art. 3 (Yolanda Polo)


Published on

Published in: Business, Economy & Finance
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

art. 3 (Yolanda Polo)

  1. 1. Strategic Management Journal Strat. Mgmt. J., 23: 245–264 (2002) DOI: 10.1002/smj.222 FOLLOWERS’ ENTRY TIMING: EVIDENCE FROM THE SPANISH BANKING SECTOR AFTER DEREGULATION LUCIO FUENTELSAZ* , JAIME GOMEZ and YOLANDA POLO Department of Management and Marketing, Universidad de Zaragoza, Zaragoza, Spain Much of the literature dedicated to the analysis of entry timing decisions has been devoted to the study of their consequences in terms of performance. However, only a limited amount of effort has been dedicated to analyzing the factors that determine these decisions. In addition, previous papers have centered their efforts on the product dimension, paying no attention to entry into new geographical markets. This paper departs from previous works in this respect and extends the entry timing literature through a consideration of the geographical side of entry. Our analysis shows that organizational size, organizational competence, and organizational experience appear as key factors when explaining the pattern of geographic diversification. Our results also indicate that diversification takes place sequentially, first proceeding to closer locations, then occupying markets further from the origin. Copyright  2002 John Wiley & Sons, Ltd. the objective of this paper is to contribute to the INTRODUCTION existing literature and to extend it, trying to shed What factors determine the whether and when of light on all these questions in a context in which entry decisions? Are entrants different from nonen- the decisions to enter into a new market are taken trants? Which firm characteristics are more likely with respect to the geographic side of entry. to be found in an early (late) entrant? Are early A review of the strategic management literature entrants endowed with the same type of resources suggests that only a relatively small amount of and capabilities as late entrants? Do market char- research has been devoted to investigating the acteristics influence the decision and the timing at factors that affect the timing of entry into an entry? These are a sample of some of the ques- industry. Most of the papers have concentrated tions that have recently been posed in the strategic on the consequences of the entry process on the management literature. The answers to them are results obtained by the firm. While the question of important not only to managers who try to under- why firms enter industries when they do is of great stand the resources and capability requirements for strategic interest, it has received little attention entering a market and the moment in which this in the literature (Schoenecker and Cooper, 1998). must be done, but also because they help estab- This paper tries to add some empirical evidence lished firms to predict both the type of entrant they to this entry timing literature and in doing so we will face (Robinson, Fornell, and Sullivan, 1992) adopt a different perspective from that taken in and the moment of entry. Against this background, previous papers. First, we insist on the importance of analyzing the entry process itself, in contrast to placing the emphasis on the relationship between Key words: entry timing; geographic diversification; Spanish banking; savings banks; survival analysis entry timing and firm performance. To do this, we *Correspondence to: Lucio Fuentelsaz, Department of Manage- consider the whole population of potential entrants. ment and Marketing, Universidad de Zaragoza, Gran Via 2, In this regard, previous works only contemplate 50005 Zaragoza, Spain. Copyright  2002 John Wiley & Sons, Ltd. Received 2 December 2000 Final revision received 11 August 2001
  2. 2. 246 L. Fuentelsaz, J. Gomez and Y. Polo the decisions taken by a sample of diversifying According to these results, diversification seems firms, but do not consider ex novo entry (entry by to have taken place sequentially, first proceeding to closer locations, then occupying markets further newly created organizations). Our data come from from the original. As previous research has shown, a sample of Spanish savings banks, a market where market structural characteristics (potential com- no ex novo entry occurs. As a consequence, the petition) and firm-specific competitive conditions potential for bias stemming from the interaction (extent of competition in core markets) also have between the diversifying companies and the new firm entrant decisions is minimized.1 a significant influence on the time of entry into a new market. Secondly, existing research has concentrated its One important aspect of our analysis to which efforts on the product case, paying no attention to attention should be drawn concerns methodologi- the entry in new geographical markets. The result cal issues. Specifically, we maintain that survival is that the validity of the hypothesis traditionally analysis techniques are more appropriate than tra- proposed in the literature has not been tested in ditional regression approaches. Our goal is to inte- the geographical context. Geographic diversifica- grate both the estimation of the determinants of the tion decisions are very important in industries like entry decision and its timing in a single equation. retail banking, in which the customer has a close Although survival analysis has been used by pop- relationship with the firm. In these cases, the spa- ulation ecology researchers, its use is far from tial dimension of the market and the number of common in strategic management and, more par- sales points is critical, and operating in a new geo- ticularly, in the literature on entry timing deci- graphical market necessarily involves the existence sions. These models are appropriate in the cases of in-market facilities from which to deliver the in which both the cross-section side of the data service or product. In this context, the decision of (firms and potential markets/products) and the lon- where to locate regionally and the timing of doing gitudinal side (time) are available. This is an espe- so appear as key strategic issues faced by firms cially interesting issue given that recent attention operating in these industries. has been devoted to the methodology used for To the best of our knowledge, our research is the the analysis of diversification decisions (Merino first attempt to cover this area. Furthermore, our and Rodr´guez, 1997), as well as the lack of ı sample is especially appropriate for this test, given dynamic studies modeling the strategic interactions that, as in the United States and other European that take place between firms (Porter, 1991; Baum countries, branching was not allowed for Spanish and Korn, 1996). Survival analysis methods have savings banks until recently. In order to test our also been proposed as a more convenient alter- predictions, we examine the entry timing proba- native when examining the problem we are con- bility of industry incumbents in the Spanish sav- cerned with (Lieberman and Montgomery, 1998). ings banks sector as a function of the bundle of In this context, we use the Cox Proportional Haz- resources and capabilities of the firms operating ards Model (CPHM) as a more general, flexible, in the industry and the structural features of the and efficient way of dealing with these situations. geographic markets. Following previous research (Barnett, 1993; Baum and Korn, 1996), we also consider some aspects of their specific competi- LITERATURE REVIEW tive conditions and the degree of competition faced in their core markets (Mitchell, 1989). Our results The literature on entry timing decisions has mainly show that organizational size, organizational com- been concerned with two questions: (1) Are there petence, and organizational experience arise as key any linkages between the moment of entry and factors when explaining the pattern of geographic the performance of the firm in the new market? diversification in the Spanish savings banks sector. (2) What factors influence entry timing decisions? Research effort has concentrated on the first of 1 As Dunne, Roberts, and Samuelson (1988) point out as one these and, as a result, much less attention has been of the criticisms of the Robinson and Fornell (1985) study, devoted to the second. The objective of this paper diversification entry only accounts for about 45 percent of the total entry in U.S. manufacturing firms. As a consequence, if is precisely to help to fill this gap in the literature, the decisions taken by the diversifying companies are affected and in this section we review the discussion that by the new firm entrant decisions, the results might be partially has taken place in the search for an answer. biased. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  3. 3. Followers’ Entry Timing 247 such a determinant aspect of firm performance as Entry timing and performance might first seem in principle. Thus, the magnitude As we have already indicated, the literature of these effects varies among categories of prod- has mainly been concerned with the effect of ucts and geographical markets. Furthermore, first- the timing of new product introduction on firm mover advantages disappear with time, although performance. It has been argued that entry they are improved by longer leading times before timing strategy is a key determinant both of other competitors enter the market. Finally, they the absolute success (or failure) of a product are weaker than price and advertising effects and and of performance relative to competitors. To they may be better modeled as interaction effects date, the theoretical discussion has focused on the rather than direct effects (Lieberman and Mont- advantages and disadvantages of being a pioneer, gomery, 1998). as opposed to a follower, and the benefits and risk associated with the decision taken (Lieberman and Montgomery, 1998). Factors affecting entry timing decisions Early entry may provide the basis for the acqui- A significantly less amount of effort has been sition of superior resources and capabilities. Pio- devoted to studying the factors affecting entry- neers may preempt competitors in the physi- timing decisions and, as Schoenecker and Cooper cal, technological, or consumer perceptual space (1998) point out, just a few papers have specifi- (Lieberman and Montgomery, 1998). They may be cally considered this issue. able to attract and retain high-quality employees, It should first be noted that research has focused occupy superior locations, or enjoy the benefits of on the product side of entry, forgetting the pos- patent protection. In addition, several studies point sibility of entry in a new geographical market.2 to the fact that the first firm to enter the market As a result, theoretical propositions and empirical obtains benefits in terms of recognition and repu- test have not been developed in the geographical tation (Lilien and Yoon, 1990). Pioneers may also context. As mentioned earlier, geographic diversi- be able to establish switching costs that make con- fication decisions are important in industries like sumers imperfectly mobile among competing firms retail banking, in which the customer has a close (Lieberman and Montgomery, 1998), or obtain relationship with the firm. In these cases, the spa- advantages through cost experience effects (Lilien tial dimension of the market and the number of and Yoon, 1990). sales points is critical and operating in a new geo- However, early entrants also have to assume graphical market necessarily involves the existence the risks stemming from the development of the of in-market facilities from which the services product and the market, and their innovations are or products are delivered. In contrast to previous subject to imitation by competitors at lower costs. studies, we do not restrict our analysis to a sit- Therefore, the decision to enter the market should uation in which the product is completely new be timed to balance the risk of premature entry and to the market, but rather to one in which the the problems of missed opportunities (Lilien and entity is. Yoon, 1990: 568). Research on the factors affecting entry-timing Empirical tests of the benefits of early entry decisions has mainly highlighted the importance have produced conflicting results. Mitchell (1991) of differential firm resources and capabilities. argues that at least some of the confusing empirical Optimal entry timing frequently depends on the results stem from the joint consideration of firms resource base of the firm (Lieberman and Mont- with different characteristics and, therefore, differ- gomery, 1998). Thus, pioneers have resources and ent entry timing incentives and effects on perfor- capabilities different from late entrants (Robinson mance. On the other hand, Vanderwerf and Mahon et al., 1992). However, the few papers published (1997) show that the probability of detecting first- in the management literature have given much mover effects largely depends on the methodology used, and Golder and Tellis (1993) point out that some of the influences are due to the fact that early 2 Some studies undertake this task in an international context (see, for example, Mascarenhas, 1992). Although conceptually followers are frequently identified as pioneers. similar, the problems analyzed in international diversification The conclusion from all this empirical literature settings are more varied and different, in many respects, from is that, while first-mover effects exist, they are not the ones analyzed here. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  4. 4. 248 L. Fuentelsaz, J. Gomez and Y. Polo less emphasis to the consideration of industry spe- FOLLOWERS’ ENTRY TIMING cific characteristics. Therefore, the literature has HYPOTHESES been mainly concerned with the hypothesis related to differences in functional skills, resources, and Our data come from a market where no ex novo organizational attributes possessed by the entrant.3 entry occurs. As pointed out in the previous The consideration of industry features should be section, the paper considers a situation in which the important when explaining entry and timing deci- firm is completely new to the geographical market sions. If the likelihood and the timing of entry are in which it wishes to operate. However, other firms influenced by the strengths and weaknesses of the may be acting as incumbents in this market. There- firm’s existing resources and capabilities, industry fore, in this sense, given that entrants are not the characteristics capture the threats and opportuni- first to operate in the market and, following the ties firms face and moderate the effect of the for- terminology of entry timing literature, they may mer variables. Following Robinson et al. (1992: be characterized as followers. 609) ‘a strategic window occurs when the best In the line of previous research on the fit arises between a market’s key success require- determinants of entry timing decisions, our ments and specific firm competencies.’ Therefore, hypotheses consider the importance of differential firm characteristics and industry features should be firm resources and capabilities. The resource- taken into account when explaining the whether based view of the firm (Penrose, 1959; Wernerfelt, and when of entry decisions. 1984; Barney, 1986) and the dynamic capabilities The third and final aspect to be highlighted in approach (Teece, Pisano, and Shuen, 1997) have this review is that, in many cases, the studies have analyzed in depth the role that resources and renounced the possibility of integrating in the anal- capabilities play in the sustainability of firms’ ysis both the decision of whether or not to enter, competitive advantages, and their consequences and the timing of that decision, with only a lim- in terms of firm efficiency. The relevant ited number of studies (Mitchell, 1989; Baum and question in this context is to determine the Korn, 1996) taking both dimensions into account. conditions that allow early entrants to accumulate As Morita, Lee, and Mowday (1993: 1431) point superior resources and capabilities (Lieberman out, ‘the joint prediction of whether and when an and Montgomery, 1998). Unlike other papers on event occurs with a single equation is statistically entry-timing decisions, we also explicitly include more complex than simply predicting whether the industry characteristics, in this way capturing the event occurs.’ This has had two immediate conse- threats and opportunities that firms face.4 quences. First, the samples used tend to consist just of firms that have actually entered the new mar- Firm characteristics that determine entry ket. Entrants are divided into three categories (pio- timing neers or first entrants, followers, and late entrants) and research has centered on the determinants of Resource constraints can be a determining aspect the order of entry. Second, the methodology used of entry and, therefore, depending on the evolution has (with some exceptions) been the traditional of the firm, of its timing. In this paper, we distin- one: either multiple regression or multinomial logit guish between two types of resources: those that models. This implies that the longitudinal side of provide the firm with financial services and those the data has frequently been ignored. As we will related to the supply of information. see, in this paper the analysis is widened through the examination of both problems at the same Firm size and firm profitability time, using survival analysis techniques and con- sidering all the firm–market–year possibilities of Size is frequently used in the literature on diver- entry. sification and entry timing decisions (Hannan, 1979; Ingham and Thompson, 1994; Merino and Rodr´guez, 1997; Schoenecker and Cooper, 1998). ı 3 In only one of the studies (Schoenecker and Cooper, 1998) the authors claim to be the first to incorporate both industry and firm characteristics as determinants of entry timing. However, 4 Although the use of industry characteristics in the analysis of although they argue for differences between the minicomputer and the PC industries sampled, the analysis does not explicitly entry is frequent in the economics literature, they have been include industry characteristics as a variable in the model. largely ignored in the entry timing literature. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  5. 5. Followers’ Entry Timing 249 The main reason is that organizational size may especially important in the presence of capital mar- be considered as a proxy for the amount of slack ket imperfections, given that these could provide resources available to a firm. Larger organizations firms with internally generated funds to finance generally possess more slack resources, which their expansion and surmount entry barriers. As in favor adaptation to changing circumstances (Cyert the case of larger firms, more profitable organiza- and March, 1963; Thompson, 1967; Haveman, tions are expected to be favored by the possession 1993a) and help them to initiate strategic change of slack resources that increase the probability of (Bourgeois, 1981). Slack resources not only make successful entry and reduce the likelihood of fail- organizations more fluid (Haveman, 1993a), but ure (Bourgeois, 1981; Hannan and Freeman, 1989). also reduce the risks associated with the process of Therefore, our two first hypotheses are proposed experimentation with new strategies, new products, in the following terms: and new markets, reducing the likelihood of failure (Hannan and Freeman, 1989, Haveman, 1993b). Hypothesis 1a: Firm size is expected to have a positive impact on the speed of entry. A related argument holds that organizational size increases market power, lowering the barriers to Hypothesis 1b: Profitability is expected to have entry in any market. Market power facilitates over- a positive impact on the speed of entry. coming the entry barriers associated with prod- uct differentiation, absolute costs advantages, and economies of scale (Bain, 1956). Thus, larger firms Proximity and sequential learning may benefit not only from the possession of slack Any organization will enter a new market quickly resources that may help to satisfy capital require- when it possesses core capabilities that may be ments associated with entry, but also from the used in it (Brittain and Freeman, 1980; Lambkin, existence of a strong brand reputation or from cost 1988). The literature on diversification and inter- reductions related to size. national business has emphasized the idea that In contrast to this view, some authors have entry into new markets may be better understood argued that, when faced with change, larger orga- as a sequential process (Chang, 1992; Chang and nizations tend to be rigid and inflexible (Have- Rosenzweig, 1998). Firms sequentially enter new man, 1993a). Larger organizations may suffer from businesses, first proceeding to related markets and bureaucratic inertia (Crozier, 1964), which may then looking for the unrelated ones as they accu- delay entry. This bureaucratization arises from the mulate experience and build new internal capa- standardization and formalization of processes that bilities. As firms proceed one step further in the take place as firms grow (Haveman, 1993a). Nev- sequence, they are able to accumulate knowledge ertheless, the empirical results usually show that and build competencies in previously unrelated size does have a positive influence on the decision markets, thereby increasing the likelihood of suc- and the timing of entry (Schoenecker and Cooper, cessful further entries and reducing the probability 1998; Robinson et al., 1992; Mascarenhas, 1992). of failure. This is the effect we will postulate in our first hypothesis. Related markets that contribute to the group of potential entrants include those that are vertically The second variable usually considered when related, those with a similar technological path, evaluating entry-timing determinants is profitabil- and those with a similar product in a different ity. Both the theoretical and empirical develop- geographical area (Geroski, 1991). In the last cate- ments in static frameworks tend to show that more competent firms are more likely to enter a new gory, physical proximity between markets arises as market (Jayaratne and Strahan, 1998; Cotterill and a key issue determining the capabilities of poten- Haller, 1992). In spite of the fact that some authors tial entrants to operate in them, the potential for success and, hence, the sequence of entries fol- point to poorly performing firms as tending to be risk seekers (Bowman, 1982; Fiegenbaum and lowed by firms. On the one hand, for a firm to be Thomas, 1986), more competent firms should have interested in a market it has to access, understand, higher expected profitability from the operation in and economically value the information about that any additional market and, therefore, should be the market. Information is costly and investment in first to respond to environmental changes stimu- information collection necessarily assumes knowl- lating entry. Differences in profitability should be edge about how to look for it. Those close to a Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  6. 6. 250 L. Fuentelsaz, J. Gomez and Y. Polo collective knowledge (Penrose, 1959). Previous market are better positioned in this respect. On the multimarket experience may serve to build capabil- other, they are not only better located to act, but ities, to create effective routines and to accumulate also to ‘listen’ to the information pointing towards knowledge about how to manage operations across an opportunity. markets. Therefore, firms with greater accumulated Cotterill and Haller (1992) follow a similar experience in managing operations across markets argument, maintaining that proximity is a mea- should exhibit a greater likelihood to enter more sure of potential entrant capability. Entrants with quickly into new geographical markets (Wilson, closer headquarters are more able to better control 1980). activities, may be known to customers, and may We expect knowledge accumulation benefits have better knowledge of objective markets. This coming from multimarket operation to accrue in a last point is important, given that uncertainty about decreasing way. For a specialized firm, previous revenues may vary from geographical to geograph- experience of operating in several markets may ical market. As pointed out by Pindyck (1991), it be a key aspect for new entry success, but this is the fact that almost any investment has (at least) is not so relevant for a highly diversified entrant. a partially sunk character which makes uncertainty By contrast, coordination and monitoring costs are so relevant. If the firm has an option to wait for expected to rise with the number of markets in new information to arrive, the delay in entry may which a firm is present. This quadratic influence of reduce the effect of uncertainty. Given that proxim- the degree of diversification on the likelihood of ity increases the knowledge of markets, this option further entries may be especially prevalent when to wait may not be so valuable, and first-mover the knowledge provided by the operation in any advantages and strategic factors may compel firms additional market is similar to that of previous to quickly enter a new market. diversification efforts (diversification into closely Therefore, entry into new geographical markets related markets). In this case, a higher marginal may be viewed as the exercise of a sequence degree of prior diversification may provide the firm of strategic options (Chang, 1995; Bowman and with few of the benefits of knowledge accumula- Hurry, 1993). When a firm enters a new geograph- tion, but all the costs. ical market, it purchases an option to enter in a pre- This reasoning suggests a quadratic pattern of viously unrelated market. As time elapses and the influence of the degree of previous diversification knowledge of closer markets increases, the option on the entry timing behavior of firms. Therefore, to wait reduces its value and the entered market our third hypothesis is proposed as follows: may be used to proceed further in the sequence of entries. On this basis, we propose our second hypothesis: Hypothesis 3: The number of markets in which a firm operates is expected to have an inverted U-shaped influence on the speed of entry. Hypothesis 2: Proximity to the objective market is expected to have a positive impact on the speed of entry. Firm-specific competitive conditions and market determinants of entry timing Experience in managing operations across As pointed out earlier, the consideration of indus- markets try features should be important when explaining entry and entry-timing decisions, given that indus- As firms proceed in the diversification process try structure, both in the origin and the host mar- they not only acquire knowledge about previously ket, will condition a firm’s competitive strategies. unrelated markets, but they also obtain experience Thus, the following factors will be considered as in managing operations across markets. Multimar- determinants of entry timing strategies. ket and multiproduct organizations are expected to be more proficient in overcoming the prob- lems derived from the difficulties in integrating Extent of competition in core markets and monitoring a set of projects in different mar- Competition stems from firms interacting and striv- kets, due to their accumulation of experience. This ing in the same environments and for the same idea agrees with conceptions of growth based on resources. The specific competitive conditions that evolutionary processes and the accumulation of Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  7. 7. Followers’ Entry Timing 251 firms face in their domains are a key aspect and Haller, 1992; Geroski and Schwalbach, 1991) determining interfirm rivalry and their subsequent and, following it, in this paper we will distinguish entry behavior (Baum and Korn, 1996). When between actual and potential rivalry. rivalry is high, firms have incentives to diversify Actual rivalry is frequently measured through in order to reduce the level of competition for the the use of concentration indexes. We could expect resources on which they depend. different effects of concentration on the entry The literature on entry timing decisions has also decision (Cotterill and Haller, 1992) depending on underlined the importance of the situation in the whether we follow different lines of argument. On core activities of the firm when explaining entry- the one hand, high levels of concentration may be timing strategies. Some authors have argued that used as a proxy for high profit expectations. In emerging markets presenting an important threat contrast, oligopoly theory suggests that a smaller to the central markets of the firm would increase number of firms may facilitate coordination prac- the probability of entering the new market earlier tices directed to deter entry (Scherer and Ross, (Mitchell, 1989; Schoenecker and Cooper, 1998; 1990). Finally, the arguments offered by the con- Cooper and Smith, 1992). In line with this litera- testable markets theory and the Chicago School ture, this present paper distinguishes between core point to the fact that concentration should not have and new geographical markets. By core geograph- any effect on the probability and timing of entry. ical markets, we refer to those markets in which For its part, potential rivalry should be disin- the firm has been traditionally operating and that centive to new entry. Different arguments indicate constitute the main source of revenues and profits. that the number of potential entrants may influ- As in the product case, and given the commitment ence competitive actions (Bain, 1956; Mitchell, of the firm to these markets, the survival of the 1989). An illustration is provided by Shermand and firm fundamentally depends on them. Therefore, Willet (1967) under a perfect information frame- we would expect that when the market domain of work: as the number of potential entrants increases, a firm becomes seriously threatened by the extent the probability of entry and that incumbents dis- of competition, the firm would tend to secure its cipline new entrants both increase, making entry survival by looking for new markets. In this sense, less attractive. rivalry in the core market may be an indicator According to this reasoning, potential rivalry about the extent to which the central activities of should have a negative effect on the timing of entry the firm are threatened and the need to look for decisions. However, the effect of actual rivalry new locations in which to develop these activities. has to be determined by the evidence offered by Therefore, our fourth hypothesis is as follows: our data. Thus, we propose the following two hypotheses: Hypothesis 4: A higher rivalry in the core geo- graphical markets of the firm is expected to have Hypothesis 5a: A higher actual rivalry in the a positive impact on the speed of entry into new objective market is expected to have a posi- markets. tive/negative/no impact on the speed of entry. Hypothesis 5b: A higher potential rivalry in the Rivalry in the objective market objective market is expected to have a negative Apart from market conditions in the home market, effect on the speed of entry. the evolution of rivalry in the new market in which the firm wishes to operate should also influence the Demand potential probability and entry timing. Mitchell (1989: 208) argues that the probability and entry timing into Following Lilien and Yoon (1990), entry timing emerging technical subfields will be influenced decisions fundamentally depend on the level of by the extent of the rivalry to secure the value market potential and its evolution (demand growth) of specialized assets. Potential followers should over time. Intensity of demand and market growth also consider the competition coming from other are also variables frequently considered in the potential entrants (Lilien and Yoon, 1990: 568). literature. A higher intensity of demand will lead The economic literature has also considered this to higher expected profits in the market. Cotterill issue when analyzing entry decisions (Cotterill and Haller (1992) point out that market growth Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  8. 8. 252 L. Fuentelsaz, J. Gomez and Y. Polo may have two simultaneous consequences on entry model for the analysis of survival data, mainly conduct: on the one hand, a higher rate of market due to its flexibility. The model gives very efficient growth can lead to excess, higher expected profits estimates as compared to a parametric proportional and easier entry; on the other, the attractiveness hazards model, even when the data actually come of entry would be reduced by the larger queue of from the parametric model (Efron, 1977). Fur- potential entrants. Given that we have controlled thermore, and contrary to other survival analysis for the second effect in our previous hypothesis, methods, the model does not require any underly- we expect the effect of increased attractiveness to ing distribution to be specified, being more general be prevalent and, therefore, a positive sign of the in nature. Its main assumption is that the hazard associated variable. functions of all the individuals are a multiple of an unspecified baseline hazard function. Therefore, the baseline hazard function is an arbitrary and Hypothesis 6a: A higher intensity of demand non-negative function in time. If Xi (t) is the vec- in the objective market is expected to have a tor of covariates for the ith individual at time t, positive effect on the speed of entry. the model assumes that the hazard for a subject takes the following form: Hypothesis 6b: A higher market growth in the objective market is expected to have a positive λ(t; zi ) = λ0 (t)ri (t) effect on the speed of entry. Table 1 summarizes both the hypotheses con- where sidered in this paper and the expected signs of the ri (t) = exp(βzi (t)) effect over the speed of entry into new geograph- ical markets. is referred to as the risk score for the ith subject, β is a vector of regression parameters and λ0 (t) SURVIVAL ANALYSIS AND THE COX is the baseline hazard function. The model does PROPORTIONAL HAZARDS MODEL not include a constant term, given that this is incorporated in λ0 (t). Its estimation is carried out To test the hypotheses just proposed this research through the maximization of the partial likelihood uses the Proportional Hazards Model proposed by function (Cox, 1975), with the procedure being Cox (1972). The Cox Proportional Hazards Model very similar to that of maximum likelihood. (CPHM) is the most commonly used regression Several extensions of the CPHM have been developed from Cox’s original. Thus, the basic model has been extended to the consideration of Table 1. Hypotheses and expected signs competing events, multiple events, discontinuous intervals or risk, different time scales or time- Hypothesis Reference Expected effect on probability and dependent covariates that are very useful in prac- timing of entry tical terms. An extension that deserves special attention is the one formulated by Andersen and + Size 1a Gill (1982). These authors consider each subject + Profitability 1b as an observation of a very slow Poisson process, + Proximity 2 in which a censored subject is taken into account as ∩ Previous experience 3 one whose event count is still zero. It is their for- + Extent of competition 4 mulation that we have chosen to use in this paper. in core markets Before applying this model to our data, we +/−/0 Actual rivalry in 5a should make one particular observation. The model objective market presented here is appropriate for survival data − Potential rivalry in 5b that have a continuous character. However, the objective market availability of continuous data is not the rule in + Intensity of demand 6b management and economics in general and, more + Relative market 6c specifically, in the analysis of entry timing deci- growth sions. In practice, data are rarely measured more Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  9. 9. Followers’ Entry Timing 253 precisely than in days and the most common case In general, the maximization of the partial like- is perhaps the one in which we have annual infor- lihood is shown to be extremely robust in the mation about the covariates and the event of inter- presence of tied deaths. What is more, the method est. This raises the practical question of whether tends to be superior to that of the maximum like- to use continuous or discrete time models. When lihood when not many events take place at the data are grouped and tied deaths are present (i.e., same time (Brostr¨ m, 1997). Therefore, with a few o when two or more observations have events at the ties in the data the continuous proportional model same time), the construction of the partial likeli- (with either the Breslow or Efron approximations) hood function is more complicated. For this reason, continues to be more appropriate for the analysis extensions of the CPHM have been proposed for of survival data, even if duration is measured the cases in which time is measured in discrete in discrete intervals. However, when many tied intervals. deaths are present, the alternative of estimating Cox (1972) gave the first alternative in his semi- a discrete time model must be used. In this last nal paper, namely that the use of logistic regression case, when probabilities are small, it does not mat- may serve as an approximation to the true model. ter which model (logit, discrete analog) is selected This method is attractive because, as the discrete (Brostr¨ m, 1998). o time units get smaller, the result converges to that of the true Proportional Hazards Model (Thomp- son, 1977). Nevertheless, two problems stem from SAMPLE DESCRIPTION this approach (Prentice and Gloeckler, 1978): first, the regression parameter no longer has a rela- Our empirical analysis will be carried out using tive risk interpretation; second, the meaning of data from the Spanish savings banks sector that the regression coefficient depends on the choice of describe branch network expansion after restric- grouping intervals. Prentice and Gloeckler (1978) tions were eliminated. Prior to deregulation, one or provide a second alternative, through the consid- two savings banks were competing in each provin- eration for the grouped data version of the Propor- cial market.5 Thus, every province had at least tional Hazards Model (Kalbfleisch and Prentice, one established saving bank, which implies that no 1973). Their extension assumes that events are province was free of incumbents. This is impor- grouped into intervals. The regression vector is tant, in the sense that it constitutes a difference allowed to be time-dependent, but it remains fixed from the usual models of entry. The risk supported within a specific interval. Estimation takes place by the new entrant is then reduced, given that the through maximum likelihood techniques. Finally, customer already has some information about the a third extension considers several assumptions characteristics of the product offered by the new about the order in which the event takes place firm. Therefore, all the firms that entered a new when ties are present. This approach has been pro- market were followers and, as a consequence, the posed by Breslow (1974) and Efron (1977) and relevant distinction we draw is between early and attempts to approximate the true partial likelihood late followers (it is not strictly possible to talk when the information contained in the data does about first movers). A last feature of this market not allow an ideal construction. The difference is that new financial entities have not been created between both approaches rests on the assumptions up to now, which implies that our data come from made about the form that the partial likelihood a sample where no ex novo entries occur. function should take. In this context, several advantages from early Therefore, in the availability of discrete data, the following could be relevant when considering the selection problem is reduced to one with two main expansion process that has taken place in the Span- options, with the presence of tied deaths being ish banking sector. As in any other industry, early the main guiding criterion: on the one hand, the entry could provide the basis for the acquisition of use of partial likelihood techniques, together with superior resources. First entrants could have been the Breslow or Efron approximations to handling able to place their branches in superior locations, ties; on the other, the alternative of estimating the model through maximum likelihood techniques 5 Spain is divided into 17 autonomous regions. Each region is (either the logit or the discrete analog to the Pro- divided into provinces, of which there are a total of 50. Data are portional Hazards Model). only available at the province level. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  10. 10. 254 L. Fuentelsaz, J. Gomez and Y. Polo preempting competitors in the physical space. This Savings Banks Confederation (CECA). The data is especially important in retailing, given that dis- have a longitudinal dimension, in that information tance to customer appears as a very important com- about the covariates and the dependent variable is petitive variable. Preemption could also take place provided every year. A savings bank and its asso- in the consumer perceptual space. Given that cus- ciated characteristics are identified and followed tomer relationships with the firm are fundamentally from 1986 to 1996. Sampling has been affected by based on confidence, recognition and reputation the mergers and acquisitions that took place at the might have played a very important role. Finally, beginning of the 1990s, which reduced the num- the existence of switching cost could also have ber of savings banks from 77 entities in 1986 to been a key issue in considering entry timing. 50 in 1996. The number of observations is given As mentioned earlier, the Spanish banking sector by all the possible combinations of savings bank- is especially interesting given that recent deregu- year-market in which the entity is not present at lation has changed the competitive environment in the beginning of the follow-up. This provides us which the firms operate, with the subsequent effect with 27,956 single entity-market-year observations on their strategies. Before branching deregulation from which 153 correspond to effective entries. took place, the geographic scope in which savings Figure 1 presents the sampling plan. The left- banks operated was restricted to a local, provin- hand side of the figure shows the years considered cial, or regional character.6 Deregulation allowed in the study, while the right-hand side records the entities to look for new markets in which to duration. If we look at the former, three basic perform their activities. Against this background, situations are possible. Case (1) corresponds to an the objective of our analysis is to shed light on the entity that has not been involved in any merger factors that have determined the decision of a firm or acquisition and, therefore, is followed through to enter or not into a new geographical market and, all the sampled period. Case (2) samples an entity given the former, the variables affecting the timing (top line) that has been integrated by a second one of the entry decision. (bottom line). Finally, case (3) shows the situation The data available for analysis belong to the in which two entities (first upper lines) merge and period 1986–96. Although the total elimination of give rise to a third (bottom line). As has been restrictions did not take place until 1989, a few mentioned earlier, in all cases an observation is years before savings banks had been authorized included in the sample either up to the time at to compete within their own Autonomous Region which the event takes place, or the time horizon and to establish some branches in the five biggest of the study ends (censoring). cities in Spain.7 This advises that we anticipate As we can also appreciate from Figure 1, sam- the time of analyses until 1986 in order to avoid pling is affected by censoring. Censoring may take the potential bias stemming from the existence place for different reasons. In general terms, it of left censoring. In any event, some entities had occurs when the information available during the opened branches (usually one or two) before that risk period is incomplete, due to limitations in the date. Therefore, in order to avoid distortions in the observation window. In the case of our sample, an analysis we only consider that an entity is present observation may be censored either when a savings in a market when it has, at least, 1 percent of the bank is involved in a merger (or acquisition) and branches of this market.8 legally disappears as a consequence of it, or due to Our study makes use of public data, coming the fact that information prior to 1986 or after 1996 from the Bank of Spain (BE) and the Spanish is not incorporated in the analysis. As mentioned before, in this context a key feature of Survival 6 In 1986, 48 savings banks (out of 77) were competing in just Models is their capacity to handle this type of lim- one market (province), 11 in two markets, 5 in three and 8 in four. The remaining 5 entities competed in five to nine provinces. itation in the data. For noncensored observations, 7 The law eliminating branching restrictions was, in fact, issued models incorporate information about the corre- at the end of 1988 (R.D. 1582/1988 December 29 Law). sponding duration. For those observations that are 8 The number of branches eliminated by the application of this censored, all we know is that the event under study criteria is 120 (1.1% of the total number of branches in 1986), mostly situated in Madrid (55), Barcelona (31), Valencia (12), (entry) has not taken place up to the censoring and Zaragoza (7). As a consequence, by operating in this way time and, therefore, the models consider that the we only eliminate 15 branches (0.14%) not situated in the four duration exceeds that limit. mentioned provinces. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  11. 11. Followers’ Entry Timing 255 Case 1 Case 2 Case 3 86 96 0 10 Years Duration Figure 1. The sampling plan Table 2. Hypotheses and measurement Hypothesis Variables H1a: Size LNSIZE: Natural logarithm of own resources understood as the sum of capital, rotation fund, reserves, subordinated financing, and retained earnings (end of previous year) INTER: Net balance (as a lender) of savings banks in the money market, normalized by the average assets of the entity (end of previous year) H1b: Profitability PROFI: Net profit divided by total assets (end of previous year). H2: Proximity PROXIC: Dummy variable that takes a value of one when the province is adjacent to the original (core) market of operation of the savings bank (end of previous year) PROXIE: Dummy variable that takes a value of one when the province is adjacent to the new markets entered by the savings bank (end of previous year) H3: Experience NMARK: Number of provinces in which the entity was operating at the end of the previous year H4: Extent of competition in COREH: Overall Herfindahl (banking system) in the provinces in core markets which the entity was initially operating (end of previous year) H5a: Current rivalry in the OBJEH: Herfindahl (banking system) in the objective market (end of objective market previous year) H5b: Potential rivalry in the NUMSB: Number of savings banks operating in the provinces objective market adjacent to the objective market (end of previous year) H6a: Intensity of demand DEPHA: Deposits per inhabitant in the objective market (end of previous year) DENSI: Density of population in the objective market H6b: Market growth MGROW: Average relative growth of bank deposits in the objective market in the 3 years previous to the one considered In relation to the variables used in the analysis, at the beginning of the year under observation in order to avoid endogeneity problems. To test the dependent variable takes a value of one for the Hypothesis 1a, we created two variables: LNSIZE entity–market–year observation in which an entry and INTER.9 LNSIZE attempts to approximate the is recorded and zero otherwise. Table 2 shows the description of the independent variables that were considered to test the hypothe- 9 As Hannan (1979) noted in his analysis of North American sis formulated above. All the covariates are taken banks, ‘although very small banks show almost no tendency Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  12. 12. 256 L. Fuentelsaz, J. Gomez and Y. Polo own resources of the entity.10 For a given size, the this, a province Herfindahl index was first devel- oped using the number of branches as a proxy relative magnitude of the position as a net lender in for market share. The core market Herfindahl the money market (INTER) will also be relevant, index was then calculated by multiplying each sin- as it provides incentives to look for new markets gle Herfindahl index of the provinces in which in which to allocate the money in the form of the entity was operating in 1986 by the relative loans at higher rates. On the other hand, profitabil- importance of the province for the entity under ity (PROFI, Hypothesis 1b) is measured through observation (the number of branches was again net profit divided by total assets. used to measure the importance of the province Hypothesis 2 expects physical proximity to play for the entity).12 Province Herfindahl indexes were a relevant role in determining the timing of new again used to proxy for current rivalry in the objec- entries. Thus, firms first proceed to closer loca- tive market (OBJEH, Hypothesis 5a). In order to tions, acquiring knowledge about local markets, and then enter further markets over time. Obvi- take into account the interaction between banks ously, the acquisition of knowledge about new and savings banks, this index has been calcu- markets takes time, which suggests a distinction lated considering the branches of both types of between the markets that were initially close to intermediaries. the core markets of the savings bank and those in Hypothesis 5b postulates that potential rivalry which proximity stems from expansion. Therefore, in the objective market should have a negative in order to test Hypothesis 2, two dummy variables influence on the timing of market entry. The mea- were created. PROXIC takes a value of one when surement of this hypothesis is complicated, given the objective market under study was geographi- the difficulties associated with the identification cally adjacent to the core market of the savings of a potential entrant and the need to follow a bank and zero otherwise. According to our reason- sensible method. In our case, the arguments antic- ing, a second variable (PROXIE) was constructed ipated in Hypothesis 2 were used and the poten- in order to take into account the proximity arising tial competition associated with each market was as a consequence of diversification. This variable defined in terms of proximity. Therefore, in order takes a value of one when the market is adjacent to measure potential competition, we first identi- to the new markets entered by the savings banks. fied the markets in which the firm was not present Following our hypothesis, we should expect the and the degree of potential competition associ- effect of this variable (PROXIE) to be of a lower ated with any market was then obtained by count- magnitude than the first (PROXIC). ing the number of savings banks surrounding it The capacity of a firm to manage operations (NUMSB). across markets (Hypothesis 3) is measured by Finally, intensity of demand in the objective the number of markets in which the organiza- market (Hypothesis 6a) is measured through two tion was operating at the end of the previous variables: DEPHA accounts for the amount of year, NMARK.11 Rivalry in the core geographi- deposits per habitant in the market, while den- cal markets (Hypothesis 4) is measured through sity of population (DENSI) is a variable fre- a Herfindahl type index (COREH). To calculate quently used in the literature on bank branching to proxy for population concentration. With respect to Hypothesis 6b (market growth in the objective to branch, large banks, particularly those with deposits in the neighborhood of $300 million or more, are frequent initiators of market), we calculated the average relative growth new branches.’ in the deposit market for the 3 years prior to that 10 Given that savings banks have a mutual character, they can- under observation (MGROW). not have access to the most common sources of funding and, therefore, the expansion has to be financed with internal funds Table 3 and Appendix 1 offer a descriptive or subordinate financing. As in other papers, we use natural log- statistics of the independent variables used in the arithms in order to reduce the skewness of the size distribution. 11 analysis. Monetary variables are measured in con- In addition to the arguments offered above, Economides, Hub- bards, and Palia (1996) point out, in a context of monopolis- stant terms (having been deflated by the IPC: Span- tic competition, a ‘large’ bank (defined as one which operates ish Inflation Index). branches in more than one market) will have, ceteris paribus, higher expected profits than a ‘small’ bank (in their definition, one with operations in just one market) and, therefore, a higher 12 Note that this measure of core market competition is firm likelihood of expanding activities towards new markets. This is specific, in the sense that it takes different values for any two due to the fact that small banks need to hold more equity capital than large banks (per branch), because they are not as diversified. different firms. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  13. 13. Followers’ Entry Timing 257 Table 3. Descriptive statistics Variable Mean S.D. Lowest Highest No. obs. H1a: LNSIZE 8.81 1.21 5.21 12.46 27,956 −0.10 H1a: INTER 0.10 0.08 0.48 27,956 −2.22 H1b: PROFI 1.02 0.57 3.53 27,956 H2: PROXIC 0.10 0.30 0.00 1.00 27,956 H2: PROXIE 0.02 0.12 0.00 1.00 27,956 H3: NMARK 2.62 2.97 1.00 49.00 27,956 H4: COREH 0.13 0.05 0.06 0.35 27,956 H5a: OBJEH 0.15 0.05 0.06 0.35 27,956 H5b: NUMSB 5.74 3.82 0.00 20.00 27,956 H6a: DEPHA 0.56 0.20 0.20 1.45 27,956 H6a: DENSI 106.71 137.81 9.02 626.45 27,956 −1.89 H6b: MGROW 5.67 3.54 23.06 27,956 Table 4. Results RESULTS Explanatory AG (1) AG (2) AG (3) Table 4 shows the results of applying the CPHM variables Efron Efron Efron over the 27,956 available observations. As was 1.399∗∗∗ 1.404∗∗∗ 1.214∗∗∗ LNSIZE mentioned in the previous section, the dependent (14.574) (12.361) (8.649) variable takes a value of one for the market and 6.601∗∗∗ 6.601∗∗∗ 7.069∗∗∗ INTER year in which the entity under observation has (3.975) (3.972) (4.238) performed an entry. The ties index takes a value of PROFI 0.116 0.116 0.023 0.004513 and, therefore, following our discussion (0.764) (0.765) (0.146) 2.464∗∗∗ 2.465∗∗∗ 2.411∗∗∗ PROXIC on continuous and discrete time models, the partial (11.349) (11.322) (10.969) likelihood method was chosen. The table includes 1.839∗∗∗ 1.845∗∗∗ 1.751∗∗∗ PROXIE the results for the Andersen and Gill formulation of (7.042) (6.753) (6.428) −0.001 0.092∗∗ the model, using the Efron (1977) approximation NMARK (0.077) (2.110) for handling ties.14 As mentioned before, the model −0.002∗∗ NMARK2 does not include a constant term because it is (−2.182) incorporated in λ0 (t). −13.885∗∗∗ −13.933∗∗∗ −12.876∗∗∗ COREH The first point to note from Table 4 is that (−4.358) (−4.289) (−3.958) −1.919 −1.910 −1.634 the model is globally significant, confirming its OBJEH (−0.969) (−0.962) (−0.822) validity. Given the high correlation between the −0.101∗∗∗ −0.102∗∗∗ −0.103∗∗∗ NUMSB variables LNSIZE and NMARK (0.51), column 1 (−3.593) (−3.593) (−3.632) presents a first model in which the second vari- −0.444 −0.439 −0.319 DEPHA able is not included in the estimations. Column 2 (−0.753) (−0.738) (−0.540) −0.0358 −0.036 −0.047 DENSI proceeds a step further, estimating a second model (−0.517) (−0.515) (−0.675) −0.269 MGROW 0.0299 0.031 (0.011) (0.011) (−0.099) i (di − 1) / (ri − 2) /n, where n 13 The tie index is defined as 762∗∗∗ 762∗∗∗ 767∗∗∗ Likelihood di is the number of events at i and ri is the number of individuals ratio test at risk at i. The index takes a value of 0 for the case in which Number of 27,956 27,956 27,956 data are truly continuous and tends to 1 as the number of ties in observations the data increases (Brostr¨ m, 1998). o 14 The results of the Efron and Breslow approximations are fairly close. However, the likelihood ratio for the Andersen and Gill ∗∗∗ ∗∗ 5%, and ∗ 10% Coefficient statistically significant at 1%, model with Efron shows a higher value for all the estimated mod- levels; t-ratios in parentheses. els. Therefore, this approximation will be taken as a reference when commenting on the results obtained. Appendix 2 shows the results of the estimation of the Prentice and Gloeckler model in which the number of markets in which a sav- for comparison purposes. As mentioned earlier, the estimates ings bank operates is expected to have a linear are also very close in value and significance to those obtained influence on the timing of entry. Finally, column 3 through the maximization of the partial likelihood function. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  14. 14. 258 L. Fuentelsaz, J. Gomez and Y. Polo presents a last model which captures the hypothe- markets takes time. Note that the expansion seems sized quadratic effect (NMARK and NMARK2). to be quicker when the core markets are used as As has been found in other studies, firm-specific a ‘platform’ to entry than when diversification is characteristics are important when explaining entry undertaken from the newly entered markets. decisions and their timing. Our results show that The experience of the organization in managing firm size (SIZE, INTER) has a positive and operations across markets (NMARK) is also shown significant impact on the timing of new market to have a significant effect over the entry timing. entry in the savings bank market. This implies As evidenced by the estimates presented in column that, once the restrictions on the opening of new 3, the number of markets in which a savings bank branches in new markets were relaxed, the biggest is operating has an inverted U-shaped influence on entities (ceteris paribus) were the first to decide the speed of entry. That is to say, as firms proceed to look for new markets in which to develop a step further in the diversification process, they their activities. At the same time, this effect is accumulate knowledge about how to effectively stronger for those savings banks that have been manage simultaneous operations in several mar- characterized as net lenders in the money market. kets. Nevertheless, as the degree of diversification As has been mentioned earlier, the reason could be increases, this positive effect is compensated by an that the existence of slack resources provides these increase in coordination and monitoring costs. entities with an incentive to look for new markets In relation to industry characteristics, increasing in which to invest these at higher rates, also concentration (less rivalry) in the market in which confirming Hypothesis 1a. These results closely the firm was initially operating (COREH) has a agree with the fact that 59 of the 82 entries negative effect on the speed of entering into a (71.95%) that took place in the first 4 years under new province. Therefore, when the core market study were made by the three biggest savings and the survival of the firm are challenged by the banks in the country, and provide evidence to extent of rivalry, it tends to secure itself through reject the view that the largest savings banks are the extension of its activities to new markets (this rigid and inflexible. confirms Hypothesis 4 and closely agrees with The profitability variable (PROFI) is not statisti- evidence for the product case). Therefore, savings cally significant at the normal levels, thus rejecting banks that were operating in more concentrated Hypothesis 1b. Therefore, return on assets, as this markets have, ceteris paribus, been the slowest in has been defined, did not have any influence on becoming involved in the expansion process. the observed pattern of entry. Current (OBJEH) and potential rivalry Both the covariates measuring proximity (NUMSB) in the objective market have also been (PROXIC and PROXIE) show positive and considered as relevant when explaining entry deci- highly significant coefficients, implying a positive sions. The coefficients associated with these vari- effect on the timing of entry, as expected ables present a negative sign. However, only the by Hypothesis 2. Spanish savings banks have coefficient associated with the NUMSB covariate undertaken their diversification activities in a (potential rivalry in the objective market) presents sequential way, first learning about markets close a significant effect over entry timing. This con- to their original scope of operation, and then firms the idea that entry probability is increased directing their activities towards markets further in the eyes of potential entrants as the number away. As was postulated when formulating this of them itself increases. On the other hand, the hypothesis, it is expected that the capabilities probability of the incumbent engaging in activities of the firms are better suited for operations directed at blockading entry also increases as the in closer markets. Savings banks in adjacent number of potential entrants is larger (Hypothe- markets may be better positioned to act and sis 5b). However, the available evidence leads us obtain relevant information about the market. to reject Hypothesis 5a.15 Additionally, activities in closer markets are easier to control and proximity makes customers in the new markets more familiar with the entity. What 15 The insignificant coefficient of the variable could be due to the correlation between the two variables used to test Hypothesis 5 is more, the larger relative value of the coefficient (0.39). When the two variables were introduced in the analysis PROXIC in all the estimations agrees with the alone, both signs were found to be negative and significant at idea that the acquisition of knowledge about new the usual levels. Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)
  15. 15. Followers’ Entry Timing 259 Finally, demand potential does not appear to DISCUSSION AND CONCLUSIONS significantly affect the probability of observing an entry. None of the three variables (DEPHA, The available literature on strategic management DENSI, MGROW) used to test this effect were reflects the lack of studies devoted to analyz- found to be statistically significant at the normal ing the determinants of entry timing decisions. levels, thereby rejecting Hypotheses 6a and 6b. The purpose of this paper has been precisely to The interpretation of the coefficients of the cover this gap in the literature and to provide an model is analogous to the classical regression. For empirical extension through considering entry tim- example, for the case of the proximity to the core ing decisions from a geographical point of view. market variable, the coefficient of 2.411 means By analogy to research carried out in the product that when the market considered is adjacent to dimension, we have developed a set of hypothe- the market in which the firm was initially oper- ses that we have tested using the Andersen and ating, the log of the hazard is increased by 2.411. Gill (1982) formulation of the Cox Proportional A more intuitive interpretation may be obtained Hazards Model. through a consideration of the exponential value Our results have several implications in terms of of the coefficient (Allison, 1984). That is to say, the questions posed at the beginning of the paper. proximity to the market in which the firm was ini- First, as we have seen in the analysis, early and tially operating multiplies the hazard 11.145 (i.e., late entrants are clearly different. Early entrants are exp(2.411) = 11.145) times. For those cases in not endowed with the same type of resources and which the variable under study is continuous, the capabilities as late entrants. Early followers seem interpretation is, again, analogous. An additional to be characterized by being large and experienced potential entrant identified in an objective mar- firms, entering from closer markets. Therefore, ket (e.g., from seven to eight potential entrants) managers in incumbent firms should expect larger, reduces the probability of entering that market experienced, and closer organizations to be the first to 0.902 (exp(−0.103(8 − 7)) times. That is to to make entry effective. say, an additional potential entrant reduces the These results provide us with new evidence on the importance of differential resources in deter- probability of entering in that market by 9 per- cent (100(exp(−0.103) − 1)). Table 5 presents the mining entry timing. On the one hand, and as mentioned earlier, they agree with the evolution- effect over the hazard of a one standard deviation ary conceptions of the development of resources change of the covariates that were significant in within firms, and highlight the relevance of learn- the previous estimations. As with the cases just ing and experience as providers of key capabilities mentioned, a ‘large entity’ (defined as one with to secure successful growth. As we have seen, a value of mean plus one standard deviation in more experienced firms not only have a higher the LNSIZE covariate) has a 4.34 times higher likelihood of entry at any time, but also exhibit a probability of entering in the average market as pattern of branch expansion that seems to resemble compared to the average sized entity (that with the neighborhood effect of spatial diffusion stud- mean values in LNSIZE). ies, first progressing to close rather than remote locations, and then entering new markets as they become contiguous (Majahan and Peterson, 1985). Table 5. Effect of a change in one standard deviation of On the other, our results provide us with new the covariates evidence on the conflictive link between organi- Explanatory AG (1) AG (2) AG (3) zational size and organizational change, in that variables Efron Efron Efron we found large organizations were the first to enter and did not show any sign of bureaucratic LNSIZE 5.435 5.468 4.345 inertia. INTER 1.696 1.696 1.760 Second, the results highlight the importance of PROXIC 2.094 2.095 2.061 PROXIE 1.247 1.248 1.234 including firm-specific competitive conditions and NMARK 1.252 market characteristics in the study of entry tim- COREH 0.499 0.498 0.525 ing decisions. As mentioned before, if the like- NUMSB 0.680 0.677 0.675 lihood and the timing of entry are influenced by Copyright  2002 John Wiley & Sons, Ltd. Strat. Mgmt. J., 23: 245–264 (2002)