CORPORATE GROWTH, AGE AND OWNERSHIP STRUCTURE: EMPIRICAL EVIDENCE IN SPANISH FIRMS

The objective of this work is to analyse fi rm mobility among the different sectors of the Spanish economy according to a statistical classifi cation of economic activities at the 1-digit level. Some of the stylised facts that we fi nd are: an inverse relation between fi rm growth and age; an increase in new entrants’ average relative size in terms of sales compared to established fi rms among the different industries and cohorts; the importance of the fi rm’s initial size in entrepreneurial activity; the favourable impact of the economy on fi rm growth; and a positive relation between non-concentration in the ownership structure and greater mobility. In this context, an effi cient corporate governance system may prove as a signifi cant policy tool for the investment and growth prospective of the Spanish economy. The regulatory framework of the Spaniard capital market has been coordinate with the EU standards. The challenge is now mostly for the fi rms to adopt the appropriate corporate governance structures, in order to achieve real convergence, in terms of productivity and competitiveness, with other developed economies.


Introduction
The concept of corporate mobility refers to the process of entry into and exit from markets of fi rms and their units. This fl ow has been one of the most useful means of explaining the evolution of companies and their adaptation to their environment. All this has been the object of much attention from the theoretical perspective, but it has not had the equivalent empirical attention until recently, perhaps due in large part to the diffi culty of measuring it statistically as Baldwin, Geroski (1989). In particular, most research on fi rm turnover and the factors characterising it focuses on the industrial sector, where authors such as Dunne et al. (1988), Acs and Audretsch (1990), Baldwin
Market entry and exit of fi rms or units is an interesting way of observing the evolution and adaptation of these productive units to their environments. In this respect, the literature appears to indicate that although the theoretical perspective has been object of attention, the empirical aspect has been somewhat neglected, with analysis of fi rm creation and survival concentrating on manufacturing sectors.
In this context, the objective of this work is to analyze corporate mobility among the different sectors of the Spanish economy (according to their 1-digit CNAE codes 1 , comparing new entrants (ex novo) and established fi rms. Within this process of fi rm mobility analyzed in this work, we attempt to provide answers to the following questions: i) how do fi rms enter markets according to the different annual cohorts and considering the sector of activity? In this respect, we consider the size of the new entrants compared to the size of the established fi rms. This leads us to ask: ii) can we explain the fi rms' evolution after entering the market (post-entry behavior? If yes, do fi rms of different sectors, ages or growth rates behave similarly? iii) In terms of fi rms' evolution in their markets, do small or medium-sized fi rms grow faster than large ones? Is the age of the fi rm a determinant of its dynamics? From work such as that of Evans (1987) and Hall (1987) we observe the existence of a positive relation between a fi rm's size and its probability of survival.
There have been relatively few studies tackling these questions in general terms and in particular for the case of Spain until recent times. In consequence, we class this work among the group of novel analyses necessary to understand the corporate spirit in Spain. If there are not many empirical works studying topics relating to fi rm creation and consolidation, there are even fewer analyzing the behavior and trajectories of fi rms beyond the initial period of mortality of young fi rms.
Analy zing the questio ns posed i n this wo rk about t he beh avior of fi rm s enterin g or exiting markets or esta bli shed fi rms, as well as the factors that ch aracterize them, s hould help company manager s underst and not onl y ho w the res ourc es and cap abil ities in term s o f size an d acc umul ated exper ie nce evolve in a sector of acti vit y, but als o ho w they be st ad just under t he perspec tives of both new entrant s and est abli shed fi rms.

T heoretical literatures in corporate dynamism: mobility and transition
The di ffere nt theorie s on corporate mobilit y provide us with guideli nes in our attempt to answer the previous questions. For example, the theory of passive learning, with Jovanov ic (1982) as its strongest support er. The m ai n argument in this ap proach is that fi rms do not a priori know their own cost st ructures. If this proves to be competitive, the fi r ms will survive; if in contrast their costs exceed the ave rage of the established fi rms, they will end up exiting the mar ket.
From active learning theory 2 (Ericson and Pakes 1990;Hopenhayn 1992), fi rms can change their c haract eristics during their t ime in a ma rke t, consequ ently varying their c hances of surv iva l. The c au ses of thes e c hanges can be of v ari ous types: technol ogical, organizational, etc.
In co nt ras t with th e imp ortance that fi rm mobilit y has for exp lain ing market functio ning, there a re sti ll m any fi elds to expl ore . Dunne et al. (1988) indicate the lac k of stud ies analyzing pattern s of beha vior of fi rms enterin g or exit ing markets and the pos t-entry behavior (performance) of the new fi r ms in term s o f anal yzing the cha ract eristics/skills necessa ry for surviva l and gro wth. In tu rn , S choene cker and Cooper (1998) point out tha t in spit e o f the strategic interes t of the issue there h as bee n re marka bly little attenti on paid to the types o f fi rm that en ter markets and whe n th ey do so.
Accor di ng Bentzen et al. (2006), the extensive empirical literature on the validity of Gibrat's law does not in general verify the law as it fi nds that fi rms' growth rates are negatively correlated with both fi rm size and age. However, some studies fi nd that Gibrat's law holds for sub-samples of fi rms such as large fi rms or fi rms belonging to special industries. It has been pointed out that these results are due to the fact that the likelihood of fi rm survival for natural reasons is positively related to fi rm size and age. Whit a representative sample of Danish fi rms this study evaluates the validity of Gibrat's law for different kinds of fi rms over the period 1990 -2003. In contrast to the majority of earlier studies this analysis corrects for the bias in the estimations by using variables related to the survival of small fi rms.
Manjon and Arauzo (2008) fi nd that, in retrospect, the econometric specifi cations used in this area have progressively become more sophisticated, addressing issues such as discrete time, unobserved heterogeneity and competing risks. These authors identify a number of fi rm-and industry-specifi c covariates that provide largely consistent results across samples, countries and periods. According Manjon and Arauzo (2008) the evidence is less clear-cut with regard to ownership and spatial factors. De Jorge et al. (2010) have investigated the determinants of fi rm size. Data was collected in face-to-face structured-questionnaire interviews of 1314 fi rm founders from 167 14 counties in Argentina. The results show that the main sets of explanatory variables related to founder characteristics (age, experience, education, and vocation) provide a full explanation of fi rm size. It has also found evidence that a high degree strategic planning and a better competitive position are positively related to fi rm size as well.
In essence , most re cent work on fi rm mobilit y takes on one of following two perspectives accordi ng Sutton (1997): (1) Firms' chances of surv iva l dependi ng on thei r a ge, si ze an d oth er individ ual charact eristics; and (2) Fi rms' growth in func tio n of thei r a ge, si ze an d oth er i ndivid ual and sec toria l charact eristics. For e xample Arauzo and Segarra (2005) explore the determinants of fi rm start-up size of Spanish manufacturing industries. Their results indicate that the variables that characterize the structure of the market, the variables that are related to the behavior of the incumbent fi rms and the rate of growth of the industries generate different barriers depending on the initial size of the entrants. Arauzo and Segarra (2008) conclude that the industries' barriers to entry affect the ability of potential entrants to enter the markets and the size range at which they decide to enter. On the other hand, several studies have analyzed entry in developed capitalist economies coming to the conclusion that entrants are usually smaller, less productive and at higher hazard than incumbents. For example Rinadi (2008) considers if this was the case also in the rather peculiar situation of those fi rms which entered during the period of transition from planned to market economy, in one of the ex-soviet countries. Additionally Rinadi (2008) considers whether or not the uncertain environment generated by transition did activate a process of entry, as situations of uncertainty are generally supposed to do. The main result of this paper is that despite the fact that incumbents were fi rms created and organized to meet the objectives of the soviet regime, they were not outperformed by subsequently-created fi rms which were formed to match the needs of a transitional/quasi market economy. These results do not support "vintage" and "liability of obsolescence" models which suggest that new comers are better fi tted to match new conditions.

Data
As Vela sco (1998) points out, the fi r st c ertainty when trying to understand the reality of fi rm creation in Spai n i s of moving in a world of statistical uncerta inty. This situation also ap plies in other European countries, hindering any international comparison. Spain is not unaffected by this problem: while s ome databases allow a partial analysis of some interesting phenomena such as survival or creation by means o f repr esentative samples, these same databases do not allow study of the causes or factors behind business success or fail ure. That is , it is not possibl e to use a stati stical source to make a clear and dir ect analys is of fi rm creation. One of the problems is to determine whether the fi rm is a new creation, or whether it is simply a new operati onal unit set up by an existing fi rm. In spite of the possibilities offered by this database to study the causes of fi rms' exit from markets, entry rates, etc., the current work uses the SABI database for its analysis. This database collects data on more than 180.000 fi rms (population) inscribed in the Mercantile Register (BORME), covering all sectors of business activity in Spain. One of the competitive advantages of this database is that it allows researchers to use variables relating to fi rm management.
The database we use here holds data on the main Spanish fi rms. It is highly representative of fi rms from the 18 S pani sh autonomous "communities" (i.e., r egions) that p resent their ac counts in the Merc antile Registers. From t he tota l po pulation of more than 180,000 fi rms, we have taken ra ndom s amples, as describe d in each of the f ollowing sections, on the basis of variables chosen in function of the objectives of the research. The un it of analy sis is the n ewly-created fi rm for the case of the new entrants in mark ets.
The statistics of Table 1 (see A ppendix 1) show some relevant data. Th e ag e of the fi rms considered in this panel is on average 13.6 yea rs, w hich implies that the fi rm s are in general relatively young, with some exceptio ns 3 . The a verage size, me asured by number of employees, ranges from 25 emplo yee s in 1996 to 37 em ploy ees in 2001, wh ich in dicates that the samp le has a signifi cant number of small and medium-sized fi rms (SMEs) 4 . This i mplies that the sample closely approximates a real market s tructur e, although logicall y we have als o con sider ed large fi rms within t he samp le. Anothe r of the b ig data base s refers t o the a nal ysis of the n ew entr ants in relat ion to the s ect ors of activ ity (see Ap pendix 2, 3). 3 The age of the SMEs, according to the European Business Study (Maroto 2001), is based on a sample of fi rms that are generally more than 15 years old, among which however the fi rms of Portugal, Spain, France and Greece appear to be younger . The statistical sources used from the SABI database, which holds information about fi rms from the BORME (Offi cial Gazette of the Mercantile Register), has introduced biases into the analysis. For example, the minimum level of turnover of the fi rms observed and retained in the database is set at €479.041. However, this limitation has certainly turned into an advantage when analyzing the fi rms and their evolution over time, as long as they survive beyond the initial stage of approximately three years of life. 4 In the above-mentioned European Business Study there appears to be a relation between on the one hand the variables average fi rm size and dominant size class of the fi rms in each country, and on the other the competitive position occupied by the countries at the international level . In the most competitive countries we fi nd Finland, the Netherlands, Sweden, Ireland and Denmark, with a total of 1,280,000 fi rms, the average size of fi rms ranges from 5 to 12 employees per fi rm ( 20-40 companies per 1,000 inhabitants ), the predominant structure is the large fi rm, and SMEs provide from 60 to 70% of total employment . In contrast, in less competitive countries such as Portugal, Italy, Greece, Spain and France, with a total of 10.085.000 fi rms , the average size is smaller, between 3 and 7 employees per fi rm ( 60-70 fi rms per 1.000 inhabitants ), the predominant structure is the micro fi rm, and SMEs generate more than 80% of total employment .

Empirical analysis and results
According Dunne et al. (1988) the importan ce of fi rm entry an d exit as deter min ants of marke t c haracteristics is widel y r ecognized. In this sec tion we carry ou t an e mpir ica l analysis to respo nd to the q ues tion s posed in the i ntr oduc tion 5 . In this respect, this wor k makes va rious contribu tions to the e mpi rica l literature on fi rm growth.
In the next sections, we exami ne the rela tion between growth a nd age. This r el ation is impor tan t because some the ories of fi rm growth p redict particul ar patterns of growt h d epending on the stage in the fi rm 's life cyc le. W e fi n d tha t fi r m growth d eclines with age . Thi s i nv erse relation between growth a nd age is c onsi ste nt with Jov anovi c's (1982) theory of fi rm growth, in which fi rms discover their tr ue effi cien cies over tim e in a pro ces s of Bayes ian learning .
Also, w e exami ne the rela tion between fi rm growth a nd age for vari ous types of fi rm, consider ing the char acte ristics of the sector o f activ ity . We fi n d tha t gro wth d eclines with the size of the fi rm for rele vant samples. This r es ult i s equal ly importan t, because some the ories (Simon a nd Bonini 1958;Lucas 1978) and special cases of Jovan ovi c (1982), among others, assume o r sugge st that fi r m growth i s indep end ent of size, as postu lated by Gibra t's law. It is pr eci sel y the vari able fi rm age that has serv ed a s a sup por t in our a tte mpt in this res earch to explain the effe cts of growt h. In thi s respect, Evans (1 987) indicates that fro m the theo reti cal point of view stu dies designed to incorpor ate age can be expected to make an important contribution to the lite rature. This a ut hor a lso rec ommen ds caution in the use of Gibrat's law 6 to expla in the dist ribu tion of fi rms by size , as does Luc as (1 978). Author s such as Evans (1 989), in the line of resea rch of empir ica l work, sug gest the impo rtan ce of age a s a det erm in ant factor o f dynam ic industri es.

Characteristics of new entrants
Understanding what happens to fi rms after entering a market is an issue of some interest, given that the effects of fi rm mobility on the sectorial structure depend not only on the number of fi rms entering or exiting the market at any given time, but also on their evolution in the market where they operate. It is particularly important to understand the rate at which fi rms disappear and how they gain market share. The scarcity of studies to the present day refl ects the diffi culties new entrants have surviving in markets, since they are generally small (Geroski 1991).
What patterns of growth do Spanish fi rms entering markets display in the different cohorts? If we can identify a particular pattern, do all the cohorts of fi rms behave similarly in their growth?
The theories explaining fi rm entry into markets are conceived from a static or dynamic perspective. The fi rst type establishes a direct relation between the new entrants and sectorial barriers. In this respect, the entry rate is positively associated with fi rms' expectations of potential profi ts and negatively associated with the profi ts sustainable in the long term, which are in turn related with sectorial characteristics. Geroski (1991) proposes that the entry rate of fi rms in a sector is related to the expected profi ts and the sectorial variables generating barriers to entry, and that these are infl uenced by the speed of response of the fi rms. On the other hand, the dynamic approaches explain fi rm mobility in terms of innovation-imitation processes, asymmetries in the expectations and the generation of economies of learning.
As can be seen in Fig. 1, the different cohorts evolve similarly in their sales 7 aggregating the sectors.
Thus the growth of the new entrants in the sample appears a priori to indicate interesting expectations for the future. This year-by-year evolution of the new entrants shows greater rates of evolution. Is the growth similar for the different cohorts (post-entry behavior)? How are the starting size and the post-entry evolution related?
Taking as basis Nelson, Winter's (1982) model andEvans (1987) in relation to the growth rate (dependent variable), which will be developed in the study of the established fi rms' growth, we propose in equation (1) the following explanatory specifi cations of the model of new entrants: where:s represents the net sales turnover in thousands of euros; t* is the fi nal year of the fi rms under analysis, corresponding to 2000; t i is the entry year of the fi rms, between 1994 and 1999; fi nally, t* -t i is the difference between the fi nal year and the initial year for each cohort. 7 In general fi rms' sales volume (in thousands of euros) has been the most used variable in this current work, both to measure size and growth. The reason for this lies in the fact that this variable is the most representative in the SABI database. In particular, it achieves 40% more year-observations than the level of employment. On the other hand, in much of the literature on entrepreneurs, as Autio et al. (2000) point out, "and even the growth in sales has allowed us to distinguish what is and what is not entrepreneurial activity". The growth is analyzed for newly-created fi rms (ex novo). We cannot distinguish any merger processes that may have occurred, and hence neither can we determine by which means the growth was achieved, whether internally or externally 8 .

Fig. 1. Evolution of average sales values of cohorts
Starting from Equation (1), the following regression models the growth of the new entrants by cohorts: where: S ti measures the sales of the year of entry into the market in the cohort being considered; S ti 2 represents the term of the quadratic evolution of the sales; and Sec is a dummy sector variable (10 sectors, according to the 1-digit CNAE code). Although this was the fi nal model chosen, in Table 1 we compare the results achieved with different alternative models, considering cumulative growth rates or not, and including logarithms in the independent variables or not. In this respect, the models incorporating logarithms are statistically signifi cant in all the variables considered, and in particular the quadratic form obtains better goodness of fi t coeffi cients for all cohorts.
The results of estimation 2 are shown in Table 2. As can be seen, the new entrants in the cohorts 1996, 1997, 1998 and especially 1999 experience higher growth rates with greater size. This can also be observed in the descriptive analysis of the distribution of the growth (see Appendix 3). In terms of elasticities, when there is an increase in the entry size of fi rms of 1% the growth evolves particularly from the cohorts of the year 1996 onwards. This growth is particularly important for cohorts 1998 and 1999. Year

Note: S = sales
Source: Author's calculation. The models are shown graphically in Fig. 2 . The minimum values of the curves, as well as the average sizes and the 95% percentiles, indicate that the relevant area of analysis is to the left of the curve s. This is where the growth rate is inversely related to the size . Over time, new entrants raise the average size of their units. Audretsch and Mahmood (1995) argue that this growth in size occurs for two reasons : i) the exit from the market of fi rms belonging to the cohorts, generally small companies; and ii) the growth of the fi rms remaining in the market .
In addition to the study of the behavior of the new entrants, we undertake a descriptive analysis of the fi rms' productivity and how this relates with the growth of the economy .
The year fi rms decide to enter in a market is probably associated with a better economic situation in Spain. If this is so, fi rms entering the market in times of economic expansion benefi t, enjoying higher growth rates than fi rms that enter the market when the economic situation is not so favorable.
As Segarra et al. (2002) points out for the manufacturing sector, the net entry rate of fi rms may be related to the economic cycle, with a positive correlation between fi rm  creation and expansionary cycles and a negative correlation during recessions . In turn, Boeri and Bellman (1995), in a study of the German manufacturing sector, fi nd no evidence that the economic cycle infl uences the exit of fi rms, at the same time as a weak sensitivity of exits to growth in terms of the number of employees in established fi rms. In this sense, when we study growth in the different cohorts of post-entry behavior the fastest growth is observed in new entrants from 1996 onwards -the point when the Spanish economy begins to enter an expansionary cycle . On the other hand, in Fig. 3 we can see the evolution in productivity (in terms of sales/number of employees ) of the new entrants by cohorts. Although in general all the new entrants in their respective cohorts show gains in productivity, it is the cohorts of the years 1996-1999 that follow the growth in the economy pro-cyclically, above all compared to the cohorts of 1994 and 1995 .
Disaggregating by sector, Fig. 4 presents the evolutions in productivity of the new entrants of each cohort in the different sectors of activity . There are generally improvements in productivity in all the sectors for all the cohorts . In particular, we can see some characteristic features. The productivity of th e new entrants is higher in the fi rst year in Sector 5 (sales, commerce, etc.) for all the cohorts, followed, by sectors 1 (food, drinks, etc. ) and 6 ( transport, post, etc.) . In Sector 5 there are signifi cant gains in average growth for all the cohorts, and the same is true for sectors 3 ( offi ce machines, electrical material, etc. ) and 4 (construction, energy, etc.), while in sectors 1, 7, 8 and 9 this evolution in growth is most marked from the year 1996 onwards. Sectors 2 (wood and cork industry, chemicals, etc.) and 6 present the profi les of least evolution in productivity .
There are two empirical facts that tend to disconcert economists when they analyze processes of market entry and corporate turnover . The fi rst concerns the a symmetric distribution of fi rms in terms of size, given that there is a clear predominance of smallersized fi rms . This could be suggesting a priori that a large number of fi rms are producing below a minimum effi cient level (Sutton 1997). The second fact is that the entry of fi rms is high even in those sectors where the economies of scale are important, which might suggest that in these sectors this phenomenon does not discourage the entry of new fi rms . Journal of Business Economics and Management, 2011, 12(1): 164-196

Entrants' size compared to established fi rms' size, by sector of activity
The statistical data analyzed in the previous section do not provide information about the diversity of fi rms entering the market in relation to the size of the fi rms already established in the sector. In Table 3 we report descriptive statistics about the new entrants by 1-digit sector, as well as the relative size of the new entrants as a proportion of the size of the established fi rms, under a longitudinal perspective ( 1994 to 1999) . Source: Author's calculation.

End of Table 3
Some of the questions that we pose in this section are : are the new entrants smaller and does this persist over time, or when they enter the market do they already have a substantial average size at the sector level? If they do remain small, does this occur in all sectors or does it differ in function of the structural characteristics of the sector, or the way the fi rms compete? If in contrast they enter the market with a particular size, which then modifi es, how long does this adjustment process take, and is it similar in all the sectors?
In the introduction of this work we discussed the criteria referring to the new entrants' expectations about the profi ts they are likely to obtain and the obstacles or barriers to entry that they are likely to fi nd . Providing answers to these questions in this section relates with aspects that the literature has in some cases already tackled: the new entrants' capacity of adjustment of their cost structures to the characteristics of the markets. The heterogeneity of the fi rms, their learning processes from their entry onwards, which uncover asymmetries in their effi ciency levels, and the differences between organizations in their development of the skills of imitation and learning or the incorporation of more effi cient capital goods, are some of the arguments that will prove useful in the analysis.
The fi ndings reported in Table 3 allow us to point out some stylized facts: (a) As in Dunne, Roberts, Samuelson (1988), the average relative size of the new entrants as a proportion of the established fi rms grows in all sectors and all cohorts. For example, the size of the new entrants in Sector 5 (mid table) is 13.7% that of the established fi rms in the 1994 year of entry. In 1995 this proportion rises to 29.3% and it continues to grow until it reaches 35.7%, (b) The pattern of evolution in the growth of the new entrants varies in function of the sector of activity. While the level of relative size reaches 36.87% in Sector 5, in other sectors such as Sector 1 it reaches 124.3% for the 1998 cohort. (c) In general, new entrants' processes of adjusting their size with respect to the established fi rms take longer than six years. Consequently, in 1999 with very few exceptions the relative size of the fi rms in proportion to the established fi rms does not exceed 50%. Geroski (1995) indicates that new entrants are small and that they take over a decade to achieve sizes comparable to the established fi rms. (d) The entry size varies in function of the sector of activity. For example, in sectors 2, 4, 7 and 9, observing the entrants of each cohort and comparing them with the sectors 1, 3 and 8. On the other hand, if survival is related to size this latter may not be acquired immediately. Some authors Audretsch (1991); Mata and Portugal, (1994); Wagner (1994); among others provide evidence of the greater variability in survival rates among different sectors than among new entrants in the same sector.

Growth of established fi rms: size and age
The theories on the relation between age and growth in fi rms are closely related with those that link size and growth, due to the demonstrated relation between age and size. However, the available evidence does not categorically confi rm that new fi rms -which are generally smaller than established fi rms -grow more rapidly. The theories of corporate growth are basically of two types. On the one hand, the stochastic theories are based on the marked asymmetry of distribution of fi rm sizes that is observed, with less importance been lent to technological or demand aspects, considering that the evolution of fi rm size is infl uenced by a large number of explanatory factors that should be treated as random variables.
The determinist theory, on the other hand, is based on the neo-classical model and holds that growth is closely linked to the idea of optimal size. According to this approach, fi rms have the objective of carrying out a process of adjustment to achieve this more or less rapidly. The main result from the analysis is that fi rms wish to reach their optimal sizes as quickly as possible, but that there are costs of adjustment that prevent them from achieving this immediately. This implies that in sectors in which the fi rms have curves of average long-term costs that are U-shaped or similar there will be an inverse relation between size and growth, since large fi rms tend to have less need to grow in size than small ones, as the costs derived from having an ineffi cient size decline the closer the fi rm is to its optimal size.
Under this perspective of the corporate growth process -which is the perspective of this section, as we shall see later -the diversity of sizes observed in the market is simply a temporary situation caused by the fact that the fi rms are all at different stages of the process of adjustment towards their optimal size. Fig. 5 suggests that the variability in growth of the sample fi rms observed between 1996 and 2001 is related to size and age. Smaller and younger fi rms tend to exhibit greater variability in their growth, considering the rate of cumulative growth (vertical axis). As the fi rms age (age groups at the top) and grow in size (Lnsales96), the variability in their growth declines. Thus, we see that in the early stages of life smaller fi rms tend to grow faster, which is in line with neo-classical growth models.
This fi nding -that large fi rms grow more slowly than small fi rms -is consistent with work carried out by Kumar (1985), Evans (1987), Acs and Audrestsch (1990), Dunne and Hughes (1994).
As we mentioned earlier, new entrants are generally smaller than established fi rms and consequently try to grow as quickly as possible in order to compensate for their size disadvantages. However, the existence of obstacles to investment, which are particularly intense in this type of fi rm (perhaps fi nancial investment being the most worrying), means that many of them cannot achieve this. The existence of greater asymmetries in their access to investment among small fi rms than among large ones appears to be the cause of the greater variability in growth. Table 4 reports the same results in a descriptive analysis. As fi rms pass from one age range to the next, their rate of growth declines while their average size increases. Our analysis demonstrates an inverse causal relation between growth and age, coinciding with Hart (1962), Mansfi eld (1962), Hall (1986), Evans (1987), Dunne and Hughes (1994). This relation is important because some theories of corporate growth predict particular patterns of growth in function of the life cycle of the fi rm.

Growth of established fi rms, age, size and sectorial characteristics
Nelson and Winter (1982) study the circumstances under which fi rms experience initial growth in relation to their size and the subsequent decline in growth. Later and Evans (1987) develops a growth model to determine the relations between these variables in the manufacturing sector. We shall apply this model to the database of established fi rms between 1994 and 2001. The model is as follows: where: t represents the period considered, t ′ > t, d = t ′ -t and e is the error term with lognormal distribution and with possibility of non-constant variance. Equation (3) suggests the following regression to estimate growth: where: μ t is a normal distribution with mean zero and possibility of non-constant variance and independent of age and sales. According to Evans (1987), taking the secondorder expansion of Ln G (Age, Sales) we obtain:  (5) The sample of fi rms that has been used for the analysis has considered the age of fi rms over three years. The literature on fi rm creation and survival holds that if companies 180 survive beyond their fi rst three years of life -the peak period of organizational mortality -their chances of survival improve considerably. Table 5 shows the results obtained. The behavior of the variables explaining growthi.e., age and size -presents a quadratic form as in Evans's (1987) model (in this case with sales rather than employment to capture the size effect). The signs of the coefficients and their statistical representativeness show that the relation between growth and size is U-shaped, independently of the age range considered, while when we consider the age of the fi rms in their respective ranges some differences are observed. The relation is U-shaped for the youngest and oldest fi rms, being inverted U-shaped for the fi rms of intermediate age (10-20 years). This behavior may be related to the life cycle of the fi rm according to a logistic trajectory. In the stages of birth and development fi rms grow when they are young, evolving in size and age. In the maturity stage the fi rms continue to grow in size and the age of the fi rms presents a convex form, until they reach the stage of full maturity, when the evolution in the size and age is similar to the initial stage, although the variation in growth is at approximately 58% of that in the earlier stages (1.21 compared to 2.08).
The positive coeffi cient of the variable age*sales indicates that the effect of the initial sales on growth is stronger the older the fi rm, and also that the effect of the age on growth is greater the higher the initial sales. This might suggest that the initial size of the fi rm, or its speed of adjustment, play an important role in its growth. Firms' greatest risk of failure and hence of abandoning the sector is associated with the smallest sizes. This implies that fi rms deciding to initiate their activity with sizes that are smaller than the effi cient level and that then attempt to achieve the optimal size by means of the necessary process of adjustment may be at a signifi cant initial competitive disadvantage.
In the case of the oldest range of fi rms this effect is inverted.

The inter-sectorial mobility of fi rms
In previous sections we an alyzed the relations between fi rm size, growth and a ge, depending on the sector of ac tivity. We investigated whether large fi rms grow more slowly than small or medium-sized fi rms, or whether conversely they grow more quickly. In this section we shall study the differentiating characteristics of the established fi rms, by sector of ac tivity, to subsequently determine the inter-sectorial mobility of fi rms.
We defi ne fi rms to be above the average of their sector when they are larger than those that are below the average, when we use both sales and employment level as measure of fi rm size. Firms that are above the average are also older -by more than 13 ye arsbeing 20 ye ars old on a verage.
On the other hand, with the data analyzed, fi rms with higher than average values are more productive (sale s/number of employees) and profi table. But their fi nancial profi tability is not so favorable, perhaps as a consequence of the higher debt levels, greater fi xed assets and higher relative labor costs of medium-sized and large fi rms.
The percentage of fi rms above the average is consequently small. Con sidering the averages of the year 1996, some 12%, of the fi rms are above the average, the minimum value is found in sectors 4 and 7 (construction and fi nancial intermediation), with only 6.2% of the fi rms above the average, and the maximum in Sector 8, with 27.6% . But it is true that in this latter case the s ize of the sample is small (94 fi rms), hence the average will be sensitive to this.
In some sectors we fi nd differences between the averages of the initial (1996 ) and fi nal (2001) yea rs considered, particularly in the v ariable sales. We recall that unfortunately we have not been able to capture processes of mergers and spin-offs that may have occurred at the sector level.
We carry out the same descriptive analysis with the median as the frontier or limit, to determine which fi rms exceed or have possibilities of exceeding the median. The conclusions drawn from the a nalysis of the mean are also valid when the frontier is the median. The differences between fi rms exceeding the median and t hose below it are that the former are larger in terms of sales, employment, produ ctivity and economic profi tability. Alt hough in this case of course the median divides the group of fi rms in two.
In a fi rst approximation Sim ilarly at the other extreme of the diagonal 22.3% of th e fi rms that were in the f ourt h quartile in 1996 remained in that quart ile four years later. Tra nsitions below the d iago nal indicate downw ard movem ents from quartiles (demo tions), w hile above the diagonal they repre sent promotions to higher quartiles.
There are fewer demotions than promotions, with the norm being promotions of levels.
For example, a total of 8.18%, 3.36% and 1.50% of the fi rms belonging to the fi rst quartile of 1996 promote to the second, third and fourth quartiles, respectively.   Of the fi rms belonging to the third quartile in 1996, a total of 2.72% were demoted to the second quartile of 2001, and 8.88% promoted to the fourth quartile of that year. In Table 7 we show the transitions disaggregated for some sectors of activity.
Sectors 2 and 3 on the left of Table 7 show similar transitions (their behavior is similar to that of the sectors that have been omitted: 1, 7 and 9). Sectors 4 and 5 on the right of the table show more dynamism (this behavior is similar to the omitted sectors 6 and 8). In particular, Sector 5 exhibits a relatively lower permanence of its fi rms in their quartiles (diagonal), which varies with respect to the rest by some 2% approximately and a greater number of transitions upwards than the rest of the sectors, particularly promotions to the highest quartile. On the other hand, the dynamism of this sector is also observed when we examine the demotions.
Having analyzed the transition of fi rms in the period of time under analysis and their dynamism, considering the sector of activity, we might ask what is the probability that a fi rm exceeds the frontier in terms of the quartiles of the sector of activity where it operates? And, to what extent does growth affect its mobility over time? Finding a response to the fi rst question could provide some evidence about the causes of fi rm survival. The answer to the second meanwhile may be more related to the question of whether growth really explains survival, and if so, how it is related with the rate at which the fi rm achieves an effi cient size to be competitive, or its possibility of catching up if it does not have adequate growth.
This approach to determine the probability that the fi rm exceeds the frontiers (in terms of sales) or not and the temporal effect can be seen in Fig. 6.
In order to analyze the probability that a fi rm will promote from its quartile both in year t o (1996) of the sample and 2001 -movements 1 and 1* of Fig. 6 -as well as to analyze the probability of improving its quartile to a higher quartile fi ve years later (position 2), we use probit models. In the fi rst case, we use an ordered probit, while in the second, where the transitions are linked to promotion or growth; we use a binary selection model. The starting equation is as follows: where: Prob is a variable taking four values in function of the quartile in which the fi rm fi nds itself; age refers to the age of the fi rm and age (square) the quadratic component of the age; communities is a dummy variable taking 18 values, according to the fi rm's autonomous community (i.e., region) of origin; Sector is a dummy variable taking 10 values in function of the fi rm's sector, according to NACE. The results are presented in Table 8.
The results considering the marginal effects are shown in Table 9.    (17) Sector ( In 1996 the probability of transitions between lower levels declines with age, while it increases in the superior levels. For example, when the age varies by 10% the probability of moving in the low levels declines by 0.46% and 0.16% in quartiles 1 and 2 and increases by 0.14% and 0.49% in quartiles 3 and 4. This effect is reversed when we consider the year 2001, when the probability of moving in the lower quartiles increases and in the higher quartiles declines. These facts may be related with the growth in the Spanish economy. In 1996 the expectations were favorable and an expansionary cycle was beginning, hence the post-entry growth of the fi rms in their sectors benefi ted from this situation. The larger the initial size of the fi rms, the greater the effect of the age, and the higher the initial sales (as we have already said), the smaller the fi rms growing in the lower quartiles. In the higher quartiles it is the medium-sized and large fi rms that experience growth. On the other hand, in 2001 the effects could be the reverse, since the expectations of growth diminish, the probability of transitions is related to an initial minimum size, and the survival of small fi rms becomes diffi cult. Meanwhile, in the higher quartiles it is the smaller fi rms that can move, probably to the extent that they have greater fl exibility and can adapt their size to market needs.
In Table 10 we report the results of the analysis of transitions of fi rms from their quartiles of 1996 to 2001 -Model 1. In Model 2 we consider the effect of the concentration of shares in the hands of the main shareholder. In this respect, we consider it relevant to examine the relation between the governance of the fi rm -measured by the control exercised by the majority shareholders -and growth. Models 3 and 4 capture the demotions produced during the same period.
Promotions from any quartile of 1996 to a higher one in 2001 are positively related with fi rm age. Moreover, the negative and statistically signifi cant sign of the dummy variable measuring shares in the hands of the main shareholder indicate that the probability of promoting is related to fi rms with a non-concentrated ownership structure. Non-concentration of the ownership fosters higher growth than when the ownership is concentrated. Work such as Zahra (1996) and Zahra et al. (2000) analyses fi rms' entry into national and international markets in relation to the ownership structure. Some of these authors' fi ndings show that the effects of ownership and governance can vary from one fi rm to another depending on their size. Marseguerra (1998) points to the importance of considering share concentration as a mechanism of management control. When there is a certain level of concentration of shares in the hands of one shareholder, this investor will have suffi cient incentive to break with their rational apathy and control the operation of the fi rm. In this sense, two conditions should coincide (Pinillos 2001) to consider the concentrated ownership as a monitoring mechanism of the management: i) that there really is a high degree of concentration of the ownership of the fi rm, to allow for an active control function to be exercised; and ii) that the shareholders are guided by performance and the return on their investments.
The absence of shares in the hands of the managers can cause opportunistic behavior, with the managers supporting projects that increase their own personal wealth and favor and ensure their job security. When the objectives of the managers and shareholders are closely aligned embarking on new activities both creates value and pursues the managers' objectives. Berle and Means (1932) point out that a concentrated ownership of a fi rm has signifi cant implications for the development of corporate strategy. Diversifi cation can imply confl ict of interests between managers and shareholders in situations where the diversifi cation only means maximizing manager wealth.
It is important to consider that agency theory warns of a negative relation between ownership concentration and strategic diversifi cation. In this respect, the shareholders' active control will favor the convergence of the managers' utility functions and the shareholders' interests. On the other hand, when control and ownership are separated, and the managers' interests are consequently directed at promotion, status, etc., expectations of company growth may improve.
With regards demotions, the probability of transitions to lower levels/quartiles diminishes with the age of the fi rm, and in this case the effect of a fi rm's ownership concentration is not signifi cant. Firms that have not promoted from their quartile in the year considered (t o ), probably because of not having suffi cient age -i.e., not having achieved suffi cient growth rate -do not catch up and the probability of exceeding the frontier increases with age (comparing 1996 with 2001), although to a decreasing extent.

Conclusions
Although we have not been able to work with data as representative as those provided by DIRCE or other sources, we have enriched the analysis by incorporating fi rm variables at the individual level (s ales, profi tability, ownership structure, etc.). This has allowed us to understand important aspects about the running of the fi rms, which have in some cases directly or indirectly suggested important facts regarding the heterogeneity of the fi rms.
The growth of the new entrants in terms of turnover has similar patterns of convergence (convex form in the relation of growth and size). The effects of economic growth are refl ected in the evolution of companies' size. While the entrants among the cohorts of 1994 and 1995 evolve similarly, in the cohorts from 1997 onwards the evolution is much more intense, with the growth increasing considerably year by year until 1999. This period corresponds to an expansionary phase in the economy. When we analyze the productivity -in terms of sales over number of employees -a similar effect is observed. This fact could suggest that although all new entrants into a market are affected by the growth in the Spanish economy (GDP), those belonging to cohorts from 1997-1999 benefi t particularly from it.
When we analyze the growth of the new entrants in relation to the established fi rms, the following common characteristics are found: (1) Similarly to the fi ndings of other authors (Dunne et al. 1988), new entrants' average relative size (in terms of sales) as a proportion of the established fi rms increases, in all the different industries and cohorts. For example, the new entrants in Sector 5 (commerce) have 13.7% of the size of the established fi rms. In 1995 this proportion rises to 29.3% and it continues to grow in the following years until it reaches 35.7%.
(2) The pattern of evolution of new entrants' growth varies among the different sectors of activity. While the average size level reaches 36.87% in Sector 5, in other sectors such as Sector 1 (food, drink and tobacco), it reaches 124%. (3) In general the processes of size adjustment of the new entrants with respect to the established fi rms take over fi ve years. Geroski (1995) indicates that new entrants are small in size and that these fi rms take more than a decade to achieve sizes comparable to the established fi rms. (4) The entry size and hence the level of resources a fi rm has at its disposal to be in a position to compete, is a function of its sector of activity. For example, sectors 2, 4, 7 and 9, observing the new entrants of each cohort and comparing them with sectors 1, 3 and 8. On the other hand, if survival is related to size, this latter may not be acquired immediately. In some works, such as Audretsch (1991), Mata and Portugal (1994) and Wagner (1994), among others, some evidence is provided of a greater variability in survival rates between different sectors than among entrants of the same sector.
When we analyze the characteristics of the new entrants and established fi rms, the following characteristics are found: (1) From a panel of data from 1996 to 20 01 we observe that the variance of fi rm growth observed is related to the size. The smallest fi rms have greater variability of growth than the larger ones. The result obtained in this work with regards the fact that large fi rms grow more slowly than small fi rms is consistent with other studies carried out by Ku mar (1985), Evans (1987), Acs and Audrestsch (1990), Dunne and Hughes (1994). (2) On the other hand, we fi nd that growth declines with age, as some authors have found (H art 1962;Mansfi eld 1962;Hall 1986;Evans 1987;Dunne and Hughes 1994). This causal relation is important, because some theories of corporate growth predict particular patterns of growth depending on the stage in the life cycle of the fi rm. The current analysis confi rms the inverse relation between growth and age. (3) The growth observed in the established fi rms -i.e., fi rms with more than th ree years of activity in the market -has a similar behavior to the life cycle of the fi rm and confi rms the results obtained by Ev ans (1987). In particular this analysis has been carried out using th ree samples of different ages. For the fi rst sample, where the group of ages ranges from 4 to 9 years, the growth relates to the evolution in sales and the age in a U shape. For the range of ages between 10 and 20 years th e sales continue to have the same form, but the age changes the trajectory to an inverted U shape. Finally, for the fi rms older than 20 years, the behavior of age and size are the same as for the youngest group, although th e variability of the growth declines (5 8 % of that of the fi rst group). (4) On the other hand, the effect of age on growth is stronger the higher the initial sales. This could suggest that the initial size of the fi rm, or its rate of adjustment, play an important role in growth. Firms' greatest risk of failure and hence of abandoning the sector is associated with smaller size. This implies that fi rms that decide to initiate their activity with sizes that are smaller than the effi cient level, and that aim to achieve the optimal size by means of the necessary learning process, may start out with a substantial competitive disadvantage.
Finally, when we analyze fi rms' probability of transition in function of th e quartile (in terms of sales volume) to which they belong in the years 19 96 an d 20 01, we fi nd that in 1996 the probability of promotion declines with age among the lower levels (fi rst and se cond quartiles), while it increases among the higher levels (th ird and fourth quartiles). This effect is inverted when we consider the year 20 01, when the probability of transition increases with age in the lower levels and decreases in the higher levels. These facts may be related with the growth in th e Spanish economy. In 19 96 th e expectations were favorable and an expansionary cycle was beginning, an d hence th e post-entry growth of fi rms in their sectors benefi ted from this situation. The larger the initial size of the fi rms, the stronger the age effect, and the higher the initial sales, th e smaller the fi rms growing in the lower quartiles. In the higher quartiles it is the medium-sized an d large fi rms that experience growth. On the other hand, in 20 01 the effects may be inverted, si nce th e expectations of growth diminish, the probability of transition is related with a minimum initial size, and the small fi rms fi nd it diffi cult to survive. In the higher quartiles it is no w the smaller fi rms that can move qu artiles, probably to the extent to which they are mo re fl exible and can adapt their size to market needs.
In the tra nsition of fi rms from quartiles of 1996 to 20 01 we have considered th e promotion of fi rms to a higher quartile, their demotion to a lower one and the effect of the concentration of shares in the hands of the main shareholder. In this respect, we consider it re levant to examine the relation between the governance of the fi rm -measured by th e majority shareholders' exercise of control -and grow th. Promotions from any quartile in 19 96 to a higher one in 20 01 are positively related with fi rm age. Moreover, th e probability of promotion is also associated with fi rms where the ownership structure is not concentrated.
To conclude this work, we propose some policy recommendations: First, an effi cient corporate governance system may prove as a signifi cant policy tool for the investment and growth prospective of the Spanish economy. Second, knowing that regulatory framework of the Spaniard capital market has been coordinate with the EU standards, the challenge is now mostly for the fi rms to adopt the appropriate corporate governance structures, in order to achieve real convergence, in terms of productivity and competitiveness, with other developed economies.  Journal of Business Economics and Management, 2011, 12(1): 164-196