- Email: [email protected]
Staying in business
inte~ational
Journal of ~~d~stria~ ~rga~~~at~o~ 9 (~99%~545-556. worth-
Final version received December 1990
The paper analyses factors which help a small, entrepreneu~al firm to stay in busines. The time-scale for the analysis is three years (1985-1988). The main issue that the paper is concerned with is the role of market and financial variables. Primary source data are used. An estimated probit model indicates that, for this sample, product range and gearing are the key ma%setand financial variables, respectively. The greater the product range, and the lower the gearing, ceteris paribus, the better the chances of staying in business.
This paper is concerned with a question that has been central to recent industrial policy [See Storey et al. (1987a)J namely what factors help a small, entrepreneurial firm (SEF) stay in business. The time scale for the survival analysis is three years (1985-1988). The main issue that the paper is concerned with is the role of market and tinancial factors. Primary source data are used. An estimated probit model indicates that, for this sample, product range and gearing are the key market and financial variables, respectively. The greater the product range, and the lower the gearing, ceteris paribus, the better the chances of staying in business. *This paper could not have been written without a grant from the Nuffield Foundation which enabled me, where possible, to trace and interview in 1988 those firms first interviewed in 1985. Ann Theresa Lawrie provided useful research assistance on the field work, as did Jacqueline Campbell, who also updated the small firm data base and disp%ayed great diligence and persistence in obtaining the fullest information possible for a few elusive cases until the end of 1988. Chris Corrie assisted on computing and on checking the summarised data. I have also benefitted from the professional computing advice of Julian Read on the data base. Professor Brian Main, Professor Gordon Hughes, Mr. Ian White and Mr. avid ~i%%iamshave provided W&U! advice on t&tireconomeiiic airi; &ttisticaI fronts. The ideas deveioped in this paper were %irstbroached while the author was on sabbati~a%at Darwin Col%ege,university of Cambridge in 1987-1988. He wishes to acknowledge in genera%the rongeniaf i~te%%e~tua% at~os~bere which made that preliminary work possible, and in particular the helpful advice of Mr. Cliff Pratten of Trinity Hall at the point of inception of the re-interview project. Pau%Geroski provided helpful the referees who comments on the final draft. None of the persons mentioned above, provided detailed criticisms, is to be held responsible for any deficiencies contain. 0%67-7187/91/$03.50 C(:)%991-Elsevier Science Publisllers B.Q. All rights reserved
G.C. Reid, Staying in business
546
2. hether a firm stays in business or exits is regarded
as an aspect of rational choice. The recent work of Baden-Fuller (1989) on the exit of firm from the U.K. steel castings industry exemplifies one approach, that of intraindustry analysis, inspired by the theoretical work of Ghemawat and Nalebuff (1985). This paper, by contrast, uses an inter-industry approach. An appropriate formula for rational exit involves specifying a net profit variable (II*), obtained by deducting exit costs, along with operational costs, from gross profit. The requirement for staying in business for a period of years is I7*>=0. Rational exit occurs when I7* CO. Net profit Z7* is an unobserved variable, and is explained by observed market and financial variables. The general hypothesis is that profit, net of exit costs, (II*), is a function of market variables (M), financial variables (F), and of other variables captured by E: II* = f(M, F; E). This function f( e) may be linearised. The decision problem involves comparing the value of the unobserved variable I?* implied by current M and F variab!es with that value which justifies staying in business according to SEF’s opportunity costs. Z7* is mapped into the variable S = 1 for staying in business and S -=O for exiting, implying a binary probit model for estimation.’ This specification is consistent with rhe theory of Holmes and Schmidt (1990) according to which the actions lying behind these choices of staying in business or exiting are an aspect of the advanced division of labour in an enterprise economy. Entrepreneurship, so conceived, consists in seeking new products to exploit unharvested economic niches. This process might in. lve changing the balance of product lines. This ‘market repositioning’ is common with SEFs and is reported upon below. A point is reached at which repositioning would take a firm into new niche, in which case the old firm is wound up, and a new one started. This is identified as the exit of a firm, and clearly it is simply an extension of market repositioning.
Data were gathered by field work methods using administered questionnaires, as explained in Reid (1987). The population was the set of SEFs in Scotland, and the sampling frame was provided by the case loads in 1985 of Enterprise Trusts and Development Corporations or Agencies in the Lothians, Fife and Strathclyde regions of Scotland: the Edinburgh Venture Trust eith Enterprise Trust (LET), the Glenrothes Enterprise Trust (EV iterative mentcd
maximum
likelihood
in the S~~~~~
procedure
ecnnometric
used is based
software
package.
G.C. Reid, Staying in business
547
(GET), the Irvine Development Corporation (I C) and the Small Business Division of the Scottish Development Agency (SDA). There were 73 firms in the sample, being representative of the case loads of the agencies concerned. Details of the sample are given in Reid and Jacobsen (1988), so the treatment here is abbreviated. Inferences drawn from this sample should have direct empirical relevance to the handling of case loads in Enterprise Trusts and Agencies in the U.K., and hopefully more general conceptual implications for the modelling of small firm survival. Most SEFs in the sample were close to financial inception and few were over ten years old. The average age at the time of interview was 42 months. The average size was very small: eight full-time workers, two part-time workers and one trainee, in terms of employment; &76,000 in terms of book value of assets at 1985 prices; and &85,000 in terms of annual sales (excluding VAT), again at 1985 prices. Firm sizes ranged from one to 90, in terms of labour force, from ~1,000 to ;E500,000 in terms of assets, and from close to zero to over &l,OOO,ooO(at 1985 prices) in terms of turnover. In terms of SIC codes there was a wide dispersion of firms’ activities. Manufacturing (SICs 22-49) accounted for 56% of the total. Timber, wooden furniture, paper products, printing and manufacturing (SICs 46,47) were well represented (11%) as were electrical and electronic engineering (SIC 34) (10%) and food, drink, tobacco and textiles (SICS 41-4) (8%). Business services, banking finance and insurance (SICs 81-83) (7%) and wholesale and retail distribution (SICs 61-65) (10%) were also well represented. Regarding firm type, 50% were prilrate companies, 20% were partnerships and 30% were sole proprietorships. In 1988, three years after the initial field work, an attempt was made to trace comprehensively all firms which were involved in the initial 1985 study. A serious effort was made to provide an exhaustive categorisation of outcomes. Again field-work methods were used to gather data, of which only one component is relevant to this paper, namely whether the firm was still in business in i988. The participation rate was high (over 800/o) and, after intensive efforts, untraced cases were few (11%). Even for untraced cases other evidence could be brought to beak=(using comments of individuals in offrices, shops, or factory sites, adjacent to the original location of the firm in question) on whether business. Generally, there wa was that stayi1.g in b have retained a substantially similar clientele (i.e. go ing a substantially similar c
G.C. Reid, Staying in business
548
Table 1 Definitions of variables used in probit equations. -.Staying in Business (S) (determined by 1988 interviews) S= 1 for still in business in 1988; S=O for exiting from business between 1985 and 1988. Market Variables (M) (determined by 1985 interviews) Adu= 1 if SEF advertises, =0 if not. DesComp= 1 if competition in main market was intense, =2 if strong, =3 if generally weak, =4 if weak. E’mploy Number of full time employees. ProdDes= 1 if main product identical to rivals, = 2 if similar, : 3 if different. Pgroup Number of nroduct groups. Financial Variables (F) (determined by 1985 interviews) AddDebt= 1 if external finance had been used since starting the SEF, =0 if not. Pusset = Book value of total assets in f’OOOsat 1985 prices at time of interview in 1985. PerFin= 1 if only personal finances used to set up the SEF, =0 if not. Pgear Equity gearing ratio at time of interview in 1985 i.e. debt (borrowing) divided by owners’ injection of finance (personal injections). Other Variables Age Age in months from linancial inception to interview in 1985.
Generally, a feature of the data used are both their numerical accuracy, and their accuracy in the sense of congruence with the definitions and questions used in the 1985 and 1988 administered questionnaires. An audit was also conducted in 1988 on the data recorded on the physical schedules as compared to the data stored in the small firms database used for econometric estimation, so the quality of data should be high.
Broadly speaking the variables used to explain whether an SEF was in or out of business in 1988 were of the market (M) or financial (F) variety (cf. equation f( -) above). That is they were concerned with products, rivals, pricing etc. or with aspects of the balance sheet like debt, equity, assets and cash %&-WV. A novelty of the approach used in this paper is that it goes beyond an explanation of small firm survival (in terms of financial ratios and qualitative features of company accounts), as used in studies like Storey et al. (1987b), to an explanation that may be more persuasive to industrial economists, in tha. tt combines market with financial and other variables. Detailed defnnitions of these variables are given in table 1. Here I concentrate on their interpretation. 4.1.
a&et variables (M) &Des),
are concerned
with pro-
G.C. Reid, Staying ipzbusiness
549
reflection of this is the size of their product range, as measured by the number of product groups. The larger the product grou range, the more flexibly can the SEF reposition in the market in terms of niche exploitation, A more narrowly based SEF with fewer product group options is therefore less likely to stay in business.2 ProdDes is a self-appraised measure of prlndrsct differentiation and represents an alternative to more tricky measures like cross elasticity of demand. It was scaled to be greater, the greater the product heterogeneity. To the extent that product heterogeneity confers local monopolistic advantages on the SEF, the greater it is, the better the chances of staying in business. A related variable DesComp is a self appraised measure of intensity of competition. The variable was coded such that low values denoted intense competition and high values denoted weak competition. Given that competitive intensity erodes profits, one might expect low values of DesComp to be associated with a diminished prospect of staying in business. However, there is also a case for thinking in terms of what Porter (1985) calls ‘good competitors’, that is, competitors who by engaging in sharp and challenging rivalry, actually promote the efficiency and innovativeness of incumbent firms and hence enhance their prospects of staying in business. The variable Adv was simply a binary variable for whether the SEF did or did not advertise. To the extent that advertising confers monopolistic advantage at low cost, one would expect the SEFs which advertised to enhance their prospects of staying in business. 3 Age is a variable which does not have a strict economic interpretation but has been widely used in empirical studies of the growth of the firm. Brock and Evans (1986; ch. 6) have shown that the probability of surviving increases with age and with size for a large U.S. small business data base. There, the size measure used was, perforce, employment, rather than the more appealing assets measure: Here full time employment (Employ) and sales turnover in 1985 (Sales85) were used as size measures from the list of market variables. 4.2. Fiw,z!al
variables (F)
The book value of assets in 1985 (Passec) was used as the size measure ‘This interpretation of the consequences of the Pgroup variable upon staying in business is reinforced by the argument of Ungern-Sternberg (1990) that diversification into several products is one of the ways SEFs try to adjust to fluctuating demands for their individual products. ‘The issue of advertising intensity as distinct from advertising per se is, however, somewhat diherent, for increasing the range and quality of advertising messages is expensive; and a fine calculation [cf. Friedman (1983, ch. 6)] must be taken of its costs vis-a-vis its likely effectiveness in terms of revenue enhancement. I found some evidence [Reid (1989)] that SEFs were tempted to over-extend their advertising intensity, possibly by ir);ng7 to mimic large firm behavior in n some sure, t markets where advertising elasticities of demand were ac;ually quite 1~ ofa ti the nat this increased advertising intensity is unrelated to output, and this ass.11 cost which
depresses
net revenue
and reduces
prospeels
for staying
in husincss;
Sirnmon:;
( i9H%
550
G.C. Reid, Sraying in business
from the list of financial variables. My financial variables were not derived in a way which directly tallies with financial ratio analysis, t relationship between the approaches will be clear. In their univ analysis Storey et al. (1987b, ch. 6) particularly emphasised the importance of gearing and profitability for success or failure prediction. Gearing appeared to be a good short-term indicator of failure, having a tendency to rise as failure approaches. Profitability too appeared as a good predictor of failure, but by my argument 10~~profitability is what requires to be explained: it is a dependent, rather than independent, variable. In multivariate ratio methods, liquidity appeared as the better success/failure discriminator. Financial variables used in my probits which can be related to those used in financial ratio analysis include equity gearing (Pgear), assets (Pussets), debt (Add Debt), and cash flow (Cfp). (Pgear) was measured in 1985 at the time of interview with the administered questionnaire, and was defined as debt (i.e. borrowing) divided by the owner’s injection of equity (i.e. personal financial injection). In a Leland-Pyle (1977) world, the value of the SEF rises with the equity stake of the entrepreneur, and the equilibrium equity position is declining in both the riskiness of the firm’s activities and in the risk aversion of the entrepreneur. Essentially in a world where entrepreneurs have private (concealed) information about the quality of projects being undertaken by the SEF, the extent ol” iinrepreneurial ownership is a signal of worth. This suggests an inverse relationship between gearing and staying in business, in that firms with more worthy business operations have a better chance of surviving.4 A distinction needs to be made between financial inception and points further down the line when the owner manager may have established credibility by surviving. At inception, according to De Meza and Webb (1988), the owners of the SEFs with the highest quality projects will seek screened debt finance exclusively, those with projects of intermediate quality will seek to put all their wealth into the SEF and top-up with the necessary outside debt finance, and those with low quality projects will funded. Because banks know that the best entrepreneurs can supply finance to themselves on better terms than the market, they will tend to require maximal self-finance. If an entrepreneur balked at this, he would signal a poor project and would be refused debt. One therefore concludes that if a firm were launched exclusively on personal finance, this would signal a poor project. Thus the binary variable (PerFin) constructed by asking the owner manager whether he did (unity) or did not (zero) use exclusively nal finance in setting up his business should be negatively associated
4This is reinforced by the argument of Rlnzenco averse have a prefewnce for fq~lify ov6.x debt.
(1984)
ht
owner
G.C. &id,
Staying
in business
551
Even though gearing and survival should be negatively relate to acquire debt is a signal of project quality. The additional ;,&UD&t) c?as csnstructed by asking owner managers if they had used external finance (debt) since starting in business. Given that the affirmative was coded as unity, and that good projects should enhance survival, one would expect a positive association between the A&Debt variable and staying in business. Finally, I come to the cash flow variable (Cfp). Owner managers were asked if they had ever had cash-flow difftculties. A high proportion said ‘yes’ (73%) and one would expect the binary variable so defined to be at least a weak predictor of exit.’
Of the 73 SEFs which had been interviewed in 1985, 54 were still in business in 1988. That is, roughly three quarters of the firms managed to stay in business. A report produced in 1987 by Business in the Community (BIC) relating to business survivors for SEFs which were part of the case loads of Enterprise Agencies in England indicated 84% staying in business over a three year time scale. This compared with 66% staying in business over the same time period for all new firms which appeared on the VAT register. Smallbone (1989) reports in a single Enterprise Agency case load sample of 39 SEFs for an outer London borough and discovered 6477% staying in business (depending on how untraced SEFs were treated) after just one year. For Scotland’s largest local economic development company, BASE (Bathgate Area Support for Enterprise), Fass and Scothorne (1990) report 84% of SEES staying in business, for those started in 1984-87, and assessed in 1989. Storey and Johnson (1987a) have established a widely quoted yardstick of small firm surviva! over a three year period of 60%. Thus my figure of 74% of SEFs staying in business over three years for a sample drawn from case loads of diverse Enterprise Trusts in Scotland is not out of line with the range of figures suggested by the contemporary literature. In order to get a statistical explanation of this survival probability interindustry probit models were estimated for the cross-section of 73 firms. Market and financial variables were used as the independent (or ‘control’) variables, and the dependent variable was S- 1 for staying in 1988, and S=O for exiting from business between 1985 an industry data were used, as described above, wit coming from rna~~~ctur~~g and cQ~str~cti~~, a and services. Fishing and farming enterprises were not i %owever. as Jovanovic’s (1982) model suggests. exit may be caus a qequmce of unfavourable shocks from Nature’s urn, and even experience CilSll
IlOW
at least streaks pIX~bh?lS
m;ty
hC
of experience: when adve iL neicshury ht IlOt ;L SII
G.C. Reid, Staying in business
552
Tabk
2
Binary probit for full set of control
Variable
Coefficient
t-ratio
variables. Weighted elasticity
-0.17397 Pgear -0.95203. lo-’ - 2.9602*** - 0.22967 Perfin - 1S763 -2.0136** -0.16625. IO-’ Passer -0.14022~10-s - 0.35692 0.68865.10 - ’ AddDebt 0.54549 1.0665 -0.11556 UP - 0.63076 - 1.1495 0.39085 2.3379** 0.89963 P&Up -0.93581.10-l -0.25149 -0.70639 DesComp -0.40625. lo-’ - 0.28305. lo- ’ ProdDes -0.12053*10-’ 0.14828 Ado 0.895 13 1.0467 0.93608.10-’ 0.18165.10-’ 1.7659* Age 0.21038*1C-’ Employ 0.14030~ lo-’ 0.44059* 10-l 0.42897. lo- ’ ConstRnt 0.18499 0.22390 Likelihood ratio test, ~2=31.7454~~~~o.001(11)~31.3 Cragg-Uhler R2 =0.53250 McFadden R2 = 0.40063 Log-likelihood = - 23.747 Survival probability = 0.767 1 Legend: t,.,, = 1.296 (-f-). t,-,.OS= !.671(*), t0.r)2s = Z.OOO(**),to.o,o=2.390(***)
Table 2 reports on a probit equation for a large set of control variables of the market and financial variety. The significant variables are the current (or present) equity gearing ratio (@ear) at the time of interview in 1985, the use or not of exclusively personal finance (PerFin) at the time of financial inception, the number of product groups (Pgroup), and the age in months from financial inception to interview in 1985 (Age). Gearing and personal finance are financial variables, product group is a market variable and age is a technical variable suggested by theories of the growth of the firm [e.g. Pakes and Ericson (1990)]. The highest weighted elasticities tend to be associated with the most significant variables, except in the case of age. Whilst an increase in Age significantly improves the chance of staying in busirless over a three year time horizon the effect is not strong. High gearing (Pgear), having initial access only to personal finance (PerFin), a narrow product range (Pgroup) and relative youth (Age) all work against staying in business. The significance and sign of the PerFin variable provide confirmation of the theory of small firms finance of De Meza and Webb (1988) that the worst small firm projects are those launched using purely personal finance. ost non-significant varia les are of the expected sign, except in cases where the I ratio is very small, and even checking for qualitative properties
G.C. Reid, Staying in business
553
Table 3 Binary probit for sub-set of control Variable
Coefficient
t-ratio
variables. Weighted elasticity . _ .___ 0.36545, -0.17705 -0.17588 -0.12940.10-’ 0.46530.10 - ’ -0.86034.10-’ 0.93527*10-’ 0.88910.10-’ -0.10837.10-’
Pgroup 0.81107 2 3994** Pgear -0.94372. lo-’ - 3:3262*** PerJin - 1.1839 - 1.8627’ Passet -0.10438. 1O-s -(i._43123 AddDebt 0.35512 0.78573 DesComp - 0.23243 - 0.67556 ProdDes 0.28062 0.74830 0.16564.10-’ 1.7469* Age Constant -0.44638.10-t -0.66638 * lo- ’ Likelihood ratio test, x2 = 29.0434> ~20.001(8) E 26.1 Cragg-Uhler R2 = 0.49564 McFadden R2 =0.36653 Log-likelihood = - 25.098 Survival probability =0.7671 Legend: to,,,= 1.296 (f), t,,,,= 1.671(*), t,,,2s =2.000(**), to.,,,,,=2.390(***)
Descomp gives some support to the ‘good competitor’ notion. Having had cash flow problems (Cfp), and the existence of little proj.uct differentiation (Pro&es), reduce the chances of staying in business. %e measures, like assets (Passer) and number of full time employees (Employ) have less obvious consequences for survival, theoretically, and are less useful predictors empirically. A more parsimonious model is reported upon in table 3. variables in the probit of table 2 that were insignificant (e.g. have been dropped though some (e.g. AddDebt, Descomp) have been retained for their intrinsic theoretical interest, as has the Age variable. The probits of both table 2 and table 3 have probability levels of about 0.001 on a likelihood ratio test, and display high R’s for models of this sort. However, the more parsimonious model of table 3 is arguably preferable. On a likelihood ratio test with three degrees of freedom for comparing the two probits the test statistic is 2.702 which is less than the appropriate critical value of ~20.0s0(3)=7.81. The probit of table 3 is therefore preferred. The weighted elasticities of the dominating variable for each class of variabl Pgroup for A4 and Pgear for F, are relatively high. There seems a sli tendency for weighted elasticities to be greater for the financial variables as a class though this would be hard to test satisfactorily, given the dorn~~at~~g variables. qualitative and based tin Several of the variables reported upon above is is true, for example, the owner manager’s assessment of a situation. the variable whit
G.C. Reid, Staying in business
554
Table 4 Binary probit for sub-set of ‘objective’ variables. Variable
Coefficient
f-ratio
Weighted elasticity
Age Employ Passet
0.010234 1.2758 0.063996 -0.11336. 10-L - 0.032546 -0.019406. 10SZ 0.11006* lo-2 0.29426 0.013624 Pgear -0.60536. lo- 2 - 3.0148*** - 0.15028 Sales85 0.12133*10-2 0.64451 0.051110 Constant 0.66904 1.8674 0.19759 Likelihood ratio test, x2 = 19.0187> ~20,00s(5)= 16.7 Cragg-Uhler R’ =0.35585 McFadden R2 = 0.24759 Survival probability = 0.7671 Legend: t,,,! = 1.296(f), rO,cS= 1.671(*), t,,02s =2.000(**), ~~,~,~=2.390(***)
sheet. One such prObit iS pieXiib& iii GibIe 4. Here the Xlfitid variables used are age from financial i::eption (Age), full-time employees (Employ), book value of assets in 1985 (Basset), the debt/equity (i.e. gearing) L.&B L; 1985 (Pgear) and sales turnover in tax year 1984-1985 excluding VAT (Sale&). In general, the probit is less satisfactory than those reported in tables 2 and 3, but generally consistent with them, nevertheless. The gearing variable is the dominating variable, being highly significant and having a relatively high elasticity with respect to probability of survival. Shedding labour improves survival, as do larger sales and assets. Older SEFs have a better chance of survival than younger SEFs. The age variable when used in a variety of probits had a positive effect on survival as predicted by theory, but was usually of marginal significance. Size, as measured by assets, typically had a positive but insignificant eflect. This was also true of size measured by sales. Age was weakly positively correlated with size, measured by assets (r=0.258), and weakly negatively correlated (I= -0.149) with gearivlg. Comparing the approach exemplified by the specification in table 4, which uses only the most obvious ‘objective’ variables, with that exemplified by the specifications in tables 2 and 3, there does seem some advantage in a more eclectic approach, balance
The novelties of this paper lie In three directions. Firstly, the use of field work methods to provide a very detailed empirical picture of the forces contributing to an S stayers in business. erspective is that of as the quest for new
G.C. Reid, Sr~yi~~ in bushes
555
unusually, subjective data (e.g. degree of product di business. Thirdly, th th financial variables (F), inch and other variables familiar from financiai ratio analysis [as in (1987b)J and market variables etition, to provide a bro The condusion of t variables have a parti for this sample of SEFs: firstly, the product group range, a market variable; and secondly, the equity gearing ratio, a financial varia ased on tbe elasticities in table 2, ceteris paribus, a one percent increase in the product group range would raise the probability of staying in business by 0.39% and a one percent reduction in equity gearing would raise this probability by 0.17%.
eferences Acs, Z.J., D.B. Audretsch and B. Carlsson, 1990, Flexibility, plant size and restructuring, in: Z.J. Acs and D.B. Audretsch, eds., The economics of small firms: The European challenge (Kluwer, Dordrecht) is!-154. Baden-Fuller, C. W.F., i 983, Exit from declining industries and the case of steei castings, Economic Journal 99,949-96 1. Birch, D.L., 1981, Who creates jobs?, The Public Interest 65, 3-14. Blazenko, G.W.. 1987, Managerial preference, asymmetric information and financial structure, Journal of Finance 42,839-862. Brock, W.A. and D.S. Evans, 1986, The economics of smail businesses (Holmes and Meier, New York). De Meza, D. and D.C. Webb, 1988, Credit market efficiency and tax policy in the presence of screening costs, Journal of Public Economics 36, l-22. Fass, M. and R. Scothorne, 1990, The vital economy: Integrating training and enterprise (Abbeystrand, Edinburgh). Friedman, J., 1983, Oligopoly theory (Cambridge University Press, Cambridge). Ghemawat, P. and B. Nalebuff, 1985, Exit, Rand Journal of Economics 16, 184-194. Holmes, T.J. and J.A. Schmitz, 1990, A theory of entrepreneurship and its application to the study of business transfers, Journal of Political Economy 98, 265-294. Jovanovic, B., 1933, Selection and evolution of industry, Econometrica 50, 649-670. Leland, H. and D. Pyle, 1977, Information asymmetries, financial structure and financial intermediation, Journal of Finance 32, 371-388. models of firm dynamics, Pakes, A. and R. Ericson, 1990, Empirical implications of alternat’ dison, WI). Mimeo. (Graduate School of Business, University of Wisconsin, Porter, M., 1985, Competitive advantage (Free Press, New York). Rasmusen, E., 1989, Games and information (Basil Blackwell, Oxford). Reid, G.C., 1987, Applying field research techniques to the business enterprise, lnternationaf Journal of Social Economics 14, 3-25. Reid, G.C., 1989, Staying in business, Edinburgh University Economics paper no. IX.
556
G.C. Reid, Staying in business
Smallbone, D, 1989, Enterprise agencies and the survival of new business start-ups. Local Economy 4, 143-147. Storey, D.J. and S. Johnston, 1987a, Are small firms the answer to u~ern~~oyrne~t? (Em Institute, London). Storey, D.J., K. Keasey, R. Watson and P. Wynarczyk, 1987b, The performance of small firms (Croom-Helm, London). Ungern-Sternberg, T. von, 1990, The flexibility to switch between dififerent products, Economica 57, 355-369.
Journal of ~~d~stria~ ~rga~~~at~o~ 9 (~99%~545-556. worth-
Final version received December 1990
The paper analyses factors which help a small, entrepreneu~al firm to stay in busines. The time-scale for the analysis is three years (1985-1988). The main issue that the paper is concerned with is the role of market and financial variables. Primary source data are used. An estimated probit model indicates that, for this sample, product range and gearing are the key ma%setand financial variables, respectively. The greater the product range, and the lower the gearing, ceteris paribus, the better the chances of staying in business.
This paper is concerned with a question that has been central to recent industrial policy [See Storey et al. (1987a)J namely what factors help a small, entrepreneurial firm (SEF) stay in business. The time scale for the survival analysis is three years (1985-1988). The main issue that the paper is concerned with is the role of market and tinancial factors. Primary source data are used. An estimated probit model indicates that, for this sample, product range and gearing are the key market and financial variables, respectively. The greater the product range, and the lower the gearing, ceteris paribus, the better the chances of staying in business. *This paper could not have been written without a grant from the Nuffield Foundation which enabled me, where possible, to trace and interview in 1988 those firms first interviewed in 1985. Ann Theresa Lawrie provided useful research assistance on the field work, as did Jacqueline Campbell, who also updated the small firm data base and disp%ayed great diligence and persistence in obtaining the fullest information possible for a few elusive cases until the end of 1988. Chris Corrie assisted on computing and on checking the summarised data. I have also benefitted from the professional computing advice of Julian Read on the data base. Professor Brian Main, Professor Gordon Hughes, Mr. Ian White and Mr. avid ~i%%iamshave provided W&U! advice on t&tireconomeiiic airi; &ttisticaI fronts. The ideas deveioped in this paper were %irstbroached while the author was on sabbati~a%at Darwin Col%ege,university of Cambridge in 1987-1988. He wishes to acknowledge in genera%the rongeniaf i~te%%e~tua% at~os~bere which made that preliminary work possible, and in particular the helpful advice of Mr. Cliff Pratten of Trinity Hall at the point of inception of the re-interview project. Pau%Geroski provided helpful the referees who comments on the final draft. None of the persons mentioned above, provided detailed criticisms, is to be held responsible for any deficiencies contain. 0%67-7187/91/$03.50 C(:)%991-Elsevier Science Publisllers B.Q. All rights reserved
G.C. Reid, Staying in business
546
2. hether a firm stays in business or exits is regarded
as an aspect of rational choice. The recent work of Baden-Fuller (1989) on the exit of firm from the U.K. steel castings industry exemplifies one approach, that of intraindustry analysis, inspired by the theoretical work of Ghemawat and Nalebuff (1985). This paper, by contrast, uses an inter-industry approach. An appropriate formula for rational exit involves specifying a net profit variable (II*), obtained by deducting exit costs, along with operational costs, from gross profit. The requirement for staying in business for a period of years is I7*>=0. Rational exit occurs when I7* CO. Net profit Z7* is an unobserved variable, and is explained by observed market and financial variables. The general hypothesis is that profit, net of exit costs, (II*), is a function of market variables (M), financial variables (F), and of other variables captured by E: II* = f(M, F; E). This function f( e) may be linearised. The decision problem involves comparing the value of the unobserved variable I?* implied by current M and F variab!es with that value which justifies staying in business according to SEF’s opportunity costs. Z7* is mapped into the variable S = 1 for staying in business and S -=O for exiting, implying a binary probit model for estimation.’ This specification is consistent with rhe theory of Holmes and Schmidt (1990) according to which the actions lying behind these choices of staying in business or exiting are an aspect of the advanced division of labour in an enterprise economy. Entrepreneurship, so conceived, consists in seeking new products to exploit unharvested economic niches. This process might in. lve changing the balance of product lines. This ‘market repositioning’ is common with SEFs and is reported upon below. A point is reached at which repositioning would take a firm into new niche, in which case the old firm is wound up, and a new one started. This is identified as the exit of a firm, and clearly it is simply an extension of market repositioning.
Data were gathered by field work methods using administered questionnaires, as explained in Reid (1987). The population was the set of SEFs in Scotland, and the sampling frame was provided by the case loads in 1985 of Enterprise Trusts and Development Corporations or Agencies in the Lothians, Fife and Strathclyde regions of Scotland: the Edinburgh Venture Trust eith Enterprise Trust (LET), the Glenrothes Enterprise Trust (EV iterative mentcd
maximum
likelihood
in the S~~~~~
procedure
ecnnometric
used is based
software
package.
G.C. Reid, Staying in business
547
(GET), the Irvine Development Corporation (I C) and the Small Business Division of the Scottish Development Agency (SDA). There were 73 firms in the sample, being representative of the case loads of the agencies concerned. Details of the sample are given in Reid and Jacobsen (1988), so the treatment here is abbreviated. Inferences drawn from this sample should have direct empirical relevance to the handling of case loads in Enterprise Trusts and Agencies in the U.K., and hopefully more general conceptual implications for the modelling of small firm survival. Most SEFs in the sample were close to financial inception and few were over ten years old. The average age at the time of interview was 42 months. The average size was very small: eight full-time workers, two part-time workers and one trainee, in terms of employment; &76,000 in terms of book value of assets at 1985 prices; and &85,000 in terms of annual sales (excluding VAT), again at 1985 prices. Firm sizes ranged from one to 90, in terms of labour force, from ~1,000 to ;E500,000 in terms of assets, and from close to zero to over &l,OOO,ooO(at 1985 prices) in terms of turnover. In terms of SIC codes there was a wide dispersion of firms’ activities. Manufacturing (SICs 22-49) accounted for 56% of the total. Timber, wooden furniture, paper products, printing and manufacturing (SICs 46,47) were well represented (11%) as were electrical and electronic engineering (SIC 34) (10%) and food, drink, tobacco and textiles (SICS 41-4) (8%). Business services, banking finance and insurance (SICs 81-83) (7%) and wholesale and retail distribution (SICs 61-65) (10%) were also well represented. Regarding firm type, 50% were prilrate companies, 20% were partnerships and 30% were sole proprietorships. In 1988, three years after the initial field work, an attempt was made to trace comprehensively all firms which were involved in the initial 1985 study. A serious effort was made to provide an exhaustive categorisation of outcomes. Again field-work methods were used to gather data, of which only one component is relevant to this paper, namely whether the firm was still in business in i988. The participation rate was high (over 800/o) and, after intensive efforts, untraced cases were few (11%). Even for untraced cases other evidence could be brought to beak=(using comments of individuals in offrices, shops, or factory sites, adjacent to the original location of the firm in question) on whether business. Generally, there wa was that stayi1.g in b have retained a substantially similar clientele (i.e. go ing a substantially similar c
G.C. Reid, Staying in business
548
Table 1 Definitions of variables used in probit equations. -.Staying in Business (S) (determined by 1988 interviews) S= 1 for still in business in 1988; S=O for exiting from business between 1985 and 1988. Market Variables (M) (determined by 1985 interviews) Adu= 1 if SEF advertises, =0 if not. DesComp= 1 if competition in main market was intense, =2 if strong, =3 if generally weak, =4 if weak. E’mploy Number of full time employees. ProdDes= 1 if main product identical to rivals, = 2 if similar, : 3 if different. Pgroup Number of nroduct groups. Financial Variables (F) (determined by 1985 interviews) AddDebt= 1 if external finance had been used since starting the SEF, =0 if not. Pusset = Book value of total assets in f’OOOsat 1985 prices at time of interview in 1985. PerFin= 1 if only personal finances used to set up the SEF, =0 if not. Pgear Equity gearing ratio at time of interview in 1985 i.e. debt (borrowing) divided by owners’ injection of finance (personal injections). Other Variables Age Age in months from linancial inception to interview in 1985.
Generally, a feature of the data used are both their numerical accuracy, and their accuracy in the sense of congruence with the definitions and questions used in the 1985 and 1988 administered questionnaires. An audit was also conducted in 1988 on the data recorded on the physical schedules as compared to the data stored in the small firms database used for econometric estimation, so the quality of data should be high.
Broadly speaking the variables used to explain whether an SEF was in or out of business in 1988 were of the market (M) or financial (F) variety (cf. equation f( -) above). That is they were concerned with products, rivals, pricing etc. or with aspects of the balance sheet like debt, equity, assets and cash %&-WV. A novelty of the approach used in this paper is that it goes beyond an explanation of small firm survival (in terms of financial ratios and qualitative features of company accounts), as used in studies like Storey et al. (1987b), to an explanation that may be more persuasive to industrial economists, in tha. tt combines market with financial and other variables. Detailed defnnitions of these variables are given in table 1. Here I concentrate on their interpretation. 4.1.
a&et variables (M) &Des),
are concerned
with pro-
G.C. Reid, Staying ipzbusiness
549
reflection of this is the size of their product range, as measured by the number of product groups. The larger the product grou range, the more flexibly can the SEF reposition in the market in terms of niche exploitation, A more narrowly based SEF with fewer product group options is therefore less likely to stay in business.2 ProdDes is a self-appraised measure of prlndrsct differentiation and represents an alternative to more tricky measures like cross elasticity of demand. It was scaled to be greater, the greater the product heterogeneity. To the extent that product heterogeneity confers local monopolistic advantages on the SEF, the greater it is, the better the chances of staying in business. A related variable DesComp is a self appraised measure of intensity of competition. The variable was coded such that low values denoted intense competition and high values denoted weak competition. Given that competitive intensity erodes profits, one might expect low values of DesComp to be associated with a diminished prospect of staying in business. However, there is also a case for thinking in terms of what Porter (1985) calls ‘good competitors’, that is, competitors who by engaging in sharp and challenging rivalry, actually promote the efficiency and innovativeness of incumbent firms and hence enhance their prospects of staying in business. The variable Adv was simply a binary variable for whether the SEF did or did not advertise. To the extent that advertising confers monopolistic advantage at low cost, one would expect the SEFs which advertised to enhance their prospects of staying in business. 3 Age is a variable which does not have a strict economic interpretation but has been widely used in empirical studies of the growth of the firm. Brock and Evans (1986; ch. 6) have shown that the probability of surviving increases with age and with size for a large U.S. small business data base. There, the size measure used was, perforce, employment, rather than the more appealing assets measure: Here full time employment (Employ) and sales turnover in 1985 (Sales85) were used as size measures from the list of market variables. 4.2. Fiw,z!al
variables (F)
The book value of assets in 1985 (Passec) was used as the size measure ‘This interpretation of the consequences of the Pgroup variable upon staying in business is reinforced by the argument of Ungern-Sternberg (1990) that diversification into several products is one of the ways SEFs try to adjust to fluctuating demands for their individual products. ‘The issue of advertising intensity as distinct from advertising per se is, however, somewhat diherent, for increasing the range and quality of advertising messages is expensive; and a fine calculation [cf. Friedman (1983, ch. 6)] must be taken of its costs vis-a-vis its likely effectiveness in terms of revenue enhancement. I found some evidence [Reid (1989)] that SEFs were tempted to over-extend their advertising intensity, possibly by ir);ng7 to mimic large firm behavior in n some sure, t markets where advertising elasticities of demand were ac;ually quite 1~ ofa ti the nat this increased advertising intensity is unrelated to output, and this ass.11 cost which
depresses
net revenue
and reduces
prospeels
for staying
in husincss;
Sirnmon:;
( i9H%
550
G.C. Reid, Sraying in business
from the list of financial variables. My financial variables were not derived in a way which directly tallies with financial ratio analysis, t relationship between the approaches will be clear. In their univ analysis Storey et al. (1987b, ch. 6) particularly emphasised the importance of gearing and profitability for success or failure prediction. Gearing appeared to be a good short-term indicator of failure, having a tendency to rise as failure approaches. Profitability too appeared as a good predictor of failure, but by my argument 10~~profitability is what requires to be explained: it is a dependent, rather than independent, variable. In multivariate ratio methods, liquidity appeared as the better success/failure discriminator. Financial variables used in my probits which can be related to those used in financial ratio analysis include equity gearing (Pgear), assets (Pussets), debt (Add Debt), and cash flow (Cfp). (Pgear) was measured in 1985 at the time of interview with the administered questionnaire, and was defined as debt (i.e. borrowing) divided by the owner’s injection of equity (i.e. personal financial injection). In a Leland-Pyle (1977) world, the value of the SEF rises with the equity stake of the entrepreneur, and the equilibrium equity position is declining in both the riskiness of the firm’s activities and in the risk aversion of the entrepreneur. Essentially in a world where entrepreneurs have private (concealed) information about the quality of projects being undertaken by the SEF, the extent ol” iinrepreneurial ownership is a signal of worth. This suggests an inverse relationship between gearing and staying in business, in that firms with more worthy business operations have a better chance of surviving.4 A distinction needs to be made between financial inception and points further down the line when the owner manager may have established credibility by surviving. At inception, according to De Meza and Webb (1988), the owners of the SEFs with the highest quality projects will seek screened debt finance exclusively, those with projects of intermediate quality will seek to put all their wealth into the SEF and top-up with the necessary outside debt finance, and those with low quality projects will funded. Because banks know that the best entrepreneurs can supply finance to themselves on better terms than the market, they will tend to require maximal self-finance. If an entrepreneur balked at this, he would signal a poor project and would be refused debt. One therefore concludes that if a firm were launched exclusively on personal finance, this would signal a poor project. Thus the binary variable (PerFin) constructed by asking the owner manager whether he did (unity) or did not (zero) use exclusively nal finance in setting up his business should be negatively associated
4This is reinforced by the argument of Rlnzenco averse have a prefewnce for fq~lify ov6.x debt.
(1984)
ht
owner
G.C. &id,
Staying
in business
551
Even though gearing and survival should be negatively relate to acquire debt is a signal of project quality. The additional ;,&UD&t) c?as csnstructed by asking owner managers if they had used external finance (debt) since starting in business. Given that the affirmative was coded as unity, and that good projects should enhance survival, one would expect a positive association between the A&Debt variable and staying in business. Finally, I come to the cash flow variable (Cfp). Owner managers were asked if they had ever had cash-flow difftculties. A high proportion said ‘yes’ (73%) and one would expect the binary variable so defined to be at least a weak predictor of exit.’
Of the 73 SEFs which had been interviewed in 1985, 54 were still in business in 1988. That is, roughly three quarters of the firms managed to stay in business. A report produced in 1987 by Business in the Community (BIC) relating to business survivors for SEFs which were part of the case loads of Enterprise Agencies in England indicated 84% staying in business over a three year time scale. This compared with 66% staying in business over the same time period for all new firms which appeared on the VAT register. Smallbone (1989) reports in a single Enterprise Agency case load sample of 39 SEFs for an outer London borough and discovered 6477% staying in business (depending on how untraced SEFs were treated) after just one year. For Scotland’s largest local economic development company, BASE (Bathgate Area Support for Enterprise), Fass and Scothorne (1990) report 84% of SEES staying in business, for those started in 1984-87, and assessed in 1989. Storey and Johnson (1987a) have established a widely quoted yardstick of small firm surviva! over a three year period of 60%. Thus my figure of 74% of SEFs staying in business over three years for a sample drawn from case loads of diverse Enterprise Trusts in Scotland is not out of line with the range of figures suggested by the contemporary literature. In order to get a statistical explanation of this survival probability interindustry probit models were estimated for the cross-section of 73 firms. Market and financial variables were used as the independent (or ‘control’) variables, and the dependent variable was S- 1 for staying in 1988, and S=O for exiting from business between 1985 an industry data were used, as described above, wit coming from rna~~~ctur~~g and cQ~str~cti~~, a and services. Fishing and farming enterprises were not i %owever. as Jovanovic’s (1982) model suggests. exit may be caus a qequmce of unfavourable shocks from Nature’s urn, and even experience CilSll
IlOW
at least streaks pIX~bh?lS
m;ty
hC
of experience: when adve iL neicshury ht IlOt ;L SII
G.C. Reid, Staying in business
552
Tabk
2
Binary probit for full set of control
Variable
Coefficient
t-ratio
variables. Weighted elasticity
-0.17397 Pgear -0.95203. lo-’ - 2.9602*** - 0.22967 Perfin - 1S763 -2.0136** -0.16625. IO-’ Passer -0.14022~10-s - 0.35692 0.68865.10 - ’ AddDebt 0.54549 1.0665 -0.11556 UP - 0.63076 - 1.1495 0.39085 2.3379** 0.89963 P&Up -0.93581.10-l -0.25149 -0.70639 DesComp -0.40625. lo-’ - 0.28305. lo- ’ ProdDes -0.12053*10-’ 0.14828 Ado 0.895 13 1.0467 0.93608.10-’ 0.18165.10-’ 1.7659* Age 0.21038*1C-’ Employ 0.14030~ lo-’ 0.44059* 10-l 0.42897. lo- ’ ConstRnt 0.18499 0.22390 Likelihood ratio test, ~2=31.7454~~~~o.001(11)~31.3 Cragg-Uhler R2 =0.53250 McFadden R2 = 0.40063 Log-likelihood = - 23.747 Survival probability = 0.767 1 Legend: t,.,, = 1.296 (-f-). t,-,.OS= !.671(*), t0.r)2s = Z.OOO(**),to.o,o=2.390(***)
Table 2 reports on a probit equation for a large set of control variables of the market and financial variety. The significant variables are the current (or present) equity gearing ratio (@ear) at the time of interview in 1985, the use or not of exclusively personal finance (PerFin) at the time of financial inception, the number of product groups (Pgroup), and the age in months from financial inception to interview in 1985 (Age). Gearing and personal finance are financial variables, product group is a market variable and age is a technical variable suggested by theories of the growth of the firm [e.g. Pakes and Ericson (1990)]. The highest weighted elasticities tend to be associated with the most significant variables, except in the case of age. Whilst an increase in Age significantly improves the chance of staying in busirless over a three year time horizon the effect is not strong. High gearing (Pgear), having initial access only to personal finance (PerFin), a narrow product range (Pgroup) and relative youth (Age) all work against staying in business. The significance and sign of the PerFin variable provide confirmation of the theory of small firms finance of De Meza and Webb (1988) that the worst small firm projects are those launched using purely personal finance. ost non-significant varia les are of the expected sign, except in cases where the I ratio is very small, and even checking for qualitative properties
G.C. Reid, Staying in business
553
Table 3 Binary probit for sub-set of control Variable
Coefficient
t-ratio
variables. Weighted elasticity . _ .___ 0.36545, -0.17705 -0.17588 -0.12940.10-’ 0.46530.10 - ’ -0.86034.10-’ 0.93527*10-’ 0.88910.10-’ -0.10837.10-’
Pgroup 0.81107 2 3994** Pgear -0.94372. lo-’ - 3:3262*** PerJin - 1.1839 - 1.8627’ Passet -0.10438. 1O-s -(i._43123 AddDebt 0.35512 0.78573 DesComp - 0.23243 - 0.67556 ProdDes 0.28062 0.74830 0.16564.10-’ 1.7469* Age Constant -0.44638.10-t -0.66638 * lo- ’ Likelihood ratio test, x2 = 29.0434> ~20.001(8) E 26.1 Cragg-Uhler R2 = 0.49564 McFadden R2 =0.36653 Log-likelihood = - 25.098 Survival probability =0.7671 Legend: to,,,= 1.296 (f), t,,,,= 1.671(*), t,,,2s =2.000(**), to.,,,,,=2.390(***)
Descomp gives some support to the ‘good competitor’ notion. Having had cash flow problems (Cfp), and the existence of little proj.uct differentiation (Pro&es), reduce the chances of staying in business. %e measures, like assets (Passer) and number of full time employees (Employ) have less obvious consequences for survival, theoretically, and are less useful predictors empirically. A more parsimonious model is reported upon in table 3. variables in the probit of table 2 that were insignificant (e.g. have been dropped though some (e.g. AddDebt, Descomp) have been retained for their intrinsic theoretical interest, as has the Age variable. The probits of both table 2 and table 3 have probability levels of about 0.001 on a likelihood ratio test, and display high R’s for models of this sort. However, the more parsimonious model of table 3 is arguably preferable. On a likelihood ratio test with three degrees of freedom for comparing the two probits the test statistic is 2.702 which is less than the appropriate critical value of ~20.0s0(3)=7.81. The probit of table 3 is therefore preferred. The weighted elasticities of the dominating variable for each class of variabl Pgroup for A4 and Pgear for F, are relatively high. There seems a sli tendency for weighted elasticities to be greater for the financial variables as a class though this would be hard to test satisfactorily, given the dorn~~at~~g variables. qualitative and based tin Several of the variables reported upon above is is true, for example, the owner manager’s assessment of a situation. the variable whit
G.C. Reid, Staying in business
554
Table 4 Binary probit for sub-set of ‘objective’ variables. Variable
Coefficient
f-ratio
Weighted elasticity
Age Employ Passet
0.010234 1.2758 0.063996 -0.11336. 10-L - 0.032546 -0.019406. 10SZ 0.11006* lo-2 0.29426 0.013624 Pgear -0.60536. lo- 2 - 3.0148*** - 0.15028 Sales85 0.12133*10-2 0.64451 0.051110 Constant 0.66904 1.8674 0.19759 Likelihood ratio test, x2 = 19.0187> ~20,00s(5)= 16.7 Cragg-Uhler R’ =0.35585 McFadden R2 = 0.24759 Survival probability = 0.7671 Legend: t,,,! = 1.296(f), rO,cS= 1.671(*), t,,02s =2.000(**), ~~,~,~=2.390(***)
sheet. One such prObit iS pieXiib& iii GibIe 4. Here the Xlfitid variables used are age from financial i::eption (Age), full-time employees (Employ), book value of assets in 1985 (Basset), the debt/equity (i.e. gearing) L.&B L; 1985 (Pgear) and sales turnover in tax year 1984-1985 excluding VAT (Sale&). In general, the probit is less satisfactory than those reported in tables 2 and 3, but generally consistent with them, nevertheless. The gearing variable is the dominating variable, being highly significant and having a relatively high elasticity with respect to probability of survival. Shedding labour improves survival, as do larger sales and assets. Older SEFs have a better chance of survival than younger SEFs. The age variable when used in a variety of probits had a positive effect on survival as predicted by theory, but was usually of marginal significance. Size, as measured by assets, typically had a positive but insignificant eflect. This was also true of size measured by sales. Age was weakly positively correlated with size, measured by assets (r=0.258), and weakly negatively correlated (I= -0.149) with gearivlg. Comparing the approach exemplified by the specification in table 4, which uses only the most obvious ‘objective’ variables, with that exemplified by the specifications in tables 2 and 3, there does seem some advantage in a more eclectic approach, balance
The novelties of this paper lie In three directions. Firstly, the use of field work methods to provide a very detailed empirical picture of the forces contributing to an S stayers in business. erspective is that of as the quest for new
G.C. Reid, Sr~yi~~ in bushes
555
unusually, subjective data (e.g. degree of product di business. Thirdly, th th financial variables (F), inch and other variables familiar from financiai ratio analysis [as in (1987b)J and market variables etition, to provide a bro The condusion of t variables have a parti for this sample of SEFs: firstly, the product group range, a market variable; and secondly, the equity gearing ratio, a financial varia ased on tbe elasticities in table 2, ceteris paribus, a one percent increase in the product group range would raise the probability of staying in business by 0.39% and a one percent reduction in equity gearing would raise this probability by 0.17%.
eferences Acs, Z.J., D.B. Audretsch and B. Carlsson, 1990, Flexibility, plant size and restructuring, in: Z.J. Acs and D.B. Audretsch, eds., The economics of small firms: The European challenge (Kluwer, Dordrecht) is!-154. Baden-Fuller, C. W.F., i 983, Exit from declining industries and the case of steei castings, Economic Journal 99,949-96 1. Birch, D.L., 1981, Who creates jobs?, The Public Interest 65, 3-14. Blazenko, G.W.. 1987, Managerial preference, asymmetric information and financial structure, Journal of Finance 42,839-862. Brock, W.A. and D.S. Evans, 1986, The economics of smail businesses (Holmes and Meier, New York). De Meza, D. and D.C. Webb, 1988, Credit market efficiency and tax policy in the presence of screening costs, Journal of Public Economics 36, l-22. Fass, M. and R. Scothorne, 1990, The vital economy: Integrating training and enterprise (Abbeystrand, Edinburgh). Friedman, J., 1983, Oligopoly theory (Cambridge University Press, Cambridge). Ghemawat, P. and B. Nalebuff, 1985, Exit, Rand Journal of Economics 16, 184-194. Holmes, T.J. and J.A. Schmitz, 1990, A theory of entrepreneurship and its application to the study of business transfers, Journal of Political Economy 98, 265-294. Jovanovic, B., 1933, Selection and evolution of industry, Econometrica 50, 649-670. Leland, H. and D. Pyle, 1977, Information asymmetries, financial structure and financial intermediation, Journal of Finance 32, 371-388. models of firm dynamics, Pakes, A. and R. Ericson, 1990, Empirical implications of alternat’ dison, WI). Mimeo. (Graduate School of Business, University of Wisconsin, Porter, M., 1985, Competitive advantage (Free Press, New York). Rasmusen, E., 1989, Games and information (Basil Blackwell, Oxford). Reid, G.C., 1987, Applying field research techniques to the business enterprise, lnternationaf Journal of Social Economics 14, 3-25. Reid, G.C., 1989, Staying in business, Edinburgh University Economics paper no. IX.
556
G.C. Reid, Staying in business
Smallbone, D, 1989, Enterprise agencies and the survival of new business start-ups. Local Economy 4, 143-147. Storey, D.J. and S. Johnston, 1987a, Are small firms the answer to u~ern~~oyrne~t? (Em Institute, London). Storey, D.J., K. Keasey, R. Watson and P. Wynarczyk, 1987b, The performance of small firms (Croom-Helm, London). Ungern-Sternberg, T. von, 1990, The flexibility to switch between dififerent products, Economica 57, 355-369.