Growth and Profitability of Small and Medium Size Enterprises:

Some Welsh Evidence

James Foreman-Peck, Gerry Makepeace and Brian Morgan
Cardiff Business School, Cardiff University, Colum Drive, Cardiff CF10 3EU

Abstract
Policy makers throughout the OECD are keen to promote the development of Small and

Medium Enterprises (SMEs) by a wide variety of means. The present study of over

1600 Welsh SMEs casts doubt on the effectiveness of doing so by grants, by

encouraging trade associations, by pursuing current initiatives in training or by

stimulating innovation as defined by SME management. On the other hand it finds

that support for selective IT investment and for marketing planning could be

beneficial. Additional results that might be addressed by policy are short-termism

by older managers- preferring current to future profits more than younger

managers, signs that fixed costs of compliance bear substantially more heavily on

the profitability of smaller SMEs, and evidence for a substantial partnership and

sole trader ‘risk premium’ between SMEs, suggesting considerable entrepreneurial

risk aversion.

JEL classification L25, D21, R32

Keywords: SMEs, growth, profitability, policy, risk, time-preference.

1
 Growth and Profitability of Small and Medium Enterprises:

Some Welsh Evidence*

The plethora of national programmes intended to stimulate Small and Medium

Enterprises (SMEs)1 prompted a ministerial level conference at Bologna in 2000

sponsored by the OECD and the Italian government (OECD 2002 p17). Subsequently the

‘Bologna process’ has become a frame of reference for governments concerned to

improve the efficiency of programmes intended to foster entrepreneurship and assist the

development and competitiveness of SMEs.

Programmes pursue a variety of objectives with an even greater range of policy

instruments. They are intended to promote innovation, remedy supposed market failures

in financing SMEs, encourage regional development, boost employment, and reduce

regulatory compliance burdens, among other purposes. Their means include encouraging

SME use of E-commerce and ICT2, management enhancement by mentoring3, promotion

of networks and ‘clusters’, for instance with ‘business incubators’ 4, and venture capital

funds5. Recently policy has tended to shift away from direct financing, favouring instead

qualitative support, partnership and collaborative approaches.

The purpose of this paper is to provide some evidence on the likely effectiveness of

current policy trends by testing for the impact of some key variables on Welsh SME

growth and profitability. The following section discusses the comparability of the Welsh

SME sector with those elsewhere. Section 2 explains and justifies growth and

2
profitability as SME performance measures. The determinants of this performance are

outlined in section 3. The data set is described in section 4 and the results presented in

section 5.

1. How Representative is the Welsh SME Sector?

The wider relevance of national or regional case studies depends on how representative

or distinctive is the sector studied. Findings of significant or insignificant covariance

within the area may carry over to other zones if the study region is typical of others. But a

focus on a distinctive region could amount to ‘sample selection’ leading to inappropriate

inferences from covariances for other areas.

In all advanced economies SMEs produce a high proportion of national output and

provide an even greater proportion of employment. The ratio of output share to

employment share is a simple measure of SME labour productivity relative to the

national average. When the ratio is unity, SME’s productivity is equal to the mean. In all

cases in Table 1 the ratio is below one. The extent to which it is below, together with the

size of the sector, gives an indication of possible gains from effective SME policies.

Wales shows a low SME manufacturing share, reflecting the high productivity of large

foreign direct investment plants. The UK as a whole is comparable to France and

Australia, with about half of manufacturing employment in SMEs, but relative

productivity is lower. Though considerably smaller than Sweden’s, Wales’ SME

manufacturing sector is more similar in the low output and employment shares to the

3
Swedish sector than to that of the entire UK. Welsh relative productivity for

manufacturing SMEs is above the UK average.

TABLES 1 AND 2 ABOUT HERE

When the whole Welsh economy is considered the SME share looks very different from

manufacturing. Welsh SMEs account for the highest proportion of total output and

employment of any region of Great Britain6. The UK all industry percentage of

employment provided by SMEs is similar to that for manufacturing but the proportion of

output is much higher. SMEs in London and the South East- where the finance boom was

prominent- actually contribute more to output than employment. As table 2 shows, this is

not true of other areas of the UK but it is clear that the labour productivity advantage of

larger firms is not especially marked. Relative productivity of Welsh SMEs is below that

for the entire UK, but not by much. The productivity gap is much lower for industry as a

whole than for manufacturing alone, both for Wales and for other parts of the UK. On

balance then, judged by relative productivity the Welsh SME sector does not appear

particularly different from the UK SME sector.

2. Measuring SME Performance

Some studies of SME performance construct composite measures of a number of

variables, for example by cluster analysis (Smith 1999; Reid and Smith 2000). Here we

consider two measures separately, - profitability and growth- on the grounds that SMEs

may reasonably choose to target one or the other. Most performance indicators, such as

the level of productivity, are merely intermediate variables influencing the one or other of

4
our two principal variables of interest in ways that will vary from firm to firm7.

Profits are necessary for survival in a competitive environment but SME management

may choose not to grow. Long-term profitability derives from the relations between cost

and revenue; it is a necessary but not sufficient condition for growth. Revenues may be

held up by entry barriers and costs pushed down by management ingenuity. A low profit

firm will lack the finance for expansion but a high profit business may conclude the risk

and rewards of expansion are inadequate- a ‘life-style’ SME for instance. Profitability

today may be traded off against profitability tomorrow. Dynamic pricing may require

initially lower profits in order to obtain higher future profits from greater market

penetration. An SME manager’s time preference is likely to determine the inter-temporal

profit trade-off.

The (static) neoclassical profit function- for the perfectly competitive maximising firm

-depends on output and input prices and ‘technology’, or, in the restricted version,

includes fixed inputs as arguments as well. Profit performance must be standardised

against the size of the operation or the resources employed. Return on capital equations

tend to adopt a specification derived from portfolio management (Söderbom and Pattillo

2002). Cross-section returns are expected to differ between sectors because of systematic

risk. Also higher accounting profits will be necessary in more capital intensive activities.

Reliable (or any) measures of capital against which rates of return could be measured are

not available for the present exercise. Consequently a profit–turnover dependent variable,

which can be derived as an equilibrium conditions from output-choosing Cournot-Nash

5
firms, is employed here (Cowling and Waterson 1976).

The second performance measure is SME growth. Managers were asked whether their

firm was a high, medium, low or no growth enterprise. The justification for the subjective

growth measure of performance employed is the lack of data that many SMEs are willing

or obliged to put in the public domain. It is also to some extent a forward looking variable

for which data in principle would be unavailable. Quite possibly management beliefs will

be systematically biased- towards optimism for example (De Meza 2002). But general

optimism bias will merely affect the scaling of the growth measure: a ‘medium growth’

response may in fact correspond with a ‘low growth’ performance8.

That the subjective growth category is an indicator of real future growth on average is

shown by a sub-sample of SME growth categories for 1998/9 which was matched as far

as possible with the growth in turnover and /or in employment data for 1998-2001

available in the FAME data base. Table 3a below shows that the ranking of the three

growth categories is correct.

TABLE 3 ABOUT HERE

A similar conclusion emerges from a comparison of historical turnover growth with

forward looking growth category. Table 3b classifies the 238 cases instances where the

FAME data base yielded three years of turnover for the 1600 cases in the sample analysed

here. The mean growth for the ‘low’ and ‘no’ growth category was a 6.6% decline but the

ordering of the three categories of growth belief or expectations corresponds with past

growth experience9.

6
The percentage of firms in each growth category is: ‘no and low’ growth (36%), medium

(47%) and high (17%). The growth variable is available for over 98% of the firms in the

sample but profit /turnover only for 41%. Since there is a large number of missing values

for profits/turnover, question arises as to whether firms with this information are typical.

The technique appropriate to answering this question requires some explanation of why

the data are missing. Fortunately the data, collected in two batches, provide an effective

instrument. 42% of the firms replied to the questionnaire without prompting, while the

remaining 58% only replied to a telephone follow-up. This response pattern was

modelled with a binary variable that takes the value 0 if the firm replied to original

questionnaire and 1 if to the telephone survey. Responding to the telephone follow-up is

significantly associated with failure to provide the profit figures.

3. Determinants of SME Performance

Identifying the determinants of SME performance necessitates controlling for a wide

range of possible contributors in an empirical model (Barkham et al 1996 Table 4.1

include 25 in their model). Restricting analysis to a small selection of variables runs the

risk of spurious correlations; the effect of one variable may be conflated with that of a

second with which it is correlated if this second variable is not also included in the

model. Unlike purely theoretical models, comprehensiveness can only be sacrificed to

coherence in empirical modelling at the expense of accuracy. The downside of

considering a multiplicity of possible contributors to SME performance is that the

7
literature survey accompanying the specification discussion for reasons of space must

necessarily be compressed and selective.

Another divergence between the demands of theoretical and empirical modeling arises

from the need to find measurable counterparts to theoretical variables in the second case.

This aspect of model specification involves sensitivity both to theoretical considerations

and to evidence of previous studies. A particular challenge is to avoid rejecting a

theoretical scheme because of an inappropriate empirical measure of a variable. At the

same time testing empirical specifications stimulates the adjustment of theoretical

speculation to evidence.

In the present section therefore we discuss a wide range of possible theoretical

contributors to SME performance and their measures, while considering some initial,

bivariate associations with firm growth in the present sample (summarized in Table 4).

Such associations can be suggestive of an impact but are not definitive; the multivariate

analysis of the following section is necessary to isolate the true effects of influences upon

growth and profitability.

A comprehensive review of the SME literature concluded growth determinants fell into

three groups; management characteristics, firm characteristics, and business strategy

(Storey 1994). To these perhaps should be added the general business environment. Static

8
theories of the firm point in particular to two ‘business environmental’ influences on

SME performance, legislation and competition.

TABLE 4 ABOUT HERE

The Business Environment

Legislation might explain the differences between economies in the contribution of SMEs

identified in Table 1. The transaction cost theory of the firm (Coase 2001) asserts that

firms exist because of the higher costs of using the market for all transactions. It is

cheaper and more effective in some circumstances to create an employment relationship

within a firm than to contract with independent workers for particular jobs when needed.

The average size of firm in an economy on this theory will be greater the higher is the

cost of using the market. Conversely the average size of firm will be smaller the more

costly are transactions within the firm. Thus taxation, employment legislation and other

government regulation will influence differences between countries in enterprise sizes.

Changes in such legislation will encourage alteration in mean business dimensions.

Exempting smaller enterprises from reporting requirements for instance should reduce

average firm size. When as is usual, smaller businesses are less productive, arithmetically

at least, legislation encouraging SMEs without raising their productivity could lower

national income.

SMEs typically cannot provide the administrative support of large firms because they

lack the turnover over which these fixed costs can be efficiently spread. They must buy in

services and/or undertake tasks with less specialised staff. An example of these handicaps

is VAT compliance costs, which are extremely regressive; 7.8% of revenue for lowest

9
turnover range of firm in 1986-7 compared with 3.69% for the average (Sandford et al

1989). Not surprisingly then there is evidence that greater regulatory costs

disproportionately force smaller firms out of business (for example Ollinger and

Fernandez-Cornejo 1998 on environmental regulation).

In the present sample high growth firms are less likely to see PAYE and VAT as

problematic than low growth SMEs, a result just significant at the 5 percent level (  2 [8]

=15.55 Prob value =0.049). The association may simply indicate the quality of

management in the different groups of firms in the face of the inevitable administrative

challenges of Welsh business life. But another interpretation is that more firms would

enter the high growth category if the tax regime was less demanding. This would be

consistent with Hansford et al’s (2003) finding that SMEs that perceive VAT compliance

as most burdensome also experienced higher compliance costs. Environmental legislation

and regulation at first sight seem to be less demanding than PAYE and VAT because there

is no significant association between the perceived burden and growth category (not

reported).

Government SME policy is also a feature of the SME business environment. Policy

includes grants and forms of support that have been analysed in a variety of studies (for

example Bennett, Robson and Bratton 2001; Bennett and Robson 2003; Gillespie et al

2001; Robson and Bennett 1999; Rodriguez-Pose and Fratesi 2004). In the present data

set neither government grants being thought important for growth, contact with the Welsh

Development Agency or using the Business Connect service are significantly linked with

10
high growth categories.

Management Characteristics

Policy makers’ growth or employment objectives are not necessarily shared by SME

management, subject to market disciplines and personal preferences. In a competitive

environment, management are obliged to maximise profits in order to survive, which may

not be consistent with growth or employment maximisation. Sometimes this explains

why SMEs are SMEs; they have found profitable niche markets, or management do not

want the extra strain involved in expansion. More congenial to growth-orientation is the

third reason for being an SME, simply that the business started small and is ‘young’ 10

(Harris and Robinson 2001).

Much may depend on managerial time horizons, the willingness to trade-off future for

present profits. This time preference rate might well be reflected in different policies by

age of management. With the passage of time and increasing awareness of finitude,

management may be more inclined to discount the future more heavily, choosing

immediate gains rather than future growth. Another reason for different objectives is that

when competitive pressures are weak, management may prefer to take profits without

making the investment that will both enhance future competitiveness and increase future

output. There is indeed a significant bivariate association between older management

(over 46 years of age) and lower growth category in the present data ( 2 [2] = 32.1

Prob value = .00000).

11
Strategy

Individual SMEs and their management might be distinguished by their strategies, their

emphasis on the policy instruments available to them. A human capital strategy would be

identified if a firm had adopted a training plan, although in each case what counted as a

plan could differ markedly. A separate budget would perhaps be a better indicator of

commitment but the significance would vary with the size of the firm’s turnover.

Adoption of the ISO 9000 standard signals a commitment to ‘quality management’. This

means an organisation’s policies to enhance customer satisfaction by meeting their

requirements and fulfilling regulatory obligations. A customer-orientated firm could be

identified as one with a marketing plan.

Information and Networks

What managers know must influence their strategy and the performance of their business.

Therefore the way they acquire and utilise information will be of vital importance.

Together with the recent rapid development of information and communications

technology, this reflection has stimulated theorising about the firm as an information

network (Bolton and Dewatripont 1994). For efficient information processing and

communication, the boundaries of the firm will be fixed to include activities where

information is most easily codified (Casson and Wadeson 1998)11.

The information approach encourages thinking about the external relations of the

successful firm as much as the internal structure. Managers scan their firm’s

environments and use the information acquired to form a business strategy. Firms and

their managers also communicate across their boundaries with customers in a variety of

12
ways that constitute marketing strategy. Before such a strategy is in place, customers do

not know exactly what the firm can supply them and management are ignorant of

precisely what customers want or how much they are willing to pay.

In the face of this reciprocal asymmetry of information, effective cooperation involves

choosing the most effective protocol for co-operation or networking. The form of this

protocol depends upon communication costs. The higher they are, the less information

will be divulged and therefore the more the supplier must anticipate the buyers needs.

Communication costs depend on management strategies and technology influence. Use of

data bases reduces the firm’s costs of learning customer preferences and CAD/CAM

techniques facilitate customers’ liaison with subcontractors to explore the feasibility of

design changes.

A common presumption has been therefore that, other things being equal, more

networking permits greater information acquisition and communication and so will be

associated with better performance. Before networking terminology assumed its present

prominence, Barkham et al’s (1996) inter-regional regression analysis of turnover growth

of 138 UK SMEs found evidence suggesting the key role of management information.

Membership of professional organisations (undefined) and management interests in other

businesses were significant contributors to growth. In their study of fast growing e-

commerce SMEs Feindt et al (2002) identified one of their general ‘critical success

factors’ as ‘The means of relationship-building with groups of like-minded

individuals/organisations by enabling the exchange of information and

13
services tailored to the needs of the community’. Cooke and Wills (1999)

discovered that ‘…for a sizeable proportion of programme-funded firms in Denmark,

Ireland and Wales (U.K.) social capital building was associated with enhanced business,

knowledge and innovation performance. Of particular importance was the opportunity

afforded to firms for linkage with external innovation networks, and the build-up of

embeddedness, or the institutional basis for the enhancement of social capital.’

Conversely poor performance is blamed on inadequate networking – for instance

Kitching ‘. .. limited networking between small-business owners and training providers in

South London is explained by a lack of embeddedness of UK small engineering firms in

the institutional framework supporting business.’ In France Delapierre et al (1998) found

‘The most successful [technology-based firms] in terms of growth belong to dense and

convergent networks through which they interact with larger firms and research

organisations.’

Networking extends to support industrial districts and learning regions (Morgan 1997)

which consist of successful clusters of SMEs (Piore and Sabel 1984). The Cambridge

high technology sector is often cited in this context (Keeble et al 1999). A spatial

adaption of this line of thought creates a theory of regional technical change and the

‘learning region’ (Lawson and Lorenz 1999).

As some of these studies imply, networking is a very general concept, whereas the theory

presupposes the communication of precise and relevant information through networks. A

distinction might perhaps be drawn between networking for a specific project or purpose

14
on the one hand and serendipity on the other. Enjoying the activities of the Rotary Club

could involve a form of networking yet the time taken from business may not be

productive. Networking may encourage uniformity, perhaps contrary to the

entrepreneurial spirit (Nijkamp 2003 401). Some networking may be even purposely

growth-retarding, as in an informal cartel, bearing in mind Adam Smith’s strictures and

Olson’s (1982) model.

Doubts about the efficacy at the margin of general networking have begun to be

supported by empirical research. A recent survey concluded that there was little evidence

to date on the success of networks, and that informal and specific concern networks were

more likely to be useful (Harris and Robinson 2001 paras 3.16-17). Subsequently

Bougrain (2002) appeared to contradict the Delapierre findings for France, maintaining

that technological co-operation did not seem to increase the chance of success of

innovative projects. Results from Havnes et al (2002) for eight small European countries

in 1991-5 question whether networks are good per se especially for SMEs. Analysis of

panel data ascertained that a substantial number of SMEs were actively networking and

that the level of networking had been maintained over a five-year period. Nonetheless

there was no evidence of associated short-term benefits, such as growth in employment or

growth in total sales, stemming from the networking activities 12. The present data set also

finds no bivariate statistical evidence that networking as measured by membership of

Trade Associations or Business Clubs is associated with growth.

Innovation

Improved information exchange in networks supposedly encourages the adoption of new

15
technology. Hughes (1997) asserted that SMEs that succeeded in growing were more

likely to have introduced product or process innovation. They were also more likely to

have developed networks of collaborative partnerships and faced up to management

development and reorganisation needs as growth proceeded. Unfortunately there are

intrinsic difficulties in forming such networks through purely market relations. These

depend upon the ability to evaluate what is to be exchanged in advance. But in the case of

information the evaluation cannot be made independently of the exchange- once a

potential buyer knows how much a seller’s information is worth, they also know what the

information is, and so need give up nothing in exchange. Because of this difficulty of

creating the property rights in information and knowledge, there may be a ‘market

failure’, a sub-optimal rate of innovation and therefore of cost reduction and/or product

development. For SMEs in England in the 1990s that did innovate there was an

association with growth (North and Smallbone 2000). Consistent with this result, both a

belief in innovation and having introduced a product or process innovation are

characteristics strongly associated with the management of growth firms in the present

Welsh SME sample (respectively chi squared (8 ) 59.4, and (2)19.9 ).

Rapid adoption of information and communication technology (ICT), and the associated

reorganisation, figures prominently among explanations of the acceleration of

productivity growth in the US during the late 1990s (Oliner and Sichel 2000; Goss 2001).

A distinction may also be drawn between defensive and aggressive innovation. The first

may be necessary merely to keep up with the competition and avoid going under. As far

as ICT is concerned, SMEs have been slower than larger firms to engage in on-line

16
transactions, and on-line purchases are more common than on-line sales (OECD 2002 13,

Table A9). This, it is suggested, should be a source of concern. On the other hand, growth

and profitability may be better served by the current slow rate of adoption- the strategy of

being ‘the world’s leading followers’ (Larsen and Rogers 1984).

An evidence-based taxonomy of ICT usage that classifies firm moves between categories

finds that an SME uses customer databases when customer dominance is low (Levy et al.

2001). When the strategic focus is high the SME shifts from ‘coordination’ to

‘innovation’ and high value added with for instance, e-commerce (with low customer

dominance still). For some SMEs motivation for ICT adoption is simply viability. There

is little inclination to grow beyond the size required to ensure an acceptable level of

profitability. In these cases, ICT investment is of limited importance. Only for more

ambitious SMEs is there a requirement for ICT systems as major process innovation - to

control the firm better. The general expectation then is that higher growth firms will in

some way employ more ICT and so the present exercise included a number of ICT

questions. Computer based accounts were strongly associated with high growth

(Chisquared [2] = 44.77 Prob value = .00000), but not with stand alone PCs

(Chisquared [2] = 2.22 Prob value = 0.33) (which were already widespread) or

computerised production control (Chisquared [ 2] = 4.39 Prob value = 0.11).

Skills and Training

Firms survive and grow in a competitive environment if they are good at providing

something the market wants. ‘Competencies’ of firms may thus be the key to growth

17
(Kay 1993). A competency approach suggests that nurturing firm-specific knowledge and

skills and investing in training of the appropriate type will be conducive to growth.

In accordance with this line, Cosh et al (2000) observe that ‘training is positively related

to employment growth, in particular when it is embedded in a wider range of human

relations practices‘. A convenient indicator of training is the ‘Investors in People’

standard. It provides a national framework for improving business performance and

competitiveness, by setting and communicating business objectives, and developing

people to meet these objectives. Supposedly, employees are therefore willing and able to

do what their organisation requires of them. The process is cyclical and is intended to

stimulate continuous improvement13. There is a significant bivariate association of

‘Investors in People’ with high growth ( χ2 [6] = 20.51 Prob value = .00224).

A great deal of other evidence supports the vital role of skills in increasing productivity,

and therefore competitiveness, of individuals and countries as well as of firms. Earnings

equations show the impact of skills by implication on productivity (for example

Chatterjee et al 2003). Skill-biased technical change, together with globalisation, is

widely identified as the cause of large increases in wage inequality in the US and the UK

and the rise in unemployment in continental Europe (Acemoglu 2002). Astute

management may therefore be expected to invest in their employees’ skills to perform

better. Firms in the present sample with a separate training budget were more likely to be

in the higher growth categories ( Chi-squared [2] = 18.20 Prob value = .00011).

18
Legal Form

Among the more fundamental strategies is the choice of legal form for the SME. A

partnership is arguably the most risky because each partner in management is

liable for the mistakes of the other but may have little control over them

(partnership is not significantly associated with growth in the present sample).

The sole trader at least lacks this vulnerability (there is a weak negative

association with growth chisquared 6.23 Prob value =0.044). A subsidiary

might be less risky because of the support that a large organisation can provide.

But headquarters may be as concerned with profitability as the market and

command relations in large organisations can be even more arbitrary. So the

relative risk for subsidiary managers is not immediately apparent. In fact

subsidiaries are significantly associated with growth in the sample (chisquare (2)

9.30 Prob value = 0.0096).

Limited liability reduces the downside risk borne by the owner manager, providing some

protection for personal property. The more risk averse entrepreneur would favour this

type of protection. Leland (1972) shows that a small increase in risk aversion lowers both

production and price14. So if the more risk averse entrepreneurs chose the limited liability

form, other things being equal, the likelihood is that profits would be lower, along with

price and output. Suppliers may be less willing to extend credit to limited liability

companies. Perhaps even banks would seek personal guarantees to circumvent the limited

liability status, if they believed the firm could not offer sufficient collateral. In such cases

19
limited liability would not offer much advantage and so would not be associated with

different behaviour. There is no significant bivariate association with growth in the

sample.

SME characteristic- Clusters

Location is a central management decision but not one that cannot be reviewed

frequently. Access to markets and suppliers depend on transport facilities as well as

proximity. Michael Porter’s (1990) extension of industry economics to the level of the

nation popularised the notion that ‘clusters’ of related firms were necessary for

‘competitive advantage’ and the associated growth of firms. Clusters are somewhat akin

to Alfred Marshall’s emphasis on external economies underlying successful industrial

districts, such as nineteenth century Lancashire with its cotton textiles. Alternatively or

additionally they might be interpreted as the spatial analogue of networking firms.

External economies plus proximity is one reason why location may be a separate

explanatory variable for SME performance; direct measures of the significant spill-overs

are less easy to identify. It is reasonable to suppose that the advantages of location for

retailers would be captured by rents, unless leases are very long. Larger businesses would

expect to pay higher site costs, say, for good access to transport facilities. But their

locational advantages may not be bid away through the land market entirely; some extra

profit might remain after paying the higher costs.

Industrial districts, clusters, agglomeration, Marshallian external economies and

20
networking all suggest that in some circumstances spatial contiguity or distance create

‘growth poles’ or depressed regions. In ‘growth poles’, concentrations of firms, specialist

suppliers and service providers in related industries offer opportunities for firms to

develop their competitive edge and also to benefit from co-operation. By grouping

together they are able to take advantage of specialisation and a supply of qualified

manpower. In such a setting a networked firm might start quite small but grow rapidly

within the network by taking advantage of sub-contracting opportunities.

However it is important to be aware that agglomeration may take place without

agglomeration economies (Parr 2001 726). Firms may cluster in particular areas by

solving similar market access-transport cost problems, rather than because of beneficial

spill-overs between adjacent enterprises. In this case, rather than encouraging networking

between firms in a given area, policy would do better to focus on transport infrastructure

provision. The Welsh economy could be succinctly described as essentially two

motorways from England pointing at the Irish Sea. The bivariate test of the M4 corridor’s

association with growth was consistent with no significant growth advantage (Chi-

squared = 4.30 Prob value = 0.12).

SME Size and Gibrat’s Law

SME size is irrelevant to growth rates according to Gibrat’s Law. Contrary to Gibrat, the

general conclusion has been that small firms grow faster than large (for example

Barkham et al 1996 who restricted themselves to SMEs employing 50 persons or fewer )

although there is some recent evidence for service firms and for some sectors of

21
manufacturing consistent with the law (Audretsch et al 2002; Piergiovanni et al 2002).

Survivor bias may explain the finding that small firms grow faster; those that closed –

and therefore were presumably not fast growers- in the observation period are typically

excluded from studies such as Barkham et al (1996). In any case, a relation between size

and growth, controlling for many other variables is not really the statistical hypothesis

that concerned Gibrat. Rather, a behavioral relationship, such as brought about by

institutional and legislative constraints upon growth, is implied, and such relationships

may be non-linear; over some size ranges constraints may be more binding than over

others. The present sample divided into seven categories of turnover suggests that the

largest firms tend to be in the higher growth categories and conversely for the smaller

(chisquared [14] = 42.75 Prob value = 0 .00009).

Business sector

The final firm characteristic included is one of 8 SIC codes. The small size of Wales’

finance sector has been blamed for laggardly economic performance (Cameron et al

2002) ) but there is only weak bivariate evidence that those few SMEs in the sector are

fast growers (Chi-squared [2] = 4.83 Prob value = 0.09).

4. Multivariate Model Specification

As already noted, bivariate associations are often excessively simply means of identifying

impacts on growth and profitability of SMEs; only multivariate methods permit adequate

controls.

22
In view of the nature of the data, growth is modelled as a latent variable. Instead of

observing the continuous variable growth of turnover, employment or assets, only the

growth categories are available. The boundaries of growth classes are parameters to be

estimated. Other parameters are the impact of the measurable influences as they affect the

probability of a firm being in a particular growth category. Unobservable influences on

growth are assumed to be normally distributed across observations.

Interpretation of the signs of estimated coefficients on the variables discussed in section 2

of this ordered probit model is not always straightforward. An SME with an attribute on

which there is a positive coefficient will for this reason be more likely to be in the high

growth group and less likely to be classified as low growth. The consequences for

probability of being in the medium growth class cannot necessarily be interpreted in this

way. For this reason it is desirable also to estimate explicitly marginal probabilities of

being in a particular class.

The Data Set

Low and selective responses are endemic in survey work. An advantage of the SME

sample of the present paper is the high rate of participation. Participating businesses

expected that co-operation with the Development Agency sponsoring the survey would

yield them dividends. The data was collected as part of a project funded by the European

Regional Development Fund that surveyed a cross-section of the SME population within

South Wales with the intention of identifying existing and potential growth firms.

23
The sample covered a cross-section of firms with between 5 and 250 employees 15,

balanced by size, sector and location16. The original questionnaire was distributed by post

to 4000 businesses. After postal reminders and telephone follow-ups, a total response rate

of 42% was obtained. There are data on 1691 individual firms (but not all questionnaires

were complete) 17.

Selection?

5. Results

There is little difference between the mean profit-turnover ratios for firms in the different

growth categories. The lowest profit/turnover ratio is for the lowest growth

category as expected but there is little difference between this number, 13.3% and

the 14.6% for medium growth. Mean profit/turnover is lower for high growth

firms than for medium growth (Table 5 below). This absence of association

between profitability and growth category is consistent either with managers

trading off present for future profits, through for example their pricing strategies,

or a lack of competitive pressures or a want of SME managerial motivation to

grow, or all of these.

TABLE 5 ABOUT HERE

Determinants of SME Profitability

Since management need long term profits for growth but can choose whether or not to

grow, profitability is the more fundamental performance measure, unless current profits

are traded off against those in the future. The profit-turnover ratio is a continuous variable

24
so that ordinary least squares (OLS) regression can be employed in the analysis. However

the variable is truncated; values are not observed for many SMEs. To avoid estimation

bias the Heckman Two Step method18 was used first. The results demonstrate that there is

no distortion from sample selection; the Heckman approach (not shown) is redundant

because the selection term is insignificant. Re-estimating the model by OLS yields

similar results (Table 6) (6b differs from 6a in the polynomials for numbers of workers to

capture the non-linearity of the relationship).

TABLE 6 ABOUT HERE

From Table 6a a marketing plan appears conducive to profits. Profitable IT strategies

include stand-alone PCs but computerised production control subtract from profits. A

wide range of potential growth drivers exercise no effect on profits. Networking is

ineffective- indeed membership of a business club or a trade association is apparently

associated with significantly lower profitability. Focussing on finance and the professions

(Table 6b) removes the negative effect of trade associations generally on the profit-

turnover ratio. A firm that is in finance or the professions is more profitable unless it is

also in a trade association as well, in which case it is of only approximately average

profitability. However the business club negative effect remains significantly negative at

the five percent level. The miscellaneous sector ‘other’ is the only other one to show

significant differences in profitability. This may stem from the profitability of novelty and

from novelty precluding the SME from falling into the traditional classification scheme.

SMEs that have introduced various types of innovation; technical change or new

organisational structures are no more profitable than those that have not. Skills and

25
training are irrelevant. Firms believing that IT, innovation and networking are important

for their success are no more likely to generate higher profits/turnover than those that do

not. But at least older management does not detract from profitability (results not

shown).

The positive effect of partnership on profitability is twice as large as that of the sole

trader, taking limited liability and subsidiaries as the base case. Since there are no

comparable impacts on SME growth it is reasonable to interpret the effects as risk

premia19. Employing people appears to be bad for current profits, though not uniformly

over the whole range of 250 persons (figure 1 derived from Table 6a). Profit/turnover first

falls up to about 50 employees and then rises, exceeding the profitability of one employee

at about 175. This pattern may reflect the additional administrative costs of employment.

Determinants of SME Growth

As indicated already, explanations of SME current profitability will not necessarily be the

same as those for SME growth. Associations with growth are identified while controlling

for other potential drivers in the ordered probit equation 20. The large number of possible

binary explanatory variables implies that collinearity will preclude precise estimates of

many coefficients in the full specification. However the equation supplies the basis on

which zero restrictions on coefficients can be tested.

The three category (Table 7) results are obtained by taking the most general specification

– including location and sector identifiers- and testing zero restrictions first with t>1 and

26
then with significance levels of 10%.

Use of computerised accounts increases the chances that a firm will have a high rate of

growth and lowers the probability that it will not be growing. Management believing that

innovation is important to the success of the company are also more likely to run ‘high

growth’ SMEs. An SME that actually does innovate or with a commitment to training is

apparently no more likely than others to be a high growth firm- in the first case contrary

to the findings of North and Smallbone (2000) for England. This may reflect a lack of

discrimination in the present data between ‘innovation’ and ‘significant innovation’, for

67 percent of the SMEs in the present sample claimed to have introduced a new product

or process in the last three years. Or the divergence may stem from the greater number of

controls in the present exercise.

Consistent with the expectation that time horizons play a role in SME objectives, firms

with older leaders are less likely to be found in the ‘high growth’ category and more

probably in the ‘no growth’ group. That this is a ‘time preference’ effect is suggested by

the earlier finding that older management does not detract from current profitability.

In Table 7 there is evidence that firms with a marketing plan are more likely to be in the

high growth category, in line with the earlier results of Barkham et al (1996). The sector

effects are restricted to finance.

TABLE 7 ABOUT HERE

SMEs with ambitions to develop skills and training might be expected to operate a simple

27
planning structure that would involve a training plan and a separate training budget. Yet

neither of these two variables is significant. Similarly, companies that are ‘Investors in

People’ do not grow any faster than others. In general, networking is unimportant, but

with one exception. Membership of a trade association significantly reduces the chance

that an SME is in the ‘high growth’ category - consistent with the cartel hypothesis. In

contrast to the profitability result, the impact of trade association membership on growth

category remains negative and significant at the 5 percent level for manufacturing, the

professions and finance interacting and separately 21. Government grants, contact with the

Welsh Development Agency and Business Connect all have insignificant effects.

Business club membership is also not significant, in contrast to the profitability result.

Marginal effects of all variables individually are small.

High turnover is helpful for growth, rather different from pseudo-Gibrat findings of

previous literature. Those managers interested in growth in general may already have

boosted their firms’ size.

Summarising, good for growth is finance (marginal effect on probability of being in high

growth group 8.7%), belief in innovation (2.9%)22, computerised accounts (7.2%), a

marketing plan (5.9%) and a high turnover (6.4%). Bad for growth is management being

over 46 (marginal effect on the probability of being in the low growth group 10.6%), and

being a member of a trade association (5.8%). There are no location effects- growth

clusters are absent.

28
A common element for growth and profitability is a marketing plan. Profitable IT

strategies appear to be different from growth-orientated IT; stand-alone PCs added to

profits but computerised production control subtracted from them. In contrast to growth,

computerised accounts and mobile phones were not helpful for profits. Not surprisingly,

current profitability is beneficial for growth; profit/turnover is significant at 5% in the

growth equation. It has the largest marginal effect of any variable (15.8% for the highest

growth category), but inclusion of the variable radically reduces the sample size (results

not shown).

5. Conclusion

Since marketing plans were a contributor to growth as well as to profits, encouraging

SMEs to think about their customers, current and potential, could well be beneficial.

OECD and other’s beliefs that more IT investment by SMEs would improve their

performance is supported by the results. IT investment- in computerised accounts - was

associated with faster growth category firms, whereas profitability increased with stand-

alone PCs but declined with computerised production control. The more ambitious IT

strategy is pursued by growth firms.

There is evidence in the sample employed that promoting information networking

through trade association membership is bad for SME growth in Wales and business club

membership is bad for profitability. Possibly a finer sectoral classification might show

that slow growth firms happen to be in mature sectors which also turn out to have trade

associations, but this cannot be demonstrated with the current data set. The

ineffectiveness of networking as measured does not mean that all networking is generally

29
of no value for SME growth in Wales. A more specific measure may yield different

results and, as the theoretical discussion suggested, variations in networking may be

determined primarily by the type of firm. But the result does at least raise questions about

the current contribution of certain institutions.

An important policy consideration is that performance preferred by SME management

may not be that desired by policy-makers. A high rate of time preference can encourage a

focus on current profits to the detriment of growth. Consistent with this proposition was

that management over 46 tend to choose lower growth categories. If older managers

showed less vitality there would also have been an adverse ‘age’ effect on current

profitability, but in fact the zero coefficient hypothesis could not be rejected in the profit

equations. Policy to reduce the short-termism implied by these results might be to address

tax disincentives to building up a ‘business dynasty’; at present the findings imply that

older management have little interest in future profits, presumably because they

themselves may not be around to enjoy them and any capitalised future profits available

to inheritors would be substantially reduced by taxation.

Profitability was strongly determined by firm type and profession whereas growth was

not- consistent with risk aversion for choice of type. There may be greater scope for

application of competition law where the professions are concerned. SMEs in finance

were growing faster than others but there were apparently neither growth nor profitability

effects of location. Mobility of resources between industries, the market working well,

could well be the explanation, with the professions being an exception.

The turnover-growth relationship together with the lack of profit in employment suggests

a threshold difficulty of employment. Once this threshold is overcome, larger firms grow

30
more. One of the most surprising findings was the lack of association with either growth

category or profitability of skills and training, indicative perhaps of the ineffectiveness of

policy initiatives such as Investors in People. Similarly of concern is that actual

innovation shows no (multivariate) association with growth or profitability in the present

sample. These results may stem from shortcomings in the measures but again there is a

prima facie case for improved policy here. The belief that innovation is important does

come through strongly as a growth driver, and therefore may serve as a means of

identifying in advance high growth SMEs.

31
References
Acemoglu, D (2002) Technical Change, Inequality, and the Labor Market Journal of

Economic Literature, 40, 1, 7-72

Audretsch, D B, Klomp, L, Thurik, A R, (2002) Gibrat's Law, Erasmus Research

Institute of Management (ERIM), Erasmus University Rotterdam Discussion Paper 150

Barkham R, Gudgin G, Hart M and Hanvey E (1996) The Determinants of Small Firm

Growth: An Inter-regional Study in the United Kingdom 1986-1990, Jessica Kingsley

London

Bennett, R J. and Robson, P J. A (2003) ‘Business Link: Use, Satisfaction and the

Influence of Local Governance Regime’ Policy Studies, December v. 24, iss. 4, pp. 163-

86

Bennett, R. J., Robson, P. J. A. and Bratton, W. J. A., (2001) ‘Government Advice

Networks for SMEs: An Assessment of the Influence of Local Context on Business Link

Use, Impact and Satisfaction’, Applied Economics, June, v. 33, iss. 7, pp. 871-85

Bolton P and Dewatripont M (1994) The Firm as a Communication Network, Quarterly

Journal of Economics 109 4 809-839

Bougrain, F, and Haudeville, B (2002) Innovation, Collaboration and SMEs Internal

Research Capacities, Research Policy, 31, iss. 5, pp. 735-47

Cameron G., Muellbauer J. and Snicker J. (2002) ‘A Study in Structural Change: Relative

Earnings in Wales since the 1970s’ Regional Studies 36, 1–11

Casson M and Wadeson N (1998) Communication Costs and the Boundaries of the Firm ,

International Journal of the Economics of Business 5 1 5-27

32
Chatterji, M, Seaman, P T.; Singell, L D., Jr. (2003) A Test of the Signalling Hypothesis

Oxford Economic Papers, 2003, 55, 2, 191-215

Coase R (2001) The Nature of the Firm, in Baumol, W. J.and Wilson, C A., eds Elgar

Reference Collection. International Library of Critical Writings in Economics, vol. 126.

Cheltenham, U.K. and Northampton, Mass.: Elgar.

Cooke, Philip and Wills, David, (1999) Small Firms, Social Capital and the

Enhancement of Business Performance through Innovation Programmes, Small Business

Economics, 13, iss. 3, pp. 219-34

Cosh A Hughes A Weeks M (2000) The Relationship Between Training and Employment

Growth in Small and Medium-Sized Enterprises, Cambridge University, Centre for

Business Research WP188

Cowling K and Waterson M (1976) Price-Cost Margins and Market Structure,

Economica 43 267-74

De Meza D (2002) Overlending? Economic Journal 112 F17-F31

Delapierre, M, Madeuf, B, and Savoy, A. (1998) NTBFs--The French Case, Research

Policy, 26, iss. 9, pp. 989-1003

Disney, R., J. Haskel and Y. Heden, (2003) ‘Restructuring and Productivity Growth in

UK Manufacturing’, Economic Journal 113 July 666-694

Feindt S, Jeffcoate, J, Chappell C (2002) Identifying Success Factors for Rapid Growth

in SME E-commerce, Small Business Economics 19: 51–62

Gillespie G, McGregor P, Swales J, Ping Yin Y, (2001) ‘The displacement and multiplier

effects of regional selective assistance A computable general equilibrium analysis’

Regional Studies 35 2 125-139

33
Goss, E (2001) The Internet's Contribution to U.S. Productivity Growth, Business
Economics, 36, 4, pp. 32-42
Hansford, A, Hasseldine, J and Howorth, C (2003) ‘Factors Affecting the Costs of UK
VAT Compliance for Small and Medium-Sized Enterprises’ Environment and Planning
C: Government and Policy, August v. 21, iss. 4, pp. 479-92
Harris R and Robinson C (2001) DTI Industrial Support Policies: A critical review of

empirical research on hindrances to business development and productivity growth and

the relative importance of different constraints on UK business DTI Ref: SEC Research

01 ITT No. SEC01

Havnes P-A and Senneseth K (2001) A Panel Study of Firm Growth among SMEs in

Networks, Small Business Economics 16: 293–302

Hughes A (1997) ‘Small Firms and Employment‘ WP 71 ESRC Centre for Business

Research, University of Cambridge

Kay J (1993) Foundations of corporate success : how business strategies add value

Oxford

Keeble D and Wilkinson F (1999) Collective Learning in regionally clustered high

technology SMEs in Europe, Regional Studies 33 295-303

Kitching, J.; Blackburn, R. (1999) Management Training and Networking in Small and

Medium-Sized Enterprises in Three European Regions: Implications for Business

Support, Environment and Planning C: Government and Policy, 17, 5 621-35

Larsen E M and Rogers J K (1984) Silicon Valley Fever: Growth of High Technology

Culture New York Basic Books

Lawson C and Lorenz E (1999) ‘Collective learning tacit knowledge and regional

innovative capacity’ Regional Studies 33 305-18

34
Leland HE (1972) ‘Theory of the Firm facing Uncertain Demand’ American Economic

Review 62 3 278-291

Levy, M, Powell, P, Yetton P (2002) The Dynamics of SME Information Systems, Small

Business Economics 19: 341–354, 2002.

Morgan K (1997) The Learning Region: Institutions, Innovation and Regional Renewal

Regional Studies 31 491-503

Nickell S J (1996) ‘Competition and Corporate Performance’ Journal of Political

Economy 104 723-746

Nijkampf P (2003) Entrepreneurship in a Modern Network Economy Regional Studies

37 4 June 395-405

North, D and Smallbone, D. (2000) The Innovativeness and Growth of Rural SMEs

during the 1990s, Regional Studies, April, v. 34, iss. 2, pp. 145-57

OECD (2002) OECD Small and Medium Enterprise Outlook , Paris OECD

Oliner S D and Sichel D E (2000) The Resurgence of Growth in the Late 1990s: Is

Information Technology the Story? Journal of Economic Perspectives 14, 4, pp. 3-22

Ollinger, M and Fernandez-Cornejo, J (1998) ‘Sunk Costs and Regulation in the U.S.

Pesticide Industry ‘ International Journal of Industrial Organization, March 1998, v. 16,

iss. 2, pp. 139-68

Olson M (1982) The Rise and Decline of Nations : Economic Growth, Stagflation, and

Social Rigidities New Haven, Yale UP

Parr J B (2002) ‘Agglomeration economies: ambiguities and confusions’ Environment

and Planning A 34 717-731

Piergiovanni, R, Santarelli, E, Klomp, L Thurik AR (2002) Gibrat's Law and the Firm

35
Size / Firm Growth Relationship in Italian Services, Tinbergen Institute Discussion

Papers 02-080/3 / Tinbergen Institute

Piore M and Sabel C (1984) The Second Industrial Divide: Possibilities for Prosperity,

Basic

Porter M (1990) The Competitive Advantage of Nations Macmillan

Reid G C and Smith JA (2000) ‘What Makes a new Business Start-Up Successful’ Small

Business Economics 14 165-182

Robson, P. J. A and Bennett, R. J. (1999) ‘Central Government Support to SMEs

Compared to Business Link, Business Connect and Business Shop and the Prospects for

the Small Business Service’, Regional Studies, November, v. 33, iss. 8, pp. 779-87

Rodriguez-Pose A and Fratesi U (2004) ‘Between Development and Social Policies: The

Impact of European Structural Funds in Objective 1 Regions’ Regional Studies February

Vol. 38.1, pp. 97–113.

Sandford, C., Godwin, M. and Hardwick, P. (1989) Administrative and Compliance

Costs of Taxation Fiscal Publications, Bath

Smith, J A (1999) ‘The Behaviour and Performance of Young Micro Firms: Evidence

from Businesses in Scotland’, Small Business Economics 13 185-200

Söderbom, M and Pattillo C (2002) ‘Managerial Risk Attitudes and Firm Performance in

Ghanaian Manufacturing: An Empirical Analysis Based on Experimental Data’ CSAE-

UNIDO Working Paper No. 3

Storey D J (1994) Understanding the Small Business Sector, Routledge

36
Table 1 Employment, Output and Relative Labour Productivity in Manufacturing
Firms Employing under 250 Persons c 1999, Selected Countries
Employment Output % Relative labour
% productivity %
Spain 73.4 54.1 73.7
Italy 72.3 61.4 84.9
Japan 69.3 47.9 69.1
UK 53.0 37.4 70.6
France 52.7 40.2 76.3
Australia 52.3 40.2 76.9
Germany 45.5 31.4 69.0
Sweden 44.5 34.2 76.8
Wales* 40.5 29.5 72.8
Source: Calculated from OECD 2002 and DTI Small Business Service Statistics Table 20
Note:* Includes mining and quarrying and Gas Water Electricity, firms employing under 200 persons, 2001

Table 2 All Industry Employment, Turnover and Relative Labour Productivity

in Firms Employing under 250 in the United Kingdom and Devolved
Administrations in 2001
Employment Output % Relative labour
% productivity %
UK 55.4 51.4 92.8
England 54.2 50.4 93.0
Highest % English region 64.8(SW) 55.7(NW) n.a.
Wales 67.1 61.4 91.5
Scotland 57.9 56.0 96.7
Northern Ireland 79.9 74.5 93.2
Source DTI Small Business Service Statistics Table 8

Table 3a Growth Categories and Subsequent Growth (%) of Turnover and

Employment (sample)
Growth Category Mean Turnover Growth 1998-2001 Mean Jobs Growth 1998-2001
‘Low’ (and No) -0.3 -15.4
‘Medium’ 2.0 5.9
‘High’ 14.0 30.3

Table 3b Growth Categories and Previous Turnover Growth % (sample) 
Growth Category Mean Growth Cases
‘Low’ (and ‘No’) -6.63 84
‘Medium’ 4.96 114
‘High’ 29.62 50

37
 Table 4: Association of selected potential growth indicators with SME growth
category

Growth driver Measure Chi-squared probability %

(d.f.)
Business PAYE and VAT 15.55 (8) 0.05 n.a
Environment problematic
Government 3.06 (2) 0.22 18
grants important
Welsh 1.80 (2) 0.41 43
Development
Agency contact
Business 0.12 (2) 0.94 7
Connect
Management Management age 32.10 (2) 0.00 56
Characteristics over 46
Strategy Marketing plan 46.10 (2) 0.0000 44
a) Customer
orientation
b) Information Business Club 1.67 (2) 0.44 23
Networking membership
Trade 0.72 (2) 0.70 42
Association
membership
c) innovation Innovation 59.3 (8) 0.00 n.a.
important
Introduced new 19.86 (2) 0.00 67
product or
process in last
three years
d) ICT Computer based 44.77 (2) 0.00 80
adoption accounts
Stand alone PCs 2.22 (2) 0.33 84
Computerised 4.39 (2) 0.11 56
production
control
e) Skills and ‘Investors in 20.51 (6) 0.002 9
training People’
Separate training 18.20 (2) 0.0001 27
budget
f) Legal form Sole trader 6.23 (2) -ve 0.04 11

Subsidiary 9.30 (2) 0.009 7

Limited liability 3.78 (2) 0.15 65

SME Characteristics M4 Corridor 4.30 (2) 0.11 50

a) Clusters location
b) Size Turnover 42.75 (14) 0.0000 n.a

38
Note: n.a signifies multipoint scale employed

39
 Table 5 SME Management Subjective Growth Assessment and Profit Performance
Profit/Turnover
Subjective growth category Mean Cases
‘High’ 14.0 106
‘Medium’ 14.6 309
‘Low’ and ‘No’ 13.3 199

40
Table 6a OLS Regression, Dependent Variable Profit/Turnover
|
Variable Coefficient |b/St.Er.| P[|Z|>z]
Constant 0.1424 8.496 .0000
Market plan 0.0245 2.382 .0172
PCs   0.0330 2.555 .0106
Computer prodn control -0.0230 -1.903 .0570
Business Club -0.0216 -2.002 .0453
Trade Association -0.0195 -2.074 .0380
Partner -.0.1183 7.512 .0000
Sole trader   -0.0600 -3.045 .0023
Firm type missing -0.0703 -3.476 .0005
Number Workers  -0.0044 -3.790 .0002
Workers squared 0.704-04 3.041 .0024
Workers cubed  -0.404E-06 -2.637 .0084
Workers4 0.75E-09 2.377 .0175
Profession     0.0535 4.173 .0000
Other (sector) -0.9763 -5.806 .0000
SIC missing -0.0464 -1.926 .0541
Adjusted R-squared = .2447
Results corrected for heteroskedasticity Breusch - Pagan chi-squared = 119.5, with 15 degrees of freedom
Observations = 645
Diagnostic: Log-L = 442.2, Restricted(b=0) Log-L = 344.1

Table 6b
Variable Coefficient b/St.Er. |P[|Z|>z]
Constant 0.1308 7.629 .0000
Market plan 2.282 .0225 0.0230
PCs   0.0332 2.590 .0096
Computer prodn control -0.02489 -2.089 .0367
Business Club -0.01991 -1.874 .0609
Trade Association 0 .00108 0.103 .9178
Partner 0.1196 7.687 .0000
Sole trader   0.06118 3.220 .0013
Firm type missing -0.0655 -4.667 .0000
Number Workers  -0.0042 -3.674 .0002
Workers squared 0.681E-04 2.960 .0031
Workers cubed  -0.393E-06 -2.583 .0098
Workers4 0.737E-09 2.343 .0191
Profession     0.08211 4.415 .0000
Other (sector) -0.0894 -5.304 .0000
SIC missing -0.0443 -1.802 .0715
Finance 0.0912 1.941 .0523
Finance*Trade Association -0.1171 -2.099 .0359
Profession*Trade Association -0.0629 -2.685 .0072
Adjusted R-squared= 0.2557
Results corrected for heteroskedasticity. Breusch - Pagan chi-squared = 124.4, with 18 degrees of freedom
Observations 645
Diagnostic: Log-L = 448.5, Restricted(b=0) Log-L = 344.1

41
Table 7 Ordered Probit Model, Growth Category (3)
Variable Coeff.t b/St.Er. P[|Z|>z
Constant -.2643 -2.357 .0184
Marketplan .2418 3.953 .0001
Computerised accounts .2954 3.931 .0001
Innovation important .1212 5.444 .0000
Trade Association -.1581 -2.649 .0081
Age46Plus -.2889 -4.979 .0000
Finance .3578 2.124 .0336
Highest turnover .2647 3.604 .0003
Turnover missing .2516 3.424 .0006
Workers missing -.1238 -1.977 .0480
Number of observations 1570
Log likelihood function -1534.400
Restricted log likelihood -1610.930
Chi-squared 153.0606
Threshold parameters for index
Mu( 1) 1.3993 32.746 .0000
                 +­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­­+
Marginal Effects for Ordered Probit
+----------+----------+----------+----------+
Variable | GROW3=0 | GROW3=1 | GROW3=2
+----------+----------+----------+----------+
Constant | .0972 | -.0330 | -.0642 |
Marketing plan | -.0890 | .0302 | .0587 |
Computerised accounts  -.1087| .0369 | .0717 |
Innovation important | -.0446 | .0152 | .0294 |
Trade Association | .0582 | -.0198 | -.0384 |
Age46Plus | .1063 | -.0361 | -.0701 |
Finance | -.1316 | .0447 | .0869
Highest turnover | -.0974 | .0331 | .0643
Turnover missing | -.0926 | .0315 | .0611 |
Workers missing | .0456 | -.0155 | -.0301
                 +­­­­­­­­­­+­­­­­­­­­­+­­­­­­­­­­+­­­­­­­­­­+

42
Profit/Turnover
Fig 1. Welsh SME Profitability and
Employment

0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
0 100 200 300
Employment

43
*
Acknowledgements to EDF and WDA, Julia Smith, Porto participants, anonymous referees.
1
Defined here as businesses employing 250 persons or fewer.
2
For example in 2000 Ireland launched the Empower initiative for e-commerce in small businesses, in Wales the
Information Society Awareness programme has demonstrated new technologies to more than 10,000 firms and Italy is
planning a project for to promote networks of firms using e-commerce.
3
Over 620,000 clients are helped annually in the United States’ 1000 small business development centres.
4
Business incubators usually comprise a physical workspace and advisory facilities. France has over 200.
5
These include Australia’s Innovation Investment Fund, Germany’s Venture Capital for Small Technology Companies
programme and the United Kingdom’s Regional Venture Capital Funds.
6
Even so Wales had 678 SMEs per 10,000 of the adult population in 1997 and 656 in 2001. The corresponding figures for
the UK as a whole are 791 and 785.
7
Aggregating a number of such indices therefore introduces an element of arbitrariness into the performance measure.
8
Management may also make mistakes of judgement or adopt different standards of , say, what constitutes ‘medium
growth’. This will reduce the explanatory of any growth equation and the precision with which coefficients can be
estimated, but does not invalidate the exercise.
9
Since respondents were not offered a negative growth category, they cannot be convicted of ‘over-optimism’.
10
Being ‘young’ in this context might mean as old as 20 years. A ‘micro’ firm, beginning with five employees and achieving
an average of 20% per annum growth in employment would need to maintain this average for 22 years before ceasing to
satisfy the definition of an SME, that is exceeding employment of 250 persons. After 21 years, employment would be
5*(1.2)21=230.
11
When the flow of information managers must process is too large, there is information overload.
12
The analyses suggested, however, that networking was associated with high growth in the geographic extension of
markets.
13
The Standard is based on four key principles: commitment to invest in people to achieve business goals, planning how
skills, individuals and teams are to be developed to achieve these goals, taking action to develop and use necessary skills in
a well defined and continuing programme directly tied to business objectives, and evaluating outcomes of training and
development for individuals' progress towards goals, the value achieved and future needs. These principles are broken down
into 12 indicators, against which organisations wishing to be recognised as an 'Investor in People' are assessed.
14
When demand is additive in price and the source of uncertainty and marginal costs are non decreasing.
15
The data includes other size firms because the sampling frame was not perfect.
16
The target was about 41% of the sample in manufacturing and 32% in the professional and financial sectors with about
1% in each of Agriculture and Fishing and Mining and Quarrying. Most firms should have come from Cardiff and the Vale
of Glamorgan (30%), Newport (11%) and Swansea (12%).
17
And in the present study businesses in the non-trading sector are omitted because they may pursue different objectives
from the others.
18
The Heckman method obtains consistent regression equation coefficients when there are missing observations on the
dependent variables. The problem is similar to that arising from the effects of an incorrectly omitted explanatory variable. It
is addressed by a selection equation to model why the data are missing; the selection equation computes a sample selection
correction term (normally called lambda). This term is added to list of regressors in the original regression equation.
Standard regression techniques applied to the augmented equation then provide consistent estimates of the regression
coefficients.
19
Another explanation would be differences in accounting practices but there seems no reason why there should be such
differences.
20
Appendices giving the full statistical results are available on request from the authors.
21
The coefficients on ‘profession’ and ‘manufacturing’ sectors in this specification were respectively -0.05 (SE=0.08) and
-0.02 (SE=0.07), while the negative coefficient on ‘trade association’ remained significant at the one percent level. A
‘manufacturing* trade association’ variable was also not significant. Some combinations of sector and trade associations
could not be estimated because of multicollinearity.
22
Replacing the Likert scale variable with a binary measure of zero for ‘innovation not important’, otherwise ‘1’, raises the
marginal probability of being in the high growth category to 6.4 percent. This modified ‘innovation important’ variable
remains significant at the one percent level.