, we take into consideration the possibility that their disturbances are correlated, and thus, we need to be careful in choosing the right estimation methodology given the existence of unobservable characteristics that affect the three stages under study. Second, we are aware that ANOVAs or cross-sectional least squares regressions are inadequate to study the relationship between diversification and performance, since these methodological approaches treat the decision to diversify exogenously, i.e. firm-level and industry-level characteristics as well as the influence of the external environment are implicitly assumed not to have any effect on firms’ diversification strategies. Consistent with Maksimovic and Phillips (2002) and Lang and Stulz (1994) among others, who show that firm and industry characteristics influence a firm’s decision to diversify, we take into account the sample selection and endogeneity issues from correlated disturbances by applying advanced parametric and semiparametric estimation methods for both static and dynamic treatments of firm-level panel data. Initially, sample selection will be tested and corrected by four estimation approaches: the standard Heckman’s two-stage method, the Vella (1998) and Wooldridge (1995) parametric 3 approach, Heckman et al.’s (1998) kernel-based propensity score matching, the Semykina and Wooldridge (2010) model. Each of these approaches is employed in the first two stages: a binary choice selection equation determining whether or not the firm diversifies, and then conditional on its diversification decision, we observe its degree of relatedness to the firm’s core business; subsequently, endogeneity of diversification degree is controlled in the firm performance equation, under the hypothesis that unobserved individual-level and firm-level characteristics might influence both firms’ diversification decision and their subsequent profitability (given their diversification decision and diversification degree choice). For this purpose we apply the GMM and Arellano-Bond (1991) difference GMM estimation approach for both static and dynamic treatments. Apart from the novelty of investigating the diversification/performance relationship in a comprehensive three-stage process and controlling for selectivity and endogeneity issues with advanced methodological approaches for firm-level panel data, this work makes another contribution as a pioneer in studying diversification activities of firms in a transition country. We argue that diversification can be a growth strategy also for firms in transition countries. We rely on the dataset of the whole population of firms in Binh Duong, the most competitive province in Vietnam after excluding state-owned and foreign firms, which is extracted from the annual enterprise survey of Vietnam General Statistics House. Key findings of the present study are: (i) factors stimulating firms to diversify do not necessarily influence their diversification degree to the same extent of sign and magnitude; (ii) firms with higher debt ratio are more likely to diversify and with stronger degree; (iii) export firms have more opportunities to recognize diversification activities, but do not necessarily diversify at stronger degree than non-export firms; (iv) diversification has a curvilinear effect on firm-level profitability: product diversification improves firms’ profit up to a point, after which a further increase in diversification is associated with declining performance; (v) firms endowed with highly skilled human capital are likely to successfully exploit diversification as an engine of growth; (vi) low industry profitability significantly stimulates firms to diversify into other business sectors, but does not have any impact on their overall performance. The paper is structured as follows. Section 2 summarizes the theoretical discussion on product diversification and its relationship to the performance of firms. Section 3 gives an overview of the dataset that we use for this paper. Section 4 presents the operationalization of variables adopted together with their descriptive statistics and correlation matrix. Section 5 develops the approach(es) that we apply to obtain final empirical estimation after relevant tests for the existence of sample selection and endogeneity. Section 6 discusses the estimation results, and finally, section 7 gives some concluding remarks for policy implication and future research. 2. Literature discussion There have been a number of researchers proposing definitions and measures of product diversification. Ansoff (1965), defines a diversification strategy as the entry into new markets with new products, whereas Kamien and Schwartz (1975) define it as a firm’s degree of product and market involvement. For Jacquemin and Berry (1979), product diversity refers to the degree 4 of relatedness among various product segments. Following Rumelt (1974), several scholars 1 view diversification as the strategy of adding related or similar product/service lines to existing core business, either through acquisition of competitors or through internal development of new products/services, which implies increase in available managerial competence within the firm. In this sense, diversification is a matter of degree of relatedness among the activities carried out by a firm. Product relatedness is defined as the extent to which a firm’s different lines of business are linked by a common skill, market, purpose, or resource (Rumelt, 1974; Luo, 2002). Recent studies have attempted to examine diversification patterns from underlying resource requirements: the degree to which two industries use the same types and proportions of human expertise (Farjoun, 1994) or rely on the same inflows of technology (Robins and Wiersema, 1995). However, these studies characterize resources only at the industry level, which limits the ability to address issues relating to heterogeneity in firms’ resource bases. Thus, in practice, diversification is normally measured as the number of activities a firm undertakes in different sectors. The degree of relatedness is then measured with reference to the system of standard industrial classification (SIC codes). While this type of measure incurs inherent limitations for not taking into account internal managerial effort or resource requirements underlying observable diversification activities and relying on proximity among SIC codes (Silverman, 1999), it has still been applied commonly in empirical parts of this work due to its availability and straight forward nature. Montgomery (1994) distinguishes three motivations for diversification: the search for market power; the solution to agency problems; and the application of bundles of resources to attain a competitive advantage (resource-based view). Focusing on the determinants of the distribution of the firm’s activities over industries beside its primary focus on vertical integration, transaction cost economics suggests that diversification is an alternative contractual method by which a firm can exploit its surplus resources (Silverman, 1999). By the same token, Grossmann (2007) submits that diversification may be a mean to extend the boundaries of a firm in the presence of internal coordination problems, which naturally arise in large firms. Multiproduct firms can increase their market power by cross subsidisation activities, i.e. market strength in one particular industry may be used to sustain low price strategies in other markets. Similarly, Teece et al. (1994), Christensen and Foss (1997), Foss and Christensen (2001) agree that diversified firms can create positive spillovers since the value of resources in one industry increases due to investment in another industry. Finally, the internal capital markets hypothesis indicates that diversified firms arise when financial market imperfections force managers to allocate funds more efficiently than the external capital market (Klein and Lien, 2009). Usually, firms diversify as long as they see the opportunity to consolidate their market power, which predicts a linearly positive relationship between diversification and profitability. Diversification strategies undertaken by growth-oriented managers may both well exploit scope economies and at the same time increase firms’ market power. An efficient way to increase firms’ market power is the multimarket contact hypothesis (Bernheim and Whinston, 1990; Scott, 1993; Spagnolo, 1999), following which firms meeting in several markets have a greater incentive to network with each other in order to sustain collective power. By diversifying in a similar way (in order to exploit cost synergies), a group of firms might create and consolidate a situation of 1 Such as Markides (1995), Markides and Williamson (1994), Tallman and Li (1996), Nachum (1999). 5 multimarket contacts where collusive practices are more likely to emerge. With respect to the effects, good performance outcomes for diversified firms are consistent with both market power, i.e. firms meeting in several markets co-ordinate to increase their bargaining power on setting higher prices, and efficiency reasons, i.e. firms diversify to exploit positive cost externalities. Whereas the market power search approach is consistent with a linear positive relationship between diversification and profitability, the agency approach predicts a negative relationship as managers use free cash flow for the sake of their own goals. This leads to considering diversification in large firms as a result of the separation between ownership and control which induces managers to pursue their own objectives at the expense of shareholders. Hoskisson and Hitt (1990) suggest that diversification, firm size, and executive compensations are highly correlated to the extent that diversification provides benefits to managers that are unavailable to investors. Diversification can also lead to the problem of moral hazard due to conflict of interest between managers having interest in costly diversification as a form of compensation and investors preferring to concentrate on the core business to maximize their returns (Bhide, 1990). Third, the same negative relationship between diversification and firm performance is also predicted by the resource-based view (Penrose, 1959). Firm is seen as a collection of sticky and imperfectly imitable resources or capabilities which strengthens its competition against its competitors (Barney, 1986). The deployment of surplus resources and free cash flows is one of the prime motives of diversification (Hoskisson and Hitt, 1990). However, asset specificity embedded in firms’ resources on one hand brings sustainable competitive power for their owner relative to competitors, but on the other hand acts as a challenge impeding firm’s ability to transfer resources to new applications or “transplant” them in a new context (Montgomery and Wernerfelt, 1988). Therefore, the value of diversification will depend on the complementarities existing between internal resources and the business/industry that the firm enters, as well as the diversifying mode that it chooses. This opens the way to several empirical predictions revolving around the concept of relatedness of diversification activities: the more closely those activities are related or complementary, the more profitable diversification is expected to be. Regrettably, the resource-based approach to diversification has generally not considered the possibility that firms can exploit resources through market arrangements rather than through expansion of corporate boundaries. Although resources can be exploited through contracts, the regularlyassumed valuable, rare, inimitable, and difficult-to-imitate resources are too “asset specific” (Williamson, 1985) to be contracted in market transactions. Finally, researchers of transaction cost economics suggest that diversification is an alternative contractual method by which a firm can exploit its resources (Silverman, 1999). However, no matter how business activities are related, the transfer of product and process technology among different industries with different characteristics normally requires certain modification and adjustment, which incurs varying degree of transaction costs as a result (Qian, 2002). When a firm moves into a market with only a weak connection to its primary line of business (unrelated diversification), it often lacks the know-how and managerial resources to prevail against the competition in this new industry. Diversification beyond a certain degree raises internal governance and administration costs to the point that performance suffers (Jones and Hill, 1988). Thus, many of the most significant failures of diversification can be traced to the failure of achieving sufficient relatedness between business sectors (Grant, 1988). The related 6 hypothesis in transaction cost approach claims that multi-business firms holding portfolios of similar related business might obtain efficiency advantages unavailable to non-diversified firms or firms with unrelated portfolios. According to Lien and Klein (2006), economic sense of relatedness implies that resources in one industry are substitutes for, or complements to, resources in another industry. Whether the firm successfully integrates new business sectors depends on the comparative costs and benefits of contracting, not on the underlying production technology. Iacobucci and Rosa (2005) suggest that group formation offers a solution to problems created by market imperfections that increase transaction costs. Empirically, the impact of diversification on firm performance is mixed (Datta et al., 1991; Hoskisson and Hitt, 1990). Some studies claim diversifying into related product markets produces higher returns than into unrelated markets, others propose that less diversified firms perform better than highly diversified firms (Christensen and Montgomery, 1981; Rumelt, 1974, 1982). Some claim that the economies in integrating operations and core skills obtained in related diversification outweigh the costs of internal capital markets and smaller variances in sales generated by unrelated diversification (Datta et al., 1991). While Prahalad and Bettis (1986) claim that it is not product-market diversity but the strategic logic applied by managers that determines the effect of diversification on performance, Montgomery (1985) argues that it is not management conduct, but industry structure that governs firm performance. Vannoni (2004) observes that diversification is normally approached focusing either on the synergies exploited by diversified firms and on the optimal organizational structure for managing a multiproduct firm (strategic management approach) or on the relationships between market/industry structure (industrial organization approach). In each discipline the empirical literature has grown with scarce contacts with the other one. Thus, the need to construct and/or reinforce an interdisciplinary dialogue is particularly evident in this area of studies. With respect to the strategic management approach, corporate strategies despite being based on various sets of management guidelines addressing the question “what is the appropriate scale and scope of the firm?”, all converge in dealing with conflicting demands of synergies and responsiveness with respect to allocating resources (Wit and Meyer, 2005). Successful corporate strategies are the result of organizational capabilities or competencies that allow firms to exploit potential synergies that large size or diversity can offer. On one hand, the synergy of interrelated businesses within a diversified firm brings in the benefit of economies of scope which arise from sharing both common tangible inputs such as markets, distribution systems, product and process technologies, or manufacturing facilities (Ansoff, 1965; Rumelt, 1974; Teece, 1980), and intangible assets such as brand names and know-how (Qian, 1997), managerial capabilities and routines and repertoires (Prahalad and Bettis, 1986; Grant, 1988). The more interrelated the businesses of a firm, the greater the potential for organizational synergy (Rumelt, 1974; Salter and Weinhold, 1981). On the other hand, synergy has harmful effects owing to responsiveness, such as higher governance costs, slower decision-making, strategy incongruence, dysfunctional control, and dulled incentives (Wit and Meyer, 2005). Thus, the fundamental challenge facing corporate diversification is the conflicting forces stemming from synergy and responsiveness, or as described by Dess et al. (2003), “managing the conflict between the new and old (business activities) and overcoming the inevitable tensions that such conflict produces for management”. Nevertheless, empirical studies of synergy and responsiveness only enable us to state whether 7 diversification has a positive (due to synergy) or a negative (due to responsiveness) effect on firm performance, or which type of diversification, related or unrelated, is more beneficial. With respect to the curvilinear relationship between diversification and firm performance, we cannot explain to what extent the positive effect from synergy fades away and will be replaced by the negative effect of responsiveness, or why moderate levels of diversification yield higher levels of performance than either limited or extensive diversification (Tran and Zaninotto, 2012). With respect to the industrial organization approach, diversification as the proxy for economies of scope is investigated in relation to firms’ innovative capabilities. Firms are assumed to have different innovative capabilities that lead them to pursue different types of product diversifications (Cohen and Klepper, 1992). A firm with a diversified portfolio of products may be better positioned to determine the general applicability of new ideas than a firm with a narrower portfolio of products, because it can capture internal knowledge spillovers. Indeed, firms selling only one category of products are less likely to engage in R&D than those selling a broader range of products (Piga and Vivarelli, 2004). Given the same competencies for the production and delivery of core products, together with the same incentives to diversify, firms possessing more dynamic capabilities will be more likely to expand their product scope (Doving and Gooderham, 2008). Regardless of which disciplinary and theoretical perspective one adopts, most studies support a non-linear relationship between diversification and profitability (for a review, see Palich et al., 2000). Appropriateness of product diversity is judged by a balance between economies of scope and diseconomies of scale, which indicates a limit on how much a firm can diversify. If a firm goes beyond this point, its market value suffers (Markides, 1992). Related diversification enables corporations to perform better if they expand their stocks of strategic assets efficiently and in consistency (Teece et al., 1994). Among the studies including measures of relatedness in order to discriminate between type and level of diversification, Vannoni (2000) presents evidence for a sample of Italian firms that irrespective of the number of industries in which they are active, they reach good performance results thanks to the relatedness of their diversifications strategies. Based on the above survey of the relevant literature, we propose the following proposition on diversification activities of firms: Proposition: Product diversification has a curvilinear relationship with profitability: positive when firms expand into industries related to their original industries and negative when they diversify beyond their original industries. 3. Data description Panel firm-level data from 2001 to 2006 were extracted from the GSO (General Statistics Office) of Vietnam’s database of annual national enterprise surveys. Since we wished to omit the effects of the unique macro-economic and institutional environment of each province, only firms located in the province of Binh Duong were used for the empirical analysis. In other words, we use the whole population of firms in an administrative unit, rather than samples of representative firms from different ones, for the analysis 2 .State-owned monopolistic firms (electricity, water 2 The rationale to select for analysis only observations in one province is: (i) to isolate the effect of institutional and macroeconomic features which are assumed to be homogenous to every firm within one province; (ii) to control for the influence of 14 process and create a predicted probability (i.e. propensity score) of diversification vs. nondiversification from a logistic regression equation and kernel-based matching. These scores then can be used to match diversification and non-diversification as a covariate in the main equation. The results from four estimation approaches: the standard Heckman’s two-stage method, the Vella (1998) and Wooldridge (1995) parametric approach, the Semykina and Wooldridge (2010) model correcting for both endogeneity and selectivity, and the Heckman et al. (1998) kernelbased propensity score matching method, are reported in table 5. Both parametric and semiparametric identification of sample selection model generally requires and “exclusion restriction” – that is, a regressor that is included in the set of regressors  for the binary variable  (the selection equation) must be excluded from the list of ? in the equation of interest (degree of diversification). This paper will adopt a dummy which indicates whether firmsare involved in export / import activities as the exclusion restriction. We believe that when conducting business transactions with foreign partners, firms will be more likely to decide to diversify their production in order to capture emerging demands from global market. However, the degree of their diversification activities will not be influenced by whether they export/import or not, but mainly by their production capacity and potentiality. The exclusion restriction test is presented in Table 1 Before estimating the selection model, we performed two tests for the presence of selection bias in the diversification degree equation. So far various tests have been proposed by Verbeek and Nijman (1992), Vella (1998), Wooldridge (1995), Das et al. (2003), Hsiao et al. (2008), and Semykina and Wooldridge (2010). While Verbeek and Nijman (1992), and Hsiao et al. (2008) propose the test for panel data models with exogenous explanatory variables, Semykina and Wooldridge (2010) use fixed effects to remove unobserved heterogeneity, and hence, permit arbitrary correlation between idiosyncratic errors and regressors. The test is based on the within transformation and has an importance advantage over alternative testing procedures because it is valid in the presence of arbitrary correlation between unobserved heterogeneity and explanatory variables. Given its novelty, we apply this test in three steps: (i) for each year, use probit to estimate <( =1\)==( +5@). Use the resulting estimates to obtain the inverse Mills ratio : > ≡:(A +5@ >); (ii) For  =1, use fixed-effect two-stage least squares (FE-2SLS) to estimate the equation:  =  +B: > + with ,: > as instruments. The interactions of : >with year dummies are included as well to allow B to be different across ; (iii) We use t-statistic to test CD:B =0 for the significance of inverse Mills ratio, and the Wald test to test CD=B=⋯=BG=0 for the significance of the interactions. The second selection bias test we apply is the non-parametric test proposed by Das et al. (2003), and applied by Hall et al. (2009) in their innovation intensity equation. First, a random effect probit model is adopted for equation (1) with ‘export’ dummy as the exclusion restriction. Second, predicted probability of diversification from the first regression and the corresponding Mills’ ratio are added in equation (2) as independent variables. Fixed effect least squares regression is adopted for the degree of diversification equation since we have a significant number of firms not diversifying (entropy=0). Selection bias is determined based on the significance of the predicted probabilities, the Mills’ ratio, their squares and interaction. Results from the test for sample selection bias are present in table 1. 15 Table 1: Sample selection test and exclusion restriction test for the variable “export” Relevance Das et al. (2003) test Semykina and Wooldridge (2010) test H ( 5 ) = 25.35 p-value = 0.0001 I J K (1,5555) = 288 p-value = 0.000 I J K ∗  . LMNO ( 5 , 5555 ) = 24.89 p-value = 0.000 Quality test (correlated with regressors) H ( 1 ) = 5.11 p-value = 0.0238 Validity test (exogeneitycondition) H ( 1 ) = 1.72 p-value = 0.19 The two sample selection tests indeed indicate the existence of selectivity issue in our panel data, which calls for relevant treatment with our estimation models reviewed above. The exclusion restriction test shows that “export” is a good exclusion since the firms’ likelihood to export is significantly related to their propensity of diversification (quality test), but does not influence how much they will diversify once they are involved in diversification (validity test). 5.2 Performance equation We consider both static and dynamic estimation models for the third stage of firms’ diversification process, i.e. outcome. In the dynamic model, lagged dependent variable is included to isolate the effect of potential performance shock as well as to account for necessary time lag until the pay-off of diversification can be recognized. For each model, we provide two specifications with respect to whether diversification degree is treated as exogenous or endogenous. Although endogeneity bias is commonly confronted in cross-sectional studies, it is less frequently considered to be a source of concern in panel data estimation. This is partly due to the concept that fixed effects estimation eliminates most forms of unobserved heterogeneity (Verbeek and Nijman, 1992). However, Vella (1998) claims that certain forms of selection bias and heterogeneity cannot be eliminated with panel FE and RE models. The Durbin-Wu-Hausman test for endogeneity later justifies the need to isolate the endogeneity bias of diversification strategies. The firm performance equation can be written as follows: PQ =PQR@+@ +?@S+T+7 (8) ( =1,2,…,; = 1,2,…,W) In equation (8) above, PQRis the 1-year lagged value of return on investment of firm i in year t. For the estimation of the static model, PQR is not included. is the diversification index of firm i. ? is a matrix of control individual-level, firm-level, and industrylevel characteristics. T an unobserved firm-specific time-invariant effect which allows for heterogeneity in the means of the PQ series across firms, and 7 is a disturbance term. A key assumption maintained throughout this work is that the disturbances 7 are independent across individuals. We also treat the firm-effects T as stochastic, which implies here that they are necessarily correlated with lagged dependent variable PQR. 16 Test for violations of estimation assumptions: (77 X  X )=YZ  =[, =[(C1) 0ℎ]^](C2) – Heteroskedasticity (H1) : the problem of heteroskedasticity is more prevalent in cross-sectional data, because they involve units and groups which are heterogeneous in nature. We apply the likelihood ratio test for heteroskedasticity in panel data and find out the strong existence of heteroskedasticity in our data. Thus, estimation with OLS is rejected, and the alternative estimation technique capable of correcting for heteroskedastic errors is “robust” regression method with standard errors corrected for heteroskedasticity by White’s method. Table 2 Test for Heteroskedasticity Heteroskedasticity test ROI Likelihood ratio test H ( 929 ) = 5522.95 p-value = 0.000 – Serial correlation in time-series data (H2): the Wooldridge test for first-order autocorrelation in panel data is insignificant even at the 5% level, which indicates the absence of first-order serial correlation in the ROI equation. Serially correlated errors will give biased estimators with increasing variances of estimated coefficients. In this case, we can feel sure that ROI as the dependent variable satisfies the assumption of no serial correlation. Table 3 Test for Serial Correlation Serial correlation test ROI Wooldridge first-order serial correlation test F (1,929) = 3.195 P-value = 0.0728 - Endogeneity of diversification index: We consider the potential endogeneity of diversificationdegree; since we suspect that some unobserved individual-level and firm-level characteristics might influence both the firm’s diversification decision and its subsequent profitability. The Durbin-Wu-Hausman test (see below) does indicate the strong presence of the endogeneity of diversification. The test begins with the reduced form regression, in which the assumed-endogenous diversification index is the dependent variable and all other observed exogenous firm-level characteristics are independent variables. The residuals predicted from this regression are then added into the structural equation (1). The endogeneity problem is determined according to the significance of the residual coefficient. Table 4 Test for Endogeneity Durbin-Wu-Hausman test H ( 1 ) = 25.04 P-value = 0.0000 Estimation methods: Several econometric problems may arise from estimating equation (8) 1. Diversification index  is assumed to be endogenous. 2. Time-invariant unobserved firm characteristics (fixed effects) T may be correlated with and ?. 17 3. The panel dataset has a short time dimension (W =6) and a large number of firms ( = 930). Thus, the presence of the lagged dependent variable PQR may give rise to autocorrelation, since it is correlated with fixed effects. It is therefore also treated as endogenous variable. OLS estimators of @ are inconsistent, since explanatory variable PQRis positively correlated with error term (T+7) due to the presence of firm-effects, and this correlation does not vanish as the number of firms in the sample increases. Within groups estimator eliminates this source of inconsistency by transforming the equation to eliminate T. However, for panels in which the number of time periods available is small, this transformation includes a non-negligible correlation between the transformed lagged dependent variable and the transformed error term: PQR − GR(PQ +⋯+PQGR) and - − GR(- +⋯+-G). Thus within groups estimator is also inconsistent here. To solve problems 1 and 2, one would usually use fixed-effects instrumental variable estimation (2SLS), but this depends on the availability and validity of exogenous instruments. We therefore decided to use the Arellano-Bond (1991) difference GMM estimator, first proposed by Holtz-Eakin et al. (1988). The lagged levels of endogenous regressor are used as instruments, which rise over time. This makes the endogenous variable pre-determined and thus not correlated with the error term in equation (8). To cope with problem 2 (fixed effects), the differenced GMM uses first-differences to transform equation (8) into: ∆PQ =@∆PQR +@ ∆ +@S∆? +∆7(9) By transforming the regressors by first differencing, the fixed firm-specific effect is removed, because it does not vary with time. Lastly, to cope with problem 3, the Arellano-Bond estimator was designed for small-T and large-N panels. For the endogenous lagged dependent variable, the first-differenced lagged dependent variable is instrumented with its past levels. Table 7 shows the estimation results from five estimation approaches: fixed-effects (FE) and random-effects (RE) model; instrumental-variable two-stage least-squares estimation (IV-2SLS) with GMM treatment for the static model; and the Arellano-Bond (1991) difference GMM estimator for the dynamic model. According to Baum and Schaffer (2003), generalized method of moments (GMM) estimation is more efficient than 2SLS when heteroskedasticity is present. Following Arellano-Bond (1991), the instrument for the endogenous diversification index is its one-period lagged value. This makes the endogenous variable pre-determined and, hence, not correlated with the error term in equation (8). 6. Estimation Results and Discussion The estimation results for diversification decision and diversification degree of firms are given in Table 5 and 6. These tables show a striking consistency in the general pattern of results obtained. Table 5 presents 4 estimation models: (i) first step probit of Heckman two-step consistent estimates; (ii) joint maximum likelihood estimation (MLE) based on the joint normality of (,]); (iii) first step of Semykina and Wooldridge (2010) method; and (iv) kernel-based propensity matching method. 18 Table 5: Probability of diversification decision Probability of diversification Variable Heckman twostep (1 st step probit) Joint MLE (1 st step) Semykina and Wooldridge (2010) (stage 1 of 2SLS) Heckman (1995) Propensity matching method Firm age 0.0205** (0.004) 0.023** (0.004) 0.02** (0.004) 0.035** (0.006) Firm’s economic size 0.0385 (0.0844) 0.052 (0.076) 0.333** (0.084) 0.098 (0.146) Firm’s economic size square -0.0032 (0.0044) -0.004 (0.004) -0.017** (0.004) -0.007 (0.007) Firm’s labour size 0.111** (0.021) 0.087** (0.021) 0.095** (0.025) 0.191** (0.035) Debt ratio 0.42** (0.071) 0.379** (0.071) 0.349** (0.088) 0.787** (0.126) Capital intensity -0.0008 (0.0008) -0.0001 (0.00006) -0.0001* (0.00005) -0.0001 (0.0001) Technological resources -0.281* (0.134) -0.172 (0.146) -0.465* (0.217) -0.489* (0.235) Export Y/N 0.125** (0.047) 0.041* (0.021) 0.128* (0.062) 0.221** (0.081) Professional education Y/N 0.351** (0.0424) 0.377** (0.040) 0.403** (0.048) 0.606** (0.073) Gender 0.155** (0.047) 0.147** (0.045) 0.082 (0.055) 0.263** (0.082) Age -0.0102** (0.002) -0.007** (0.002) -0.007** (0.002) -0.0166** (0.003) Average industry ROA -3.898** (0.817) -3.879** (0.873) -2.878** (0.932) -7.078** (1.471) : > 0.666** (0.148) : > * year dummies H ( 5 ) =399.32**  ) a a a H ( 9 ) =380.13** Intercept -0.15 (0.407) -0.316 (0.364) 0.07 (0.403) -0.111 (0.705) Likelihood ratio test H ( 12 ) 240.17** 242.33** Observations 5580 5580 5580 5580 Note: (*) significant at 5% level; (**) significant at 1% level; Standard errors in brackets Both the Shapiro-Wilk W and the Shapiro-Francia W test for normality assumption of the error terms in diversification decision equation indicate the rejection of the null normality hypothesis at 1% significance level. Thus, the Heckman two-step procedure requiring normality assumption may not be an efficient estimation method for our analysis. If we compared estimation results across different methodological treatments in Table 5, sign of estimated parameters are quite consistent although the statistical significance seems to be stronger with coefficients obtained from the first stage of Semykina and Wooldridge (2010) approach. Note that the estimated equations here refer to the decision to diversify or not, rather than to the degree of diversification, so the estimated coefficients will carry a different meaning, i.e. the likelihood of undertaking a diversification policy. There is consistency in significant effect of firm age, firms’ labour size, debt ratio, technological resources, export, professional education, gender and age of firm owners and 19 average industry ROA. Significant and positive parameters of debt ratio indicate the leverage effect of loans on stimulating firms’ diversification activities. Technological resources proxied by the rate of technical employees in the whole labour force act as a barrier to firms’ entering new business sector. It is likely that technical employees with specialized knowledge in the core business are less willing to absorb new knowledge and skills required for crossing firms’ business boundaries, which is actually a significant source of change resistance in incumbent firms. We find evidences to support the quality test of ‘export’ as the exclusion restriction in sample selection treatment. Those firms exporting their products / services into foreign markets are more likely to undertake diversification in order to capture emerging demands, advanced technology and resources from international markets. Table 6: Diversification degree of firms Degree of diversification (Entropy index) Variable Heckman two-step Joint MLE Semykina and Wooldridge (2010) 2SLS Vella and Wooldridge (1994) Firm age 0.0149* (0.0065) 0.013** (0.003) 0.002 (0.002) 0.028** (0.002) Firm’s economic size 0.181** (0.066) 0.171** (0.076) 0.151** (0.0405) 0.087** (0.032) Firm’s economic size square -0.0095** (0.0035) -0.009** (0.003) -0.007** (0.002) -0.0093** (0.0017) Firm’s labour size -0.034 (0.033) -0.044** (0.013) -0.025** (0.011) -0.144** (0.011) Debt ratio 0.193 (0.13) 0.193** (0.045) 0.08** (0.027) 0.702** (0.041) Capital intensity -0.001 (0.0008) -0.0006 (0.0004) -0.0002 (0.0004) -0.0001** (0.00008) Technological resources 0.074 (0.127) 0.007 (0.097) 0.251** (0.073) 0.82** (0.063) Professional education Y/N 0.242* (0.105) 0.221** (0.028) 0.0363* (0.02) 0.102** (0.0206) Gender 0.111* (0.058) 0.118** (0.029) 0.015 (0.02) 0.263** (0.022) Age -0.0013 (0.0032) -0.037** (0.0015) -0.004** (0.001) -0.008** (0.001) Average industry ROA -2.218* (1.219) -2.018** (0.59) -1.987* (0.928) -2.759** (0.362) : > 0.711* (0.365) -0.136** (0.033) : > * year dummies H ( 5 ) =10.7*  ) * 1.497** (0.059) Intercept 0.281 (0.456) 0.504* (0.25) 0.07 (0.403) -1.781** (0.184) Wald LR H ( 11 ) = 31.05** H ( 11 ) = 130. 28** H ( 17 ) = 98.28** H ( 13 ) = 5555** Observations 5580 5580 5580 5580 (1186 uncensored) Note: (*) significant at 5% level; (**) significant at 1% level; Standard errors in brackets 20 With respect to the effects of individual-level characteristics on the likelihood of firm diversification, significant impact could be found for all three control variables: other factors remain constant, firms having higher educated owners, or female owners, or younger owners are more likely to diversify than their counterparts. Finally, we witness the negative relationship between industry profitability, proxied by average industry ROA, and firms’ diversification decision. Apparently, firms facing low profitable opportunities in their own industry will have to search for opportunities in other industries: one of the motives for firms’ diversification activities is to compensate for their loss or poor performance in their core business. We again observe the consistency in the general pattern of results obtained. It is worth noting that the effects of some variables, for instance technological resources and firm size, in this equation contradict with their effects in the above diversification decision equation, which indicates that factors stimulating firms to diversify do not necessarily influence their diversification degree to the same extent and vice versa. Firm age is statistically positively related to degree of diversification of firms. Older firms with accumulating knowledge and experience tend to take risks of expanding their product portfolio generously rather than stay persistently within their core business. Economic size of firms, as expected, positively stimulates their diversification activities. Firms possessing larger asset pool have more favourable conditions to invest in R&D, and hence, are more incentivised to bring up radical innovations. However, the effect of economic size on the firm’s diversification intensity is nonlinearly decreasing to scale. Further increase of total asset to a certain level will not create the proportional positive effect on the likelihood that firms diversify into unrelated knowledge base. However, firm labour force size has a statistically negative impact on firms’ relatedness of diversification although it has a stimulating effect on firms’ likelihood to diversify: larger firm size on one hand facilitates firms’ diversification activities by transferring a part of their rich human capital into other business sectors, but on the other hand impedes the diversification into unrelated business activities which require higher degree of flexibility and responsiveness of business units. This is because there has been no evidence that large firms considered in the analysis undertake their diversification policy by creating independent business units. Finally, debt ratio and technological resources both significantly stimulate firms to increase their involvement in areas unrelated or just marginally related to their current domain of competence and corresponding opportunity set. Technical employees do not encourage firms to enter new business sectors; but once firms overcome this change resistance by taking part in some forms of diversification, they act as a stimulative source for firms to take risks in joining unrelated industries. Consistent with above findings, owners who are more educated, or female, or younger are more likely to go into unrelated diversification. We also find the negative relationship between industry profitability and firms’ degree of diversification. Firms may want to get out of their stagnant industry by diversifying into a completely new industry. Table 7 lists the estimation results for both static and dynamic models. The former adopts panel-data estimation: fixed-effects and random-effects regressions when diversification is assumed exogenously; and the GMM technique when it is assumed endogenously. The latter adopts differenced GMM estimation and measures diversification exogenously and endogenously respectively. Given the panel data structure and diagnosis tests performed above, the dynamic model with GMM treatment is the preferable estimation, based on which results are interpreted. 21 Table 7: Firm performance (ROI as the dependent variable) Firm return on investment (ROI) Static model PQ  =   @  + ?  @ + T  + 7  Dynamic model PQ  = PQ  R  @  +   @ + ?  @ S + T  + 7  FE 1 RE 1 GMM 2 GMM exogenous 3 GMM endogenous 4 ROI (t-1) 0.375** (0.068) 0.364** (0.062) Entropy 0.595** (0.121) 0.267** (0.052) 0.255** (0.0411) 0.798** (0.188) 0.78** (0.18) Entropysquared -0.257** (0.094) -0.073 (0.056) -0.085* (0.043) -0.38** (0.133) -0.369** (0.129) Technologicalresources 0.046* (0.025) 0.006 (0.02) 0.007 (0.021) 0.042 (0.056) 0.0448* (0.056) Leverage (debt ratio) -0.056** (0.0214) -0.068** (0.015) -0.083** (0.011) -0.118** (0.036) -0.116** (0.036) Capital intensity -0.0002 (0.0002) -0.0004 (0.00001) -0.0001** (0.00005) -0.0001 (0.0001) -0.0001 (0.0001) Export 0.025** (0.0064) 0.032** (0.0057) 0.0349** (0.0052) 0.034** (0.011) 0.035** (0.011) Firmage 0.0008 (0.0008) 0.0005 (0.0006) 0.0002 (0.0004) 0.0006 (0.001) 0.0007 (0.001) Economicsize -0.098* (0.051) -0.087** (0.033) -0.084** (0.0185) -0.145* (0.06) -0.144* (0.061) Economicsizesquared 0.0053* (0.0025) 0.004** (0.0017) 0.0043** (0.0009) 0.008* (0.003) 0.007* (0.003) Laborsize 0.022** (0.0062) 0.0049 (0.0038) 0.005* (0.0027) 0.0088 (0.011) 0.009 (0.011) Professioneducation Y/N 0.170** (0.011) 0.125** (0.008) 0.083** (0.005) 0.14** (0.016) 0.14** (0.016) Gender 0.025* (0.01) 0.024** (0.0065) Age -0.003* (0.0017) -0.0003 (0.0003) -0.0001 (0.0002) -0.002 (0.002) -0.003 (0.02) Averageindustry ROS 0.018 (0.089) 0.066 (0.078) 0.128* (0.061) 0.108 (0.088) 0.101 (0.089) Intercept 0.462* (0.241) 0.399* (0.157) 0.426** (0.087) 0.685* (0.324) 0.698* (0.328) F-value F(13) =29.04** F(14) =45.75** Wald Chi-sq H ( 13 ) = 408.23** H ( 14 ) = 295.15** H ( 14 ) = 309.74** Hausman test H ( 12 ) = 229** Hansen J statistic of excluded instruments H ( 1 ) = 3.2 P = 0.072 Observations 5580 5580 4650 3720 3720 Notes: (**): significant at 1% level; (*): significant at 5% level; Standard errors in brackets 1: FE and RE estimators assume that the diversification index is exogenous. 2: GMM estimators assume that the diversification index is endogenous. 3: The differenced GMM estimator assumes that all explanatory variables, apart from the lagged dependent variable, are exogenous; robust standard errors are used to control for heteroskedasticity and serial correlation. 4: The differenced GMM estimator assumes that the diversification index and lagged dependent variable are endogenous; robust standard errors are used. 22 From Table 7, we can see that the coefficients of lagged ROI are statistically significant in both regressions, which indicates the superiority of the dynamic model with endogeneity treatment of entropy index. It is plausible that ROI has significant lag effect since firms base their investment decision this year contingent on the investment return of last year. The ROS equation is placed in Appendix C. In general, both ROS and ROI equations show substantial consistency in the pattern of regressor significance and size of coefficients. For the analysis of estimation results, we base our interpretation on the ROI equation. Due to the significant of lagged ROI, the dynamic model with GMM treatment is the better estimation technique. The Hausman test also indicates the superiority of the dynamic model with ROI as the dependent variable at 1% significance level 4 . We first discuss the estimation results from FE and RE regressions in which diversification is measured exogenously. The criteria for selecting the RE or FE model is often based on whether (T,b)=0. The Hausman test with the null hypothesis that(T,b)=0 can be rejected, whichindicates that the RE model is preferable. With ROI as the measure of profitability, both FE/RE and GMM estimations tend to find a positive relationship between entropy index and firm performance. Generally, more highly focused firms tend to have lower profitability or, equivalently, greater diversification raises profitability. In other words, positive effects occur as firms move from a single-business strategy to a diversification strategy. However, the significant parameters of the square of the entropy index indicate the non-linear influence of diversification: the positive effects of diversification gradually fall as the firm moves further and further away from its core business. These findings support our hypothesis and are consistent with most of those summarized by Palich et al. (2000). Based on previous evidences, they conclude that performance increases as firms move from single-business strategies to low-scaled diversification, but that the effect deteriorates as firms move away from the low end of related diversification to the high end of unrelated diversification. As Qian (1997) suggests, the relative costs and benefits of product diversification are likely to depend on how different business activities of a firm are related to each other. If they are loosely linked or poorly structured, they are less likely to complement or supplement each other, and hence, synergy will not exist. Obviously, the profitability of a firm can mainly be accelerated by increasing innovativeness through its accumulated technological resources. According to resource-based theory, a manufacturing firm’s technology resources are valuable assets for its survival and development and can differentiate its performance. Thus, it is not surprising that the number of R&D and technical personnel as a proxy of technological resources of a firm has a strong positive effect on its profitability. The relation between corporate performance and the debt-to-assets ratio has long been established in corporate finance literature (McConnel and Servaes, 1990; Lang et al., 1991; Harford, 1999). According to this paradigm, capital structure choice is a trade-off between the 4 H =50.58 p-value = 0.000 23 costs and benefits of debt. Although there is agreement among academics and practitioners on the benefits of debt, it may be argued that large firms are more likely to receive more benefits than medium-sized or small firms, at the same level of debt ratio. While loans significantly stimulate firms to be involved in diversification activities, their burden imposes a serious impediment to firm value, especially for small and medium-sized firms. Therefore, it is plausible that debt ratio is estimated to have statistical negative effect on firm profitability. We find a significantly convex relation between the economic size of firms in terms of total assets and their investment return. Larger firms realize lower return on investment than their smaller counterparts. Owners tend to face more challenges in allocating resources efficiently in large firms. It is worth noting that the majority of total assets of firms in Vietnam are fixed assets including land, machinery, equipments, etc. Their “asset specificity” made them difficult to be transferable to other business sectors (high transaction cost); and thus imposes a limitation in diversification pay-off, and in turn outcome that large firms can obtain. This tends to limit the firm's economic size to the extent that owners-managers achieve optimum efficiency. We do find the significant effect of quadratic coefficient of economic size to indicate a curvilinear influence on performance. Those firms exporting their products / services are likely to outperform their non-exporting counterparts. Finally, with respect to individual characteristics of owners, consistent with our previous finding (Santarelli and Tran, 2013) and a number of studies on the return to education (for instance, Cooper et al., 1994; Parker and Van Praag, 2006), professional education has a significant and positive effect on firm performance. We do not find evidences for a significant impact of industry profitability on firm performance. 7. Conclusions Research tends to focus on performance outcomes of types and degrees of diversification rather than on what determines diversification in the first place (Hoskisson and Hitt, 1990; Doving and Gooderham, 2008). Although researchers share a common consensus that diversification is to deploy surplus resources and cash flows, they still fail to account for the antecedents of resource deployment, and in turn of diversification decision. Various approaches, resting on differing assumptions, justify divergent relationships. However, these assumptions all converge in dealing with conflicting demands of synergies and responsiveness with respect to diversification. Their investigation will enable us to understand whether diversification has a positive or negative effect on firm performance. Empirical results seem to be consistent with a resource- (or competence)-based view, which maintains that a positive relationship between diversification and profitability depends on the relatedness of diversified activities. However, as the driving forces of diversification and its profitable pay-off are resources or prior competences underlying diversification decision, it is still not clear what factors determine firms’ decision to diversify and to what degree (relatedness of their activities). This paper is pioneer in investigating firm diversification in a transition country in three interrelated and consecutive stages: decision, degree, and outcome. Controlling for individuallevel, firm-level, and industry-level characteristics we find out that (i) factors stimulating firms to undertake diversification decision do not necessarily influence their diversification degree to the same extent of sign and magnitude; (ii) firms with higher debt ratio are more likely to diversify and diversify with stronger degree; (iii) export firms have more opportunities to recognize diversification activities, but do not necessarily diversify at stronger degree than non- 30 Maksimovic, V. and Phillips, G. (2002), “Do Conglomerate Firms Allocate Resources Inefficiently Across Industries? Theory and Evidence,” Journal of Finance, 57 (2), 721-767. Markham, J. W. (1973), Conglomerate Enterprise and Economic Performance, Cambridge Mass.: Harvard University Press. Markides, C. C. (1992), “Consequence of Corporate Refocusing: Ex ante Evidence,” Academy of Management Journal, 35(2), 398–412. Markides, C. C. (1995), “Diversification, Restructuring, and Economic Performance,” Strategic Management Journal, 16(2), 101-118. Markides, C. C. and Williamson, P. J. (1994), “Related Diversification, Core Competences and Corporate Performance,” Strategic Management Journal, 15(S2), 149-165. McConnell, J. J. and Servaes, H. (1990), “Additional Evidence on Equity Ownership and Corporate Value,” Journal of Financial Economics, 27(2), 595-612. Montgomery, C. A. (1985), “Product Market Diversification and Market Power,” Academy of Management Journal, 28(4), 789-798. Montgomery, C. A. (1994), “Corporate Diversification,” The Journal of Economic Perspectives, 8(3), 163-178. Montgomery, C. A. and Wernerfelt, B. (1988), “Diversification, Ricardian rents, and Tobin’s q,” Rand Journal of Economics, 19(4), 623-632. Murphy, G. B., Trailer, J. W. and Hill, R. C. (1996), “Measuring Performance in Entrepreneurship,” Journal of Business Research, 36(1), 15-23. Nachum, L. (1999), “Diversification Strategies of Developing Country Firms,” Journal of International Management, 5(2), 115-140. Opler, T. C. and Titman, S. (1994), “Financial Distress and Corporate Performance,” Journal of Finance, 49(3), 1015-1040. Palich, L. E., Cardinal, L. B. and Miller, C. C. (2000), “Curvilinearity in the Diversification Performance Linkage: An Examination of Over Three Decades of Research”, Strategic Management Journal, 21 (2), 155-174. Parker, S. C. and Van Praag, C. M. (2006), “Schooling, Capital Constraints, and Entrepreneurial Performance: The Endogenous Triangle,” Journal of Business and Economic Statistics, 24, 416–431. PCI. (2005) (2006) (2007) (2008), “The Vietnam Provincial Competitive Index,” Vietnam Competitive Initiative (VNCI) Policy Paper No. 10, 11, 12 & 13. http://www.vnci.org/publications.html. Penrose, E. (1959), The theory of the growth of the firm, Oxford: Blackwell. Piga, C. A. and Vivarelli, M. (2004), “Internal and External R&D: A Sample Selection Approach,” Oxford Bulletin of Economics and Statistics, 66(4), 457-482. Porter, M. (1976), “Please Note Location of Nearest Exit: Exit Barriers and Strategic and Organizational Planning,” California Management Review, 19(2), 21-33. Prahalad, C. K and Bettis, R. A. (1986), “The Dominant Logic: A New Linkage between Diversity and Performance,” Strategic Management Journal, 7(6), 485-501. Qian, G. (1997), “Assessing Product-market Diversification of US Firms,” Management International Review, 37(2), 127-149. Qian, G. (2002), “Multinationality, Product Diversification, and Profitability of Emerging US Smalland Mediumsized Enterprises,” Journal of Business Venturing, 17 (6), 611-633. Robins, J. and Wiersema, M. F. (1995), “A Resource-based Approach to the Multibusiness Firm: Empirical Analysis of Portfolio Interrelationships and Corporate Financial Performance,” Strategic Management Journal, 16(4), 277–299. Robinson, P. (1988), “Root-N-Consistent Semiparametric Regression”, Econometrica, 56(4), 931-954. Rumelt, R. P. (1974), Strategy, Structure, and Economic Performance, Cambridge MA: Harvard University Press. Rumelt, R. P. (1982), “Diversification Strategy and Profitability,” Strategic Management Journal, 3(4), 359–369. Salter, M. S. and Weinhold, W. A. (1981), “Choosing Compatible Acquisitions,” Harvard Business Review, 59, 117-127. Santarelli, E. and Tran, H. T. (2013), “The Interplay of Human and Social Capital in Shaping Entrepreneurial Performance: the Case of Vietnam,” Small Business Economics, 40(2), 435-458. Scott, J. (1993), Purposive Diversification and Economic Performance, Cambridge University Press. Semykina, A. and Wooldridge, J. (2010), “Estimating Panel Data Models in the Presence of Endogeneity and Selection,” Journal of Econometrics, 157 (2), 375-380. Shadish, W. R., Cook, T. D. and Campbell, D. T. (2002), Experimental and Quasiexperimental Designs for Generalized Causal Inference. Boston, MA: Houghton Mifflin. Sharp, B. M., Bergh, D. D. and Li, M. (2013), “Measuring and Testing Industry Effects in Strategic Management 31 Research: An Update, Assessment, and Demonstration,” Organizational Research Methods, 16(1), 43-46. Shepherd, W. G. (1979), The Economics of Industrial Organization, Englewood Cliffs, N.J., Prentice Hall. Silverman, B. S. (1999), “Technological Resources and the Direction of Corporate Diversification: Toward an Integration of the Resource-Based View and Transaction Cost Economics,” Management Science, 45(8), 1109-1124. Spagnolo, G. (1999), “On Interdependent Supergames: Multimarket Contact, Concavity and Collusion,” Journal of Economic Theory, 89(1), 127-139. Tallman, S. and Li, J. (1996), “Effects of International Diversity and Product Diversity on the Performance of Multinational Firms,” Academy of Management Journal, 39(1), 179-196. Tanriverdi, H. and Venkatraman, N. (2005), “Knowledge Relatedness and the Performance of Multibusiness Firms,” Strategic Management Journal, 26(2), 97-119. Teece, D. J. (1980), “Economies of Scope and the Scope of the Enterprise,” Journal of Economic Behavior and Organization, 1(3), 223-247. Teece, D. J., Rumelt, R., Dosi, G. and Winter, S. (1994), “Understanding Corporate Coherence: Theory and Evidence, Journal of Economic Behavior and Organization, 23(1), 1–30. Teece, D.J. (1982), “Towards an Economic Theory of the Multiproduct Firm,” Journal of Economic Behavior and Organization, 3(1), 39-63. Tran, H. T. and Zaninotto, E. (2012), “Product Diversification, Corporate Entrepreneurship and Firm Performance: An Empirical Study of Vietnamese Firms,” in Prodotto, consumatore, e politiche di mercato quarant’anni dopo, eds. S. Borghini, A. Caru, F. Golfetto, S. Pace, D. Rinallo, L. Visconti and F. Zerbini, Egea S.p.A: Milan, pp. 407-428. Vannoni, D. (2000), “Diversification, the Resource View and Productivity: Evidence from Italian Manufacturing Firms,” Empirica, 27(1), 47-63. Vannoni, D. (2004), “Causes and Effects of Multimarket Activity from Theory to Empirical Analysis,” Managerial and Decision Economics, 25(3), 163-174. Vella, F. (1992), “Simple Tests for Sample Selection Bias in Censored and Discrete Choice Model,” Journal of Applied Econometrics, 7, 413-421. Vella, F. (1998), “Estimating Models with Sample Selection Bias: A Survey,” Journal of Human Resources, 33 (1), 127-169. Verbeek, M. and Nijman, T. (1992), “Testing for Selectivity Bias in Panel Data Models,” International Economic Review, 33(3), 681-703. Williamson, O. E. (1985), Economic Institutions of Capitalism, Free Press: New York. Wit, B. D. and Meyer, R. (2005), Strategy Synthesis, London: Non Basic Stock Line. Wooldridge, J. M. (1995), Selection Corrections for Panel Data Models under Conditional Mean Independence Assumptions, Journal of Econometrics, 68 (1), 115-132. Wooldridge, J. M. (2002), Econometric Analysis of Cross-Section and Panel Data, Cambridge, MA: MIT Press. 