LIMITS ON THE PROSPECTS FOR SMALL FIRM LED EMPLOYMENT GROWTH: REGIONAL DEVELOPMENT AND THE SMALL FIRM Harvey Johnstone and David Kirby

Abstract Using Great Britain as a case study, a method of classifying environments with respect to their conduciveness to new-firm-formation (NFF) is developed. It is shown that the majority of localities needing to restore lost employment also occupy environments that are among the least-conducive to the formation of new firms. Thus, according to the small firm literature, these depleted localities¹ are at a marked disadvantage because employment growth has been associated with growth in firm registrations. However, in these particular cases, the situation appears to be more complex than the literature suggests. Associations between rates of NFF and rates of net employment change are very weak in depleted localities from the leastconducive environments. It is shown that, in these complex situations, the extent to which the employment contributions of new small firms are observable depends upon employment changes in the manufacturing sector. Implications for policy and research are explored. Introduction Since the late 1980s researchers in several countries have observed marked variations over space in rates of new firm formation (NFF). Studies examining the reasons for these observed differences, have suggested that factors, such as antecedent population change (Keeble, 1993), levels of self employment (Moyes and Westhead, 1990) and levels of home ownership (Ashcroft, Love and Malloy, 1991), have exerted considerable influence on rates of NFF 2 (see also -- Reynolds, Storey, Westhead, 1994). Localities possessing these factors are believed to offer environments that are conducive to the formation of new firms and enjoy high rates of NFF. Other localities providing less conducive environments have recorded relatively lower rates of NFF. Research of this kind fosters the notion that the small firm sector is not robust. That is, it assumes that the small firm sector is not able to 'overcome' the constraints imposed by the immediate environment. Certainly, the empirical evidence gathered to date would support the notion that robustness is not common. Nonetheless, it would be excessive to ignore the possibility of robustness. Of course, environmental differences and issues related to robustness will be of little consequence unless it can be shown that variations in rates of NFF matter. Are localities with the higher rates of NFF reaping benefits that are not enjoyed by other localities? Researchers have tried to assess the possible impacts that marked variations in NFF may have on host regions or host localities. It is believed that higher rates of NFF will lead to greater industrial diversification, and greater flexibility in coping with technological, market and economic change (Mason, 1991). Furthermore, it has been argued that higher rates of NFF will eventually lead localities to a firm-size-structure that is dominated by small firms; this, in turn, should give such areas a greater capability to generate more new businesses (Lever, 1987). There is evidence, as well, of new-firm-led-growth. In the UK, linkages between rates of new firm registration during the 1980s and rates of employment change for the same period have been explored. Ashcroft and Love (1994) reported on this association; finding a strong link between county rates of NFF, lagged by one year, and changes in total employment (1981-89) at the county level.

Therefore, the literature on small firms does provide reasons to believe that environmental differences and issues related to robustness matter. This is especially true if, as has been argued (Storey and Johnson, 1986), the small firm sector is the principal and perhaps the only source of new jobs. With respect to regional development policies, the literature suggests that, without prospects for new-firm-led-growth, localities have few prospects for job growth at all. This paper explores the implications of these findings with a particular focus on Britain's depleted localities. What kinds of environments do these localities occupy? With a demonstrated need for new employment, what are their prospects for employment growth? Do depleted localities require specialized policies in order to see employment restored? Determining the conduciveness of the environment Researchers in the UK have developed models that explain observed variations in rates of new firm formation across the 66 counties of Great Britain (Keeble, et. al., 1993). Using a similar approach, the stepwise regression reported here employs five variables: chgpop measures antecedent population change, pcntmanu measures the percentage of the labor force in manual occupations, prfindex measures relative peripherality, pcntownr measures the percentage of residents owning their own homes and pcntself measures the percentage of the population who were self-employed. A summary of the regression results is presented in Table 1. Table 1 -- Summary of Stepwise Multiple Regression Results Adjusted R Squared

0.78968

Standard Error

0.85995

F (5,60), sig = 0.0000

49.8115

Durbin Watson

2.1997

Y = 7.96 + 0.277V1-0.1229V2 + 0.415V3 +0.0255V4 + 0.2629V5 {(5.04)*** (7.10)*** (-4.99)*** (3.13)** (2.53)* (3.86)***}

Sig. *** = 0.000 ** = .003 * = .014

V1=PCNTSELF; V2=PCNTMANU; V3=PCNTOWNR; V4=PRFINDEX; V5=CHGPOP; CONSTANT=7.96. The five variables in the solution yielded an adjusted R Squared of 0.78968 (indicating considerable explanatory power in the equation), a standard error of 0.85995 and F=60.7 (5,60) sig.=0.0000. Multi-colliniarity is not likely to be a problem as the highest correlation between independent variables was 0.4113. The coefficients of the variables were of the expected signs. The high levels of significance achieved by all values of T along with the high value of F indicate considerable explanatory power. The Durbin Watson statistic was calculated as 2.1997 indicating that autocorrelation is not a likely feature of the data.

The outcome of this regression is comparable to other reported regressions of county level firm registrations with county level environmental factors. Thus, the results suggest that the five variables in this equation are able to "explain" a substantial proportion (almost 80%) of the variation in firm registration rates that occurred across the counties of Great Britain. In combination, these variables appear to be able to portray "something" about the counties that makes some of them more or less conducive environments for the development of small firms. These same variables can be used to operationalize the notion of conduciveness. The predictor variables used in the multiple regression just described were introduced into a discriminant analysis procedure. Discriminant analysis partitioned the 66 counties into groups. Each group possesses a particular combination of environmental factors. The technique is used to provide a means of combining these factors so as to distinguish counties with the most-conducive environments from counties with the least-conducive environments. Under this approach the question of robustness is still open and can be tested for. That is, under the discriminant analysis approach, a county could be labeled least-conducive and still record a new firm registration rate that was in the top tertial of all rates. If environmental labels were assigned solely on the basis of registration rates, then robustness would be impossible by definition. As part of the procedure, counties were classified as belonging to one of three groups based on whether their NFF rates belonged to the first, second or third tertial of the array of VAT registration rates. With such a division it was expected the group composed of counties in the lowest tertial would include many, but not necessarily all, of those counties whose environments are among the least-conducive. Similarly, the group composed of counties within the third tertial is likely to include many, but not necessarily all, of those counties whose environments are among the most-conducive. 'Predictions' generated by the discriminant analysis procedure were taken as a reasonable and accurate measure of what is meant by conduciveness. Thus, where counties are "predicted" to belong to the lowest group (group=0) they are considered to be the counties with the least-conducive environments. If, in fact, any of them have registration rates that rank higher than the 33rd percentile, then the performance of their small firm sector would be considered robust. In such cases, the sector's performance with respect to registration rates would be viewed as having 'overcome' the limits imposed by its environment. Similarly, counties predicted to belong to the highest group (group=2) would be considered to have the most-conducive environments. Again, under the discriminant analysis approach, it is at least possible for a county labelled as most-conducive to have a registration rate that is ranked below the third tertial. Results of a Wilk's Lambda discriminant analysis procedure using the five variables that appeared in the solution to the stepwise regression are reported in Table 2. As Table 2 shows, all three F statistics between pairs are significant at the 0.000 level with values for F as follows: Groups (0,1) F=7.8165, Groups (0,2) F=21.803 and Groups (1,2) F=4.937. As expected, the value for F on pair (0,2) is greatest since this pairing represents the extreme cases. Of the two Canonical Discriminant functions, clearly function number one is the most important, explaining 95.22% of the total variance while function number two explains the remaining variance. The eigenvalue for function one is 20.1 times greater than that of function number two. The primary importance of function one is also indicated by the Chi-Squared statistic (69.52) that is significant at the 0.000 level. Because of its overwhelming importance, comments will be confined to function one.

Table 2 -- Results of Discriminant Analysis Using Wilk's Lambda and Five Defining Variables Reported in the Stepwise Multiple Regression Fnc

Eigen

%Var

Canon. Corr.

Wilk's Lambda

Chi Squared

Level of Significance

1

1.859

95.22

.8064

.3199

69.52

0.000

2

.093

4.78

.2922

.9146

5.443

0.2448

Pairs

F (Significance)

NFF Rank

Actual Number

0 Least-conducive Predicted

1 Indeterminate Predicted

2 Most-conducive Predicted

(0,1)

7.8165 (0.000)

0

22

19 (86.4%)

2 (9.1%)

1 (4.5%)

(0,2)

21.803 (0.000)

1

22

4 (18.2%)

15 (68.2%)

3 (13.6%)

(1,2)

4.937 (0.0008)

2

22

1 (4.5%)

4 (18.2%)

17 (7.3%)

Overall

77.27%

Environments of Depleted Localities: Standardized canonical discriminant function coefficients indicate the relative importance of each of the five variables in each function. In function number one, PCNTSELF contributes most to the discriminant score with a coefficient of 0.77469. It is followed in importance by PRFINDEX with a coefficient of 0.608351 and CHGPOP with a value of 0.5455. PCNTMANU had a standardized canonical discriminant function coefficient of -0.53971. The first discriminant function, when evaluated at the group centroid for group 0 (those counties with the lowest registrations), equals -1.70166; this suggests that some of these counties are areas with declining populations and uncommonly high concentrations of people in manual or craft work. As Table 2 indicates, twenty-four of Great Britain's counties were categorized as the least-conducive environments. Twenty-one counties were considered to be among the most-conducive and the remainder were classified as indeterminate. Since the discriminant functions of environmental factors "predict" with 77.3% accuracy it can be seen that in only a handful of cases do actual county registration rates exceed or fall short of expected rates. So there was little evidence of robustness. Based on this information it is possible to determine the kinds of environments Britain's depleted localities occupy. Table 3 shows the distribution of depleted localities among the three environmental categories. Comparisons of depleted and non-depleted localities show the non-depleted group to be better positioned. Almost half (46.1%) of the non-depleted localities occupy the most-conducive environments while only 23.5% of depleted localities are in the most-conducive environments. In contrast, only 18.1% of the non-depleted localities are in the least-conducive environments while more than half (54.4%) of the depleted localities are in the least-conducive environments. Based on these figures it can be concluded that for more than half of Britain's depleted localities, the prospects of newfirm-led growth are relatively poor.

Table 3 -- Distribution of Depleted and Non-depleted Localities According to Conduciveness of Environment Least-Conducive Environments

Indeterminate Environments

Most-Conducive Environments

56

122

132

Mean rate of NFF per 100 of labor force

5.40

7.03

8.36

Mean rate of FTE employment change 1984-89

7.76

10.39

9.64

81

33

35

Mean rate of NFF per 100 of labor force

4.51

6.01

7.16

Mean rate of FTE employment change 1984-89

1.49

8.80

4.13

Non-depleted Localities

Depleted Localities

Small-Firm-Led Growth The small firm literature strongly suggests that localities occupying environments hostile to the formation of new firms will be disadvantaged (Mason, 1991; Lever, 1987; Ashcroft and Love, 1994). Table 3 shows the mean values of NFF rates and rates of FTE employment change for each type of environment. The table compares two groups: depleted and non-depleted localities. Most striking is the very sharp drop in average FTE employment growth for depleted localities occupying the least-conducive environments. This result seems to bear out the associations predicted by the small firm literature; namely, that as rates of NFF decline, so do rates of employment change. But, if poor employment growth among the depleted localities from the least-conducive environments has been caused solely by the lower rates on new firm formation, there should be evidence of a linear relationship between net employment change and the numbers of firms registered. That is, as rates of new firm registrations rise or fall, there should be some commensurable observable rise or fall in rates of net FTE employment. In fact, there is very little evidence of this. Figure 1A (omitted) shows a plot of these variables for the 81 localities that were both depleted and in the leastconducive environments. The plot in Figure 1A shows that new firm formation is not consistently linked to net employment change in these localities. The correlation between rates of new firm formation and rates of net FTE employment growth in these 81 localities was weak.

Figure 1B (omitted) shows the relationship between new firm registrations and total employment change for the same localities. Again there is little evidence of linearity. These outcomes raise the possibility of a more complex situation involving other influences. The Influence of Manufacturing A very weak association between employment and firm registrations suggests that many of these localities may have some additional feature that makes net increases in FTE employment less likely, irrespective of the level of new firm activity. What kind of feature might do this? One possible source of insight is provided by researchers exploring Britain's de-industrialization (Champion and Townsend, 1990; Martin and Rowthorn, 1986; Allen and Massey, 1988; Massey, 1984). They have emphasized the added importance of particular jobs -- those belonging to the 'economic base3'. For example, Champion and Townsend (1990) report a strong positive correlation between employment changes in sectors comprising the economic base and changes in total employment. In positive circumstances, jobs added to the economic base can have multiplier effects and beget other jobs. Under negative conditions, jobs lost from the economic base can ultimately place other jobs in jeopardy. During this period Britain's manufacturing sector, a key component of the economic base, experienced considerable upheaval and the distribution of these negative impacts was spatially uneven. So we should look to the manufacturing sector. If there are indeed two factors influencing net employment growth, this may help to explain the very low rates of FTE employment change experienced by depleted localities from the least-conducive environments. Those particular localities may have felt the combined effects of lower rates of NFF and manufacturing employment losses. Since the intention is to show how the manufacturing sector influences employment, growth rates of FTE employment change 1984-89 are plotted against rates of manufacturing employment change 1984-89. This is done for the 81 localities that were both depleted and occupying environments least conducive to the formation of new firms. Two distinct cases are explored: first, the case of those localities with FTE manufacturing employment above the group median {Figure 2(A)}; second, the case of those with rates below the median {Figure 2(B)}. Each case will be discussed in turn. In Figure 2(A) it can be seen that 40 localities experienced rather minor changes in manufacturing employment between 1984 and 1989. Furthermore, the correlation between manufacturing employment change and overall employment change was very weak (R= -0.01). In these cases, the manufacturing sector provided stability but did not lead employment growth. FTE employment growth in these localities averaged about 6.6%. Contrasting with this is the case of depleted localities with lower than average manufacturing employment growth {Figure 2(B)}. Here, manufacturing employment changes tended to be negative. Also, in contrast to the first case, changes in manufacturing employment appear to have exerted a much stronger influence on total employment change. The two rates are strongly linked as evidenced by a positive correlation (R=0.6513, Sig. 0.000). In these cases manufacturing did affect total employment but caused it to fall. These outcomes suggest that the association between new firm formation and net employment may only be observable when changes in manufacturing employment are relatively stabilized.

To test this, once again a plot of FTE employment change with new firm registrations was made for each group. The results are presented in Figure 3A and Figure 3B (both omitted). For depleted localities whose rates of manufacturing employment were above average there was evidence of a linear relationship (Figure 3A--omitted). The coefficient of determination indicated that over half of the variation in net employment could be explained by levels of new firm formation {R2=0.811}. The slope of the regression line {m = 1.09231} indicates that for every new firm registering, on average, net employment increased by nearly one FTE job. It is important to emphasize that these are not cases where manufacturing 'led employment growth'. Correlations between manufacturing employment change and overall employment change are quite weak in these localities. Rather, it seems more appropriate to characterize these as cases where the 'stability' provided by the manufacturing sector made smallfirm-led growth possible. In Figure 3B (omitted) there is little evidence of a linear relationship between the number of new firm registrations and employment change; R2=0.0504 and the slope of the fitted line is -0.1573. Therefore, in some depleted localities from the least-conducive environments the conditions necessary for a 'combined effect' were in place. The combination, of lower levels of new firm formation and a manufacturing sector that was unable to provide stability, would help to explain why rates of FTE employment change were so much lower in these particular localities (3.5%). Implications for Depleted Localities from the Least-conducive Environments Even in the most hostile environments new small firms contributed to job creation. They may be a key source of new jobs in all environments and some argue that they have more potential to create jobs than large firms (Robson and Gallagher, 1994). Whether these contributions would be detected really depends upon the stability of the baseline employment, particularly in the manufacturing sector. The research suggests that, on their own, small firms cannot be expected to reverse the effects of enormous structural change in the short space of time examined here. It seems that at best they can do is ameliorate, but not overcome, employment losses arising from structural changes and industrial decline. So small firm led growth is contingent on the stability of existing employment. In other words, with respect to local recovery, there are limits on what may be expected from the small firm sector. The sector cannot provide a panacea for every problem. Generally the research reveals little or no evidence of robustness. Rates of new firm registration from the leastconducive environments were significantly lower than registration rates in other environments. Only seven of these localities had rates of new firm registration greater than the national median rate. Of these seven, only one had an employment rate above 7.0%. So, not only were there very few cases where new firm registration rates were higher than the environmental conditions would suggest they "should be", but, furthermore, even when this did occur, there was no guarantee that an abundance of new firms would lead to net employment growth! Conclusion Evidence gathered here suggests that a policy intending to create employment through a strategy of new firm formation is unlikely to be successful unless there are concurrent efforts to stabilize existing employment, particularly in the manufacturing sector. Existing enterprises under threat should be identified and, where feasible, attempts should be made to secure their positions within the industry. In Canada, for example, a community based venture capital company has been successful in rescuing faltering enterprises (MacLeod and Johnstone, 1995). A study of SMEs and manufacturing employment in the UK indicates that efforts should be focused on ensuring that these small manufacturing firms survive as a high proportion of job loss was attributed to firm deaths (North, Smallbone, Leigh, 1994). Without steps like these, which are aimed at stabilizing existing employment, the impact of new small firms may be quite limited.

This research has established that high rates of new firm registration are neither a necessary nor a sufficient condition for net employment growth. As such, this raises (yet again) the issue of cost effectiveness of blanket policies aimed at stimulating new firm startups (Storey, D., 1992 and Barkham, 1990). In the UK considerable research effort has gone into determining which factors influence rates of firm registration. Unfortunately, for those intending to stimulate new firm registration rates, many of the factors seen to influence rates of new firm registration are difficult to alter (Reynolds, 1993). But, even if policy could stimulate higher rates of firm formation the result may not be what was hoped for. What the current research suggests is that more information is needed about the differences in conditions between localities where net employment grows when firm registration rates are high and localities where net employment does not grow when firm registration rates are high. Secondly, more research must be devoted to identifying methods of overcoming the particular problems faced by the pool of potential recession-pushed-entrepreneurs. The evidence presented here also implies that many redundant workers are currently unable to make the transition to recession pushed entrepreneurship. Lack of information on this topic is a weakness in the literature dealing with small firms and their impact on employment growth. Appendix: Chart Of Variables Used In Regression And Discriminant Analysis REGPR100

VAT registrations in all vat-industry sectors for the period from 1980 to 1989. Source, Nomis: datta=vat, year=1980-1989, vat-industry=11, lad-1-459.

PCNTMANU

Percent of employees in manual occupations in 1981. Source, Nomis: data=occ, year=1981, ocstatus=6-7, broadwoc=6-7.

PCNTOWNR

Percentage of privately owned homes in 1981. Source, Nomis: data=sas, year=1981, ratio=5408/4952 county=1-66.

PCNTSELF

Percentage of labor force who were self employed in 1981. Source, Nomis:data=occ, year=1981, ocstatus=1-2, broadwoc=1-6.

CHGPOP

The percentage change in population between 1975 and 1981 in each county. Source, Regional Statistics No. 14, 1979.

PRFINDEX

A measure of the relative accessibility of each of the counties in 1981. Source, Owen, D. and Coombes, M., 1983, An Index of Peripherality for local areas in the United Kingdom, Regional Development Studies, University of Newcastle upon Tyne.

References Allen, J. and Massey, D., eds. (1990), The Economy in Question, London: Sage. Ashcroft, B., Love, J., (1994), "Employment Change and New Firm Formation in GB Counties: 1981-89". Paper presented at the New Firm Formation and Regional Economic Development Conference, Craigie College, Ayr, Scotland. Ashcroft, B., Love, J.H., Malloy, E. (1991), "New Firm Formation in the British Counties with Special Reference to Scotland", Regional Studies, vol. 25, pp. 395-409.

Barkham, R. (1990), "Entrepreneurship and New Firm Growth", Discussion Paper in Urban and Regional Economics, Series C, no. 49, Department of Economics, University of Reading, March. Champion, A. and Townsend, A., (1990), Contemporary Britain: A Geographical Perspective, London, Edward Arnold. Coombes M. and Raybould S., (1989), "Local 'Enterprise Activity' Potential (LEAP): Variability in Small Firm Growth", Northeast Regional Research Laboratory Research Report 8913, University of Newcastle Upon Tyne. Daly, M. (November 1991), "VAT Registrations and Deregistrations in 1990", Employment Gazette, pp. 579-88. Daly, M., Campbell, M., Robson G., Gallagher, C. (August 1992), "Job Creation 1987-89: preliminary analysis by sector", Employment Gazette, pp. 387-92. Gallagher, C., Daly, M., Thomason, J. (February 1990), "The Growth of UK Companies 1985-87 and their Contribution to Job Generation", Employment Gazette, pp. 92-8. Keeble, D. (1990a), "Small Firms, New Firms and Uneven Regional Development in the United Kingdom". Area, vol. 22, pp. 234-45. Keeble, D. (1990b) "New Firms and Regional Economic Development: Experience and Impacts of the 1980s". In G. Cameron, B. Moore, D. Nicholls, J. Rhodes and P. Tyler (Eds.), Cambridge Regional Economic Review, Cambridge: Dept. of Land Economy, Cambridge University. Keeble, D., Walker, S., and Robson, M., (1993), "New Firm Formation and Small Business Growth in the United Kingdom: Spatial and Temporal Variations and Determinants", Employment Department, Research Series No. 15, Sheffield UK. Lever, W.F., ed. (1987), Industrial Change in the United Kingdom, Essex: Longman Scientific & Technical. MacLeod, G., Johnstone, H. (1995), "The BCA Experiment in Finance", Proceedings of the 25th Annual Atlantic Schools of Business Conference, Sydney N. S., Canada. Mason, C. (1991), "Spatial Variations in Enterprise: the geography of new firm formation", in R Burrows (ed) (1991) Deciphering the Enterprise Culture: Entrepreneurship, Petty Capitalism and the Restructuring of Britain, London: Routledge, pp. 74-106. Massey, D. (1984), Spatial Divisions of Labour: Social Structures and the Geography of Production, London: MacMillan Education Ltd. Moyes, A., Westhead, P., (1990) "Environments for New Firm Formation in Great Britain", Regional Studies, vol 24 pp. 123-136. North, D., Smallbone, D., Leigh, R., (1994), "Employment and Labour Process Changes in Manufacturing SMEs during the 1980s" in D. Atkinson and D. Storey, (eds.) (1994), Employment, The Small Firm and The Labour Market, London: Routledge.

Reynolds P., (1993) "Regional Characteristics Affecting Business Volatility in the United States, 1980-4". In Karlsson C., Johnnnisson B. and Storey D., (Eds.), Small Business Dynamics: International, National and Regional Perspectives, London: Routledge, pp. 78-116. Reynolds P., Storey D.J., Westhead P. (1994) "Cross National Comparisons of the Variation in New Firm Formation Rates", Regional Studies, Vol 28, No. 4, pp. 443-456. Robson G. and Gallagher C. (1994), "Change in Size Distribution of UK Finns", Small Business Economics, Vol. 6, No. 4, pp. 299-312. Storey, D.J., Johnson, S.G. (1986), "Job Creation in Small Firms: A Review", International Journal of Small Business, Vol. 4, No. 4 pp. 26-41. Storey, D., (1992), "Should We Abandon Support For Start-Up Businesses?", A paper presented to the 15th National Small Firms Policy and Research Conference, Southhampton, November 1992. About the Authors David Kirby, Dean and Pro Vice-Chancellor, University of Middlesex Business School, London, UK Harvey Johnstone, Associate Professor of Business and Director of SME Institute, School of Business, University College of Cape Breton Contact person: Harvey Johnstone School of Business University College of Cape Breton P.O. Box 5300 Sydney, Nova Scotia, B1P 6L2 Canada. Tel. 902 563 1178 Fax: 902 563 1453 e-mail: [email protected]

1

A locality refers to a local authority district. A locality was considered depleted if its rate of full time equivalent employment change fell below the national rate for the period 1981-1984 and if its 1981-1987 employment figure was below the expected value based on national rates. Britain has 66 counties containing 459 local authority districts. 2 3

In Britain, NFF rates are frequently measured by registrations for VAT normalized by labor force.

The economic base is comprised of divisions 0-4 of the 1980 Standar Industrial Classification as wel as producer services.