The impact of employee involvement on small firms' financial performance.

Bryson, Alex

Introduction

Most firms operating in Britain have some form of employee involvement (EI), that is, an employee relations practice which management believe encourages employees' commitment to managerial goals and the success of the enterprise. Recent empirical research indicates that different El practices have very different effects on performance, and that these effects can vary when El practices exist in combination (Ichniowski et al., 1997; McNabb and Whitfield, 1998).

But there are few studies analysing EI's impact on small firms' performance. Consequently, it is only possible to infer what that impact might be from studies based on larger firms and firms in general. This is hazardous because there are theoretical reasons to suggest that the performance effects of EI may differ according to establishment and firm size.

This study begins to fill this gap in the empirical literature by exploring the impact of EI on small firms' performance in Britain, comparing it to the impact on larger firms. We focus on two issues.

(i) How does EI's impact on financial performance differ by firm size?

(ii) What combinations of EI practices have significant effects on small firms' performance?

In addressing these questions we also consider whether the effects of EI can be explained, at least in part, by the nature of small firms that practice EI.

Theory and hypotheses

In seeking to explain a link between EI and firm performance, the empirical literature draws on emerging but limited theoretical propositions. However, most authors identify improved labour productivity as the main mechanism by which El affects performance. This occurs if EI elicits greater discretionary effort by increasing workers' intrinsic rewards from working, or by raising organisational commitment. There is empirical support for the proposition that El raises worker effort (Green and Mcintosh, 1996). Performance-related payments may elicit greater effort by linking effort and reward more closely (Mitchell et al., 1990). Higher productivity may also come about when El alters organisational structures in such a way as to enhance employees' ability to control how their own roles are performed. This can provide employees with opportunities for skill enhancement which raise the firm's human capital stock, and improve task efficiency by drawing on workers' tacit skills and knowledge. But EI can also affect productivity through its impact on management, offering managers greater control over the production process through improved information flows which enhance management's job knowledge. Through upward-problem solving, EI may contribute to a more productive reallocation of jobs and job redesign. Finally, financial participation schemes can reduce shirking and thus the resources devoted to monitoring effort, so reducing the cost per unit of output (Marsden and Momigliano, 1995). Table 1. Mean values on independent variables for small firms and large firms Variables Small Large firms firms Establishment characteristics Size of establishment ESTSIZ1 (0,1): under 50 employees 0.63 0.49 ESTSIZ2 (0,1): 50-99 employees 0.28 0.27 ESTSIZ3 (0,1): 100-199 employees 0.09 0.13 ESTSIZ4 (0,1): 200-499 employees - 0.10 ESTSIZ5 (0,1): 500+ employees - 0.02 EMPFALL1 (0,1): employees fell by 5%+ in last year 0.20 0.18 EMPRISE1 (0,1): employees rose by 5%+ in last year 0.47 0.38 EMPSAME1 (0,1): emp'ees stable in last year 0.33 0.43 Size of firm SIZ1 (0,1): under 100 employees 0.80 - SIZ2 (0,1): 100-199 employees 0.20 - SIZ3 (0,1): 200-499 employees - 0.15 SIZ4 (0,1): 500-1999 employees - 0.16 SIZ5 (0,1): 2000-9999 employees - 0.17 SIZ6 (0,1): 10000-49999 employees - 0.26 SIZ7 (0,1): 50000+ employees - 0.26 Industrial sector SIC1 (0,1): energy, water supply 0.00 0.02 SIC2 (0,1): extraction etc 0.03 0.05 SIC3 (0,1): metal goods etc 0.14 0.09 SIC4 (0,1): other manufacturing industries 0.16 0.10 SIC5 (0,1): construction 0.11 0.02 SIC6 (0,1): distribution, hotels and catering 0.21 0.44 SIC7 (0,1): transport, communication 0.08 0.04 SIC8 (0,1): banking etc 0.15 0.20 SIC9 (0,1): other services 0.12 0.04 LMKTSHAR (-.5.83,3.77): log of estab's share of 4-digit SIC employment -3.16 -3.10 Age AGE05 (0,1): 0-5 years 0.18 0.20 AGE610 (0,1): 6-10 yrs 0.25 0.13 AGE1120 (0,1): 11-20 yrs 0.18 0.19 AGE21P (0,1): over 20 yrs 0.38 0.48 MULTI (0,1): if multiple-establishment organisation 0.27 0.98 COMPANY (0,1): if company 0.75 0.92 UKOWNED (0,1): if UK owned 0.90 0.92 HITECH (0,1): microelectronics used in design, control, handling, or storage 0.40 0.49 Nature of the product market Number of competitors NOCOMP (0,1): dominates the market 0.05 0.10 FEWCOMP (0,1): 5 or fewer competitors 0.29 0.28 MANYCOMP (0,1): more than 5 competitors 0.66 0.62 Location of market LOCAL (0,1): primarily local 0.30 0.49 REGIONAL (0,1): primarily regional 0.20 0.14 NATIONAL (0,1): primarily national 0.35 0.27 INTERNAT (0,1): primarily international 0.14 0.10 Product diversification SINGPROD (0,1): produces a single product/service 0.52 0.40 PROD25PC (0,1): main product/service accounts for 25%+ of sales 0.40 0.38 DIFPRODS (0,1): no single product/service accounts for 25%+ of sales 0.08 0.22 Workforce characteristics PCMAN (0,100): % manual workers 61.4 51.8 PCFT (2,100): % working full-time 81.2 77.6 PCUNSKIL (0,99): % unskilled workers 22.9 18.6 Internal procedures/structures ONBOARD (0,1): if someone with personnel responsibilities on the board 0.54 0.76 SPECIAL (0,1): if a personnel specialist 0.03 0.21 MERITPAY (0,1): if any merit pay 0.33 0.49 INDIVPBR (0,1): if any individual payments by results 0.27 0.28 Unionisation UNIONREC (0,1): at least one recognised union 0.24 0.50 UREC2PLS (0,1): 2+ recognised unions 0.07 0.17 USTRONG (0,1): if closed shop exists, or management strongly recommends union membership or 80%+ workers covered by collective bargaining 0.13 0.22 Information provision INFOPROD (0,1): management provide info. on productivity at least annually 0.14 0.26 INFOWAGE (0,1): management provide info. on wage costs at least annually 0.20 0.33 FININFO (0,1): if management gives employees any financial information 0.42 0.82 LOTFINFO (0,1): if management gives employees a lot of financial information 0.15 0.50 Sample sizes weighted 351 354 unweighted 178 386 Note: Values are based on weighted data

Whether any of these effects actually feed through to better firm-level performance depends on the costs of El adoption and maintenance (which may outweigh any benefits), and offsetting negative incentives for some groups of workers (Mitchell et al., 1990; Lazear and Rosen, 1981). Furthermore, alternative theories offer opposing predictions as to the impact of different types of EI practice. Much depends on the nature of the scheme, how it is introduced, and prevailing conditions within and beyond the firm (Levine and D'Andrea Tyson, 1990).

But there are strong theoretical grounds for believing that the effects of EI on performance will differ by firm size. These form the basis for the following hypotheses.

First hypothesis: the impact of EI practices differs by firm size

The extent to which a particular EI practice can improve performance depends on a range of structural, technological, procedural and market characteristics that are internal and external to the firm (Kessler and Purcell, 1992; Huselid, 1995). Its actual impact will depend on the degree to which it 'fits' with other human resource management practices, on the one hand, and its 'fit' with the firm's competitive strategy on the other (Baird and Meshoulam, 1988).

According to Huselid (1995: 643): "The external fit perspective raises the conceptual issue of whether any particular human resources policy can be described as a best practice, or whether, instead, the efficacy of any practices can only be determined in the context of a particular firm's strategic and environmental contingencies."

Although each firm faces a unique set of internal and external conditions, those facing small firms differ markedly from those faced by larger firms, such that one might reasonably expect the El practices 'fitting' the small firms environment to be quite different from those appropriate for larger firms. (See Table 1 for differences between small and large firms on variables used in the analyses).

The costs in maintaining effective El: The relative cost of adopting and maintaining more formal, sophisticated El practices is greater for small firms. Larger firms benefit from economies of scale since the adoption/maintenance cost per worker is lower, all other things being equal. Large-firm establishments also tend to benefit from the firm-wide support systems and financial resources available in multiple-establishment organisations. These can facilitate the smooth running of complex El practices while personnel specialists can ensure that a scheme is adequately maintained and fits with other policies and procedures. Where these are absent, management often involves "informal routinisation" (Scott et al., 1989) characterised by ad hoc responses to short-term exigencies. In consequence, less formal and adaptive practices survive and flourish in small firms, whereas more formal procedures are frequently undermined by short-term operational pressures (Marchington et al., 1993). So more formal El practices may 'fit' better in a large firm environment than in a small firm environment. The overall effectiveness of such schemes may be impaired in small firms where they do exist; this is less likely to be the case for simpler EI practices with minimal cost implications.

The nature of employee relations: Increasing trust in management can engender organisational commitment, leading to better employee relations and higher productivity (Hammer, 1988). In turn, this may feed through to improved financial performance. But the 'trust building' process is different in small firms where the management chain is shorter and employee relations are more often conducted face-to-face rather than through representation. This closer proximity to the decision-making process may lead to higher levels of employee satisfaction in small firms as compared with large firms, resulting in higher organisational commitment. But, equally, if this opportunity is spurned by small firm managers, and decision-making takes little account of employee input, or information exchange is minimal, the potential for demotivating staff is high.

There are no clear predictions from these observations as to the likely effect of consultative committees and union representation on small firms' performance where they do exist. Much will depend upon why the union or committee originally started operating at the firm, and the precise function it: performs.

Impact of firm size on financial incentives: With n workers, the extra income-sharing reward associated with a single worker's marginal effort is diluted by a factor of 1/n (Weitzman and Kruse, 1990) so that, other things being equal, the incentive effects of income-sharing are greater in small firms. The net effect of financial participation on performance will depend on whether these effects outweigh the greater costs small firms face in adopting and maintaining such schemes.

Individual performance-related payments, unlike financial participation schemes, are not treated in the literature as a form of EI. However, they may have a greater impact on small firm performance than group-based incentives. Wage responsiveness to individual-level productivity, which is a key factor in determining the impact of incentive schemes on performance, is greater in small-firm establishments than large-firm establishments (Bishop, 1987). Higher monitoring costs and more firm-specific skill differentials mean large-establishment employers are less inclined to compensate more productive workers fully.

The countervailing influences of financial incentives upon small firm performance mean that it is difficult to predict from theory what the net effect will be.

Second hypothesis: combinations of EI practices differ from isolated EI practices in their effect on small firms' performance

Recent research indicates that, although individual EI practices can be effective in enhancing performance, it is the combination of practices and, in particular, the combination of financial participation and substantial forms of non-financial EI, which has the most powerful effects on performance (e.g. McNabb and Whitfield, 1996; Cooke, 1994). These findings are consistent with Benner and Jones' (1995) proposition that combining decision-making rights ('control' rights) and financial rewards ('return' rights) can ensure the net benefits for individuals are positive.

Where there is no link between worker involvement in decision-making and financial consequences for individuals, workers may be less careful in reaching decisions which affect surplus and may prefer to prioritise decisions that improve their own working conditions. However, as Ben-ner and Jones (1995) hypothesise, divorcing 'control' and 'return' rights may be less problematic in small firms where continued employment is more easily understood to be contingent on the financial well-being of the organisation, and where 'bail out' funds may be harder to come by. In these circumstances workers may be less inclined to jeopardise their employment by pursuing their own interests.

Even if financial incentives improve workers' preparedness to participate constructively in EI, these incentives do not need to take the form of group or firm-level financial participation. Indeed, individual-level performance payments may offer more powerful incentives in small firms (see above).

Furthermore, there are no theoretical grounds for supposing that the pay-off which workers may want in return for constructive participation in EI should be confined to financial incentives. Concerns a bout job satisfaction, job security, and 'voice' in the workplace may be equally significant and, perhaps, more so in small firms where workers are less likely to be part of a financial participation scheme, and where employment prospects are more closely tied to short-term fluctuations in the firm's fortunes.(1)

Small-firm workers may be motivated by different intrinsic rewards compared with large-firm workers. Ingham (1970) argues that, relative to large-firm workers, small-firm workers place a lower emphasis on economic rewards, and a high emphasis on non-economic rewards such as satisfying interpersonal relations.(2) We might therefore expect small-firm workers to be more responsive than large-firm workers to El practices which improve interpersonal relations and the general working environment, such as improved direct communication and consultation.

Where workers feel they have a voice which is taken into account by managers they may have greater trust and confidence in the decision-making process (Freeman and Medoff, 1984).(3) Conversely, where schemes are imposed on employees by management, with little or no discussion, in the hope of tapping into worker knowledge to improve productivity, and without enhanced and meaningful participation in other respects, they can damage performance.(4) These propositions hold for small and large firms alike, although the failure to communicate and consult may be particularly damaging in small firms if workers join small firms expecting better treatment, or if they are faced by managerial diktat which often accompanies the unitarist managerial perspective adopted by owner-managers (Rainnie, 1989; Scott et al., 1989).

The data

We use data from the management interview of the 1990 Workplace Industrial Relations Survey (WIRS) to test the above hypotheses concerning El and financial performance.(5) It is one in a series of nationally representative surveys of British establishments with 25 or more employees. WIRS 1990 has a high response rate (84 per cent) and contains a rich variety of information on EI practices and other relevant factors, making it ideally suited to testing our hypotheses.

The sample

The unit of analysis is the firm, meaning the whole of an organisation engaged in economic activity under common ownership. Although WIRS is a nationally representative sample of establishments, not firms, it contains data on the number of employees in the firm, as well as the establishment, so that we are able to distinguish small-firm establishments and large-firm establishments.

We define small-firm establishments as those belonging to firms with less than 200 employees. WIRS does not contain establishments with fewer than 25 employees, so the small firm size band is 25-199 employees.(6)

Measure of financial performance

The analyses estimate the association between EI and financial performance. WIRS asked managers: "How would you assess the financial performance of this establishment compared with other establishments and firms in the same industry?", with responses coded on a five-point scale. The distribution of responses for the sub-sample used in our analyses is given in Table 2. Table 2. Managerial assessment of financial performance Small Large All firms firms firms A lot above average 23.7 26.4 25.0 A little above average 27.4 38.7 33.1 About average 41.1 30.5 35.8 A little below average 6.1 3.2 4.6 A lot below average 1.7 1.3 1.5 Number of establishments weighted 351 354 705 unweighted 178 386 564 Note: Figures are column percentages.

Although there are disadvantages in relying on manager evaluations of their establishment's performance, 'objective' performance data are rarely available at establishment level. Furthermore, as long as there is no systematic bias in self-reporting and reporting errors are not related to explanatory variables, then qualitative measures of this sort provide unbiased estimates of performance. Diagnostic tests have indicated that the subjective measure of financial performance used in the WIRS series is a satisfactory indicator of financial performance at establishment level (Machin and Stewart, 1990; Machin and Stewart, 1996).

Measuring employee involvement

Ideally, one would wish to establish the impact of El on performance using measures of its incidence, degree of usage and coverage. The strength of WIRS for our current purpose lies in its measurement of the incidence of a wide range of El practices (Table 3),(7) but there is little information on El usage or coverage.

The El typology is similar to that devised by Marchington et al., (1992) and adopted in previous WIRS analyses (McNabb and Whitfield, 1998).(8) However, to take account of the possible sensitivity of results to alternative El definitions of financial participation and direct communication, we experimented with multiple measures.

Another advantage of the data is its identification of El initiatives, that is, EI practices introduced within the preceding three years.(9) These merit attention, since the 'stock' of established practices may over-represent practices which managers perceive as successful.

In treating El practices as heterogeneous, and in testing for interaction effects arising from 'bundles' of practices, we are addressing two of the main sources of bias in estimates of EI's impact on performance (Bryson and Millward, 1997).

Analyses

Modelling approach

The analyses of financial performance are based on 564 private sector establishments providing valid non-missing data on financial performance and the independent variables in our models.(10) Public sector establishments are excluded because they were not asked the financial performance question. We also excluded establishments which were not making or providing goods or services (head offices and other purely administrative offices) because the hypotheses discussed above are most pertinent to firms operating in a commercial environment.

The dependent variable is the managerial assessment of the establishment's financial performance relative to other establishments in the same industry, as detailed above. For the purposes of analysis the last two categories identified in Table 2 have been collapsed into a single category so that the ordered variable runs from 0 (a little or a lot below average) to 3 (a lot above average). The financial performance variable is a categorical indicator defined in terms of ordered responses, so that we estimate our models using an ordered probit estimator.

We present models for small firms and large firms separately, rather than all firm models with firm size interactions because there is evidence of segmentation between the small firm and large firm sectors: they operate under different labour and product market conditions, and according to different internal systems of bureaucracy. This sample split is statistically merited according to a likelihood ratio test.(11)

All models are run on data weighted by the inverse of the sampling probability. Because large establishments had a higher probability of selection than smaller establishments (Airey, Tremlett and Hamilton, 1992), the weighted models provide better estimates for EI's effects on smaller establishments which are more common in the population than larger establishments (Skinner, 1997). This is appropriate here since we wish to fit models which offer a good approximation of El effects on small firms' performance. The use of probability weights also guards against estimation bias which can arise through differential sample selection probabilities.(12) Because the data we use are survey data we employ the Huber-White robust estimator of variance.(13) Table 3. Incidence of El practices among small and large firms Variables Small Large firms firms Financial participation FINPART1 (0,1): if has one of - profit- related pay scheme, establishment/ organisational level payments-by-results, any share ownership scheme 0.34 0.73 FINPART2 (0,1): as FINPART1 but excludes executive share schemes 0.33 0.71 ANYSHAR1 (0,1): if any share ownership scheme 0.06 0.51 ANYSHAR2 (0,1): as ANYSHAR1 but exc. executive share schemes 0.02 0.47 EXECSHAR (0,1): if any executive share scheme 0.03 0.24 ANYPRP (0,1): if any profit-related pay schemes 0.30 0.57 Direct communication DIRCOM3 (0,1): if has one of- regular meetings between junior managers/ supervisors and workers for whom they are responsible; regular meetings between senior managers and all sections of workforce; regular newsletters distributed to all levels of employees 0.51 0.84 MANCHAIN (0,1): if systematic use of management chain for communication with all employees 0.47 0.66 Upward-problem solving UPS (0,1): if has one of- problem-solving groups (regular meetings among work- groups or teams to discuss aspects of performance); suggestion schemes; surveys or ballots of employees' views or opinions 0.34 0.55 PROBSOLV (0,1): if regular meetings among work-groups or teams to discuss aspects of their performance 0.23 0.35 SAYSCHEM (0,1): if any suggestion schemes 0.14 0.35 SURVEYS (0,1): if any surveys or ballots of employees' views or opinions 0.05 0.21 Representative participation JCC (0,1): if has any joint committees of managers and employees primarily concerned with consultation rather than negotiation (apart from single issue committees) 0.12 0.26 INITIV (0,1): if management has made a change in last 3 years with the aim of increasing EI in the operation of the establishment 0.38 0.49 EI scores EICOUNT (0,10): number of the following - JCC,MANCHAIN,SAYSCHEM, PROBSOLV, ANYPRP, ANYSHARI, regular newsletter to all staff, regular meetings between junior managers and their workers, regular meeting between senior managers and all staff, survey/ballot of employees' views or opinions 2.11 4.46 SQRTEIC2: square root of EICOUNT+1 1.70 2.28 EI3PLS (0,1): if has 3 or more of the practices listed under EICOUNT 0.35 0.76 EI combinations DIRCOM3xUPS 0.25 0.51 UPSxMANCHAIN 0.19 0.41 INDIVPBRxDIRCOM3 0.14 0.22 Sample sizes weighted 178 386 unweighted 351 354 Note: Values are based on weighted data.

Control variables

Small firms are heterogeneous. Among the smallest firms, owners are likely to be actively involved in day-to-day management of the business, normally on a full-time basis. The owner/manager will be a generalist, dealing with problems of finance, marketing, production and so on without the help of managerial specialists (Atkinson and Meager, 1994). Recruitment will be rare and generally done by word of mouth. Day-to-day management of the workforce will be personal, ongoing and face-to-face, often with workers who are family members and friends from non-work social networks. By contrast, in firms close to our 200-employee threshold there is likely to be a separation between owners and managers: most managers will have been employed as managerial specialists, rather than managing by virtue of their ownership. These 'larger' small firms have a managerial hierarchy and specialist managers, sometimes including 'personnel' managers. Recruitment is a recurrent activity with much less reliance on family ties and personal contacts and a greater use of impersonal recruitment and selection procedures. Relations between management and employees involve a greater use of formal rules and procedures, with less face-to-face contact being the natural consequence of the larger numbers involved.

Firms that have grown rapidly may still resemble the micro-business run by the founding entrepreneur; others will have contracted from much larger entities and still retain most of their characteristics. The sector includes firms where the employer is highly dependent upon the skills and commitment of the existing workforce, as well as firms where the opposite is the case. It includes firms where employees are able to resist the exercise of proprietorial prerogative, as well as others where they cannot (Goss, 1988). It includes businesses where management-employee relations are generally harmonious (Curran et al., 1993) as well as others with more conflictual relations (Rainnie, 1989).

It is important to account for these differences in so far as they may have a bearing on small firms' financial performance. Failure to do so may result in estimation bias arising from omitted variables (discussed below). Table 1 gives definitions of these variables but we now briefly introduce and explain them.

Establishment characteristics: Establishment size indicates the extent to which an organisation can exploit economies of scale to reduce operational costs. Employment change over the last year is used as a proxy for change in demand for an establishment's products/services. Taken together, the size of the establishment, its age, and recent employment shifts can indicate what point the business has reached in its 'life-cycle'. Barring sectoral shocks, we might expect variance in performance to diminish with the age of the establishment, so that firms reaching operational maturity may be less likely to be relatively high or low performers. Multiple-establishment organisations are identified: they may be able to offer financial and other resources to the establishment in pursuit of better performance which may not be available to the single-establishment firm. Size of establishment relative to total employment in the sector is used as an indicator of the degree of market power.(14) Industry dummy variables are used to capture unmeasured industry effects. Some 'cultural' differences in the running of firms may be picked up by a dummy variable distinguishing UK and foreign-owned firms.

Nature of the product market: Competitive pressures are captured by the number of competitors the establishment faces in the market for its primary/sole product or service, and the geographical extent of that market. The degree of dependence on, or exposure to, that market is proxied by degrees of product diversification.

Workforce characteristics: The percentage of unskilled workers in the workforce is used as a proxy to distinguish labour-intensive and capital-intensive firms and technology levels across firms. It may also help control for differences in the receptiveness of workforces to EI, and the potential costs and benefits of introducing EI.

The impact of trade unions on financial performance and the differential impact of El in union and non-union firms, are well-documented (Cooke, 1994; Fernie and Metcalf, 1995; McNabb and Whitfield, 1997). Machin and Stewart (1996) show that, by 1990, the negative impact of recognised unions on financial performance was confined to establishments with strong union organisation. Accordingly, we differentiate establishments with 'strong' recognised unions from others with recognised unions.

Pay systems: We identify incentive pay systems which include merit pay or individual performance-related payments to isolate the effects of financial participation schemes on performance. Union recognition identifies establishments where pay for (at least some) workers is subject to collective bargaining, either at establishment, firm, sectoral or national level.

Information provision: The flow of information from management to workers may determine worker responsiveness to past under(over)-achievement, and the degree of trust and confidence in management. We therefore control for the impact of financial information, and regular information on wage costs and productivity.

Tackling possible sources of estimation bias

Since El is not randomly distributed across establishments, an association between El and performance may result from unobserved and unobservable differences between El and non-EI establishments which affect performance independently. We can be more confident that an association exists if appropriate data are used to capture differences between firms or employees which might be correlated with performance. So the models incorporate a range of variables which previous WIRS analyses identify as having significant effects on firm performance, together with other potentially important influences (Holl and Pickering, 1991; Holl and Dassiou, 1993; Machin and Stewart, 1996; Machin and Stewart, 1990; McNabb and Whitfield, 1996; McNabb and Whitfield, 1998: see above and Table 1).

Factors leading to the adoption or retention of El may themselves affect firm performance. Failure to correct for this will bias the estimates of EI's impact on performance. For example, if more extensive use of El practices is simply an indicator that a firm is well-managed across a range of functions, this may upwardly bias the estimated effects of El on performance (Huselid and Becker, 1996). We address this problem in two ways. Table 4. Determinants of the number of El practices (SQRTEIC2), all establishments Workplace characteristics ESTSIZ2 0.06 (1.18) ESTSIZ3 0.22 (2.88) ESTSIZ4 0.14 (1.63) ESTSIZ5 0.17 (1.65) LMKTSHAR -0.05 (1.86) SIZ2 -0.02 (0.25) SIZ3 0.08 (1.06) SIZ4 0.18 (2.30) SIZ5 0.35 (4.40) SIZ6 0.54 (6.62) SIZ7 0.85 (11.00) SIC1 -0.40 (2.49) SIC2 -0.16 (1.15) SIC3 -0.16 (1.39) SIC4 -0.29 (2.47) SIC5 -0.30 (2.20) SIC6 -0.10 (0.89) SIC7 -0.21 (1.71) SIC8 -0.06 (0.56) UKOWNED -0.04 (0.52) AGE21P -0.05 (1.13) COMPANY 0.03 (0.38) HITECH 0.18 (4.05) Product market NATINTER 0.18 (3.56) MANYCOMP 0.00 (0.07) SINGPROD -0.05 (1.24) Workforce characteristics PCMAN -0.003 (3.20) Union recognition UNIONREC -0.05 (1.00) UREC2PLS 0.19 (3.10) Personnel function ONBOARD 0.11 (2.27) SPECIAL 0.23 (4.41) Pay systems INDIVPBR 0.04 (0.86) MERITPAY 0.11 (2.68) Constant 1.56 (9.58) F f(33,2028) = 29.17 [R.sup.2] 0.569 Notes: OLS using weighted data. t-statistics in parentheses. 864 establishments weighted (667 unweighted). Reference categories are: ESTSIZ1, SIZ1, SIC9.

First, we add a fitted El score (PREDSQRT) to our performance models. This is generated by an OLS first-stage model estimating El scores for all establishments in our data set (Table 4). The dependent variable (SQRTEIC2) is the square root of a scale running from 1 to 11, where 1 is an establishment with no El practices, and 11 is an establishment with all ten El practices which make up EICOUNT.(15) With an [r.sup.2] of 0.57 the model accounts for much of the variation in the dependent variable.

The predicted EI score is effectively an index of characteristics which go to make up the propensity to involve employees. It can be seen as the degree of 'EI-ness' which one might expect at an establishment, given its observed characteristics. This score is then incorporated into the second-stage models which estimate performance in small and large firms respectively. It sits alongside the El variables which identify whether or not each establishment is observed as having various types of EI.

A significant coefficient on the predicted EI score would indicate that there are characteristics of EI firms which are associated with performance outcomes, and are independent of El effects. If the EI effects are unchanged with the introduction of the predicted EI score, they may be said to be independent of those factors associated with the introduction and retention of EI (Maddala, 1983).

Our second test for selection effects takes all the variables used to estimate the number of El practices which do not appear in the performance model, and simply adds them to the performance model. If EI effects are robust to their inclusion, this is further evidence that EI selection effects are not significant.

Another concern is that the introduction or modification of El occurs in response to changes in performance, as well as affecting subsequent performance (that is, EI is endogenous). For example, it may be that firms with higher productivity introduce EI practices, perhaps as a reward for good performance. If so, the selection of El by high-performing firms will overstate EI's effect on performance. Equally, if poorer performing firms adopt El in order to raise performance, estimates of EI's effects are likely to be underestimates of its real impact. Ideally, one would draw on information about the conditions under which El was adopted, but this is not available in our cross-sectional data. In a later paper we hope to address this issue by modelling financial performance treating El as potentially endogenous. Here we simply introduce a prior measure of performance, namely whether productivity increased over the preceding three years, and assess whether this has any effect on the El coefficients in the model.

In treating El practices as heterogeneous, and in testing for interaction effects arising from 'bundles' of [TABULAR DATA FOR TABLE 5 OMITTED] practices, we are addressing two of the other main sources of bias in estimates of EI's impact on performance (Bryson and Millward, 1997).

Results

'Baseline' models

Table 5 presents two types of model for small and large firms separately showing the impact of six types of El on financial performance. Columns 2 and 3 show the EI effects without control variables. Columns 4 and 5 show how these effects change once our 'baseline' control variables are added. The EI effects differ markedly for small and large firms.

Small-firm establishments benefit from systematic use of the management chain and regular use of direct communication methods, like team briefings. These are the least bureaucratic EI practices, and the least costly to maintain. The effects become more pronounced with the introduction of the control variables. These EI forms do not benefit large firms: indeed, once control variables are added, direct communication is significantly associated with lower financial performance among large firms. These findings are consistent with the hypothesis that direct communication methods 'fit' better in a small-firm environment. They also challenge the conventional wisdom that, as fairly 'low intensity' forms of EI, they are liable to have little effect on performance. In small firms regular contact and communication with decision-takers who are in close proximity may form the basis for meaningful participation in decision-making and trust-based employee relations. Financial participation schemes have no significant effect on small firms' performance, suggesting that the costs in implementing and maintaining such schemes counter any added incentive effect which may arise from fewer income sharers.(16) Financial participation appears to be more effective in large firms. Although the significant effect in the 'EI variables only,' model disappears with the addition of control variables, it is apparent again in the better fitting large firm model discussed below (see Table 8).

Individual performance-related payments significantly improve small firms' financial performance, but have no significant effect on large firms' performance. They may be more suited to the small-firm environment, where monitoring individual effort is less costly and wages are more responsive to individual productivity. However, merit payments made largely at the discretion of managers reviewing past performance have no such effect in small or large firms, a finding consistent with the view that incentive effects are greater where the criteria for determining performance payments are objective and explicit, and applied consistently.

Upward problem-solving has a strong negative impact on small firms' performance, whereas the coefficient is positive and non-significant for large firms. But, as discussed below, the impact of upward problem-solving on small firms' performance is contingent on the package of EI practices the firm has adopted.

Small firms introducing an El initiative in the previous three years were performing more poorly than other small firms. EI initiatives were not significantly associated with large-firm performance. It is possible that small firms take such initiatives in response to deteriorating performance. However, the effect on small-firm performance remained after controlling for productivity change over the previous three years as a proxy for past performance. An alternative explanation is that the initial costs in devising and implementing an initiative outweigh the short-term benefits.

Joint consultative committees (JCCs) are the only form of El identified in Table 5 which appear to have no significant effect on small-firm or large-firm financial performance.(17) Union recognition is a stronger form of representative participation for employees than JCCs because it entails negotiation with management over pay and conditions. But it is not normally regarded in the literature as a form of EI, despite the fact that British employers are at liberty to decide whether or not they recognise unions. This is probably because, although unions can engage in constructive dialogue with employers, giving 'voice' to employees' concerns and offering useful support and assistance in achieving the firm's objectives, unions can also mobilise worker support for action in support of claims which managements perceive as contrary to the short-term or long-term interests of the firm. The 'constructive' side of union recognition predominates among small firms where union recognition is significantly associated with better performance; the coefficient for strong unions is negative but non-significant. In large firms, on the other hand, strong unions have a significant negative effect on performance, whereas other recognised unions have a positive but non-significant effect.

Interaction models for small firms

Table 6 presents four models estimating financial performance for small firms. All contain EI interaction terms to capture the effects of EI in isolation and in combination. Model 1 in Table 6 adds EI interaction terms to the small firm 'baseline' model presented in Table 5. The coefficients in the El variables row represent the main effects of each El practice in isolation, whereas the [TABULAR DATA FOR TABLE 6 OMITTED] interaction terms represent the effects of EI in combination. These estimates show that the impact that an EI practice has on a small firm's performance differs depending upon whether it exists in isolation or, if not, the other EI practices with which it combines.

In isolation, direct communication has a positive, significant effect on small-firm performance similar in magnitude to the effect identified in the baseline model. But, in combination with upward problem-solving, it has a much stronger positive effect.(18) Without direct communication, however, upward problem-solving has a significant negative effect on performance which is stronger than the effect identified in the baseline model. It seems that attempts to tap into worker knowledge, and to elicit greater worker effort through problem-solving EI, can prove counterproductive it they are not combined with regular two-way communication between management and workers. Managers can use regular direct communications to explain their plans for problem-solving, while workers can use them to convey concerns and alternative approaches. This may foster the mutual trust which, through organisational commitment and job satisfaction, can have positive effects on firm performance. In essence, small-firm workers may feel (and actually be) more involved in decision-making processes because there is two-way communication. Direct communication may be seen as a non-financial pay off for the extra effort workers are being asked to make on behalf of the firm.

In isolation, the systematic use of the management chain has a positive significant effect on small-firm performance which is appreciably larger than the average effect of management chain use identified in the baseline model. However, it has a negative effect on performance in combination with upward problem-solving. Fernie and Metcalf (1995) identified the posture effect the management chain has on productivity levels, despite a negative influence on the climate of industrial relations. So, one-way downward communication of management instructions can bring financial rewards for small firms, but what Fernie and Metcalf (1995) term this authoritarian approach appears to conflict with the more collaborative approach required to make upward problem-solving work.

The model confirms significant positive effect of individual payments-by-results on small firm performance but, contrary to expectations, they are not used as a financial pay-off in combination with EI to improve firm performance. Their only significant effect in combination with EI is the negative effect they have in conjunction with direct communication. There may be a conflict between the use of direct communication which is group- or establishment-wide, and incentive payments targeted at individuals.

Financial participation schemes have no significant effect on small firms' financial performance, either in isolation or in combination with other EI practices.(19) So the hypothesis that EI will have the greatest impact on performance when control rights are combined with financial rewards does not hold for small firms. Furthermore, there is no one EI practice which, if combined with others, can guarantee improved financial performance for small firms. Direct communication and upward problem-solving have positive effects in combination with one another, but both combine with other practices to worsen performance.

Information provision: It is sometimes suggested that EI communication techniques can improve firm performance by improving information flows between management and workers which, in turn, increases trust in management, improving organisational commitment. It is unlikely that workers will respond to financial incentives if they have limited information on their performance. To test for the impact of information provision we introduced dummy variables identifying whether management provided information to workers on wages, productivity, or financial circumstances, at least annually. Only financial information was significantly associated with improved small-firm performance, and then only if managers provided what our management respondent considered to be "a lot of financial information" (see Model 2 in Table 6). But the introduction of the financial information variables has very little impact on the EI main effects and interactions: the sign and significance of the El coefficients remains the same, and their magnitude is little changed, indicating that the positive effect of financial information provision is independent of EI effects.(20) However, it is another indication that open workplace governance based on good communication between management and workers is associated with better small-firm performance.

Biases in El estimation arising from selection into El status: In Model 3 we add a fitted El score (PREDSQRT) using the method discussed earlier. The predicted EI score is positively associated with small firm performance, as expected. However, it is not statistically significant, and it has no bearing on the impact of the EI variables. This indicates that factors determining an establishment's propensity to adopt El are not biasing the El estimates. (21)

We approach the same issue in a different way in Model 4. This time we take all the variables used to estimate the number of EI practices used which do not [TABULAR DATA FOR TABLE 7 OMITTED] appear in the performance model, and simply add them to the performance model. There are five such variables, only one of which has a significant effect on small firms' performance. Again, the EI effects are unaltered, lending further support to the view that the EI effects are not subject to selection biases.

Testing for EI endogeneity: EI may be introduced as a reward for good performance, or as a spur to good performance in poorly performing organisations. As such, it may be endogenous. This paper does not attempt to address this issue rigorously. We simply introduce a prior measure of performance, PRODINC, a dummy variable identifying those who thought that labour productivity was "a lot higher" than it was three years ago. The coefficient for this variable is always positive, as expected, but not statistically significant. The EI effects are unaltered, supporting the proposition that the EI estimates on small firm performance are not biased by prior economic performance.(22)

Interaction models for large firms

The large-firm equivalents to the models reported above are presented in Table 7. None of the EI variables and their interactions have a significant effect on large-firm performance. Indeed, all four models are a poorer fit than the baseline model in column 5 of Table 5. The models simply confirm the very different impact that El has in small-firm and large-firm establishments.(23) However, it would be wrong to conclude' that El has no significant impact on large-firm performance.

Because these models are good fit models for small firms, as opposed to large firms, we also present a better fit model for large firms in Table 8.

The EI effects in this model are quite marked. Contrary to expectations, EI combinations are not associated with better large-firm performance, whereas isolated EI practices do seem to improve performance. Financial participation and upward problem-solving both have strong positive effects on performance when operating in isolation but, when combined, the additive effect is negative but non-significant. This is contrary to previous findings in the literature (Bryson and Millward, 1997). JCCs also have a significant positive effect on large-firm performance when operating in isolation, while the coefficient for direct communications is positive but non-significant: when combined, their additive effect on performance is negative and significant.

Quantifying the effects of El on small firms' financial performance

El practices have different, and sometimes opposing, effects on small-firm performance in isolation and in combination with other practices. There is no one EI practice which, if combined with others, can guarantee improved financial performance for small firms. So what sort of EI package would most benefit small firms? Table 8. Better fitting model estimating large firms' financial performance EI variables FINPART1 0.61 (2.43) DIRCOM3 0.19 (0.91) UPS 0.71 (1.96) MANCHAIN 0.32 (1.60) JCC 1.48 (3.47) INITIV 0.09 (0.55) Interaction terms FINPART1xUPS -0.63 (1.49) DIRCOM3xJCC -1.58 (3.35) Workplace characteristics ESTSIZ2 -0.31 (1.35) ESTSIZ3 -0.07 (0.26) ESTSIZ4 0.32 (1.02) ESTSIZ5 0.03 (0.06) LMKTSHAR 0.05 (0.48) SIZ3 -0.14 (0.45) SIZ4 0.36 (1.16) SIZ5 -0.32 (1.30) SIZ6 -0.40 (1.48) EMPRISE1 -0.01 (0.08) SIC1 -0.79 (1.57) SIC2 -0.21 (0.35) SIC3 -0.05 (0.10) SIC4 -0.19 (0.38) SIC5 1.68 (2.22) SIC6 0.28 (0.71) SIC7 -0.06 (0.12) SIC8 0.26 (0.57) AGE21P -0.26 (1.47) Product market INTERNAT -0.80 (3.64) MANYCOMP -0.09 (0.48) SINGPROD -0.41 (2.34) Workforce characteristics PCUNSKIL -0.00 (1.20) Unionisation UNIONREC 0.11 (0.47) USTRONG -0.31 (1.40) Information provision INFOPROD -0.63 (3.09) LOTFINFO -0.09 (0.49) Thresholds _CUT1 -2.00 (2.87) _CUT2 -0.42 (0.60) _CUT3 0.87 (1.23) F f(35,2026) = 3.22 Notes: t-statistics are in parentheses. Model is weighted. 386 weighted large firm establishments (354 unweighted).

In Table 9, we use the coefficients from Model 2 in Table 6 to illustrate the impact of six different EI regimes on performance. The regimes range from no EI at all, through to a combination of three EI practices, a recent EI initiative, and individual performance-related payments.(24) The effects of each package are evaluated for three different types of small-firm establishment: those observed as having few EI practices (scoring under 2 on EICOUNT); those with medium levels of EI (scoring 2 or 3 on EICOUNT); and those with a high number of El practices (4 or more on EICOUNT).

The estimated effects are calculated by multiplying the coefficients from our preferred statistical model by the mean scores on each independent variable for that firm type. In this way it is possible to simulate the effect of shifts in the EI regime holding other characteristics constant.

So, for example, in panel 1 of Table 9, the bottom line in the 'No EI' row relates to 'High EI' establishments. The figures in that row show the change in the probabilities that a 'High EI' small-firm establishment would have of being in each of the four performance categories if it switched to being a 'No EI' establishment, ceteris paribus.

The 'No EI' row shows the negative impact that removing existing EI practices would have on small firm performance. Even among those with 'Low, EI', a shift to 'No EI' would reduce the probability of performance 'a lot above average' by 13 percentage points to 8 per cent. The best EI package for small-firm establishments is a combination of direct communication and systematic use of the management chain (panel 2) - regardless of current levels of El activity at the establishment. 'Low EI' small firms would increase the probability of recording performance 'a lot above average' by 34 percentage points to 55 per cent if they adopted this package. 'Medium EI' small firms would see a 28 percentage point increase to 56 per cent, and 'High El' firms a 32 percentage point rise to 53 per cent.

The addition of individual performance-related pay (panel 4) is of no net benefit compared with the direct communication/management chain combination because the positive effect of individual performance-related payments in isolation is offset by the negative effect it has in combination with direct communication. Table 9. The difference EI can make to small firms' financial performance A lot/ About Little A lot little average above above below No EI Low EI -0.01 +0.18 -0.03 -0.13 Med EI -0.01 +0.21 -0.01 -0.20 High EI +0.02 +0.19 -0.06 -0.14 DIRCOM3+MANCHAIN Low EI -0.09 -0.29 +0.04 +0.34 Med EI -0.08 -0.26 +0.05 +0.28 High EI -0.06 -0.28 +0.03 +0.32 DIRCOM3+MANCHAIN+UPS Low EI -0.08 -0.10 +0.10 +0.08 Med EI -0.07 -0.07 +0.11 +0.02 High EI -0.05 -0.08 +0.08 +0.06 DIRCOM3+MANCHAIN+INDIVPBR Low EI -0.09 -0.28 +0.05 +0.31 Med EI -0.08 -0.24 +0.06 +0.26 High EI -0.06 -0.27 +0.04 +0.30 DIRCOM3+MANCHAIN+UPS+INDIVPBR Low EI -0.08 -0.07 +0.10 +0.06 Med EI -0.07 -0.04 +0.11 0.00 High EI -0.05 -0.06 +0.07 +0.04 DIRCOM3+MANCHAIN+UPS+INDIVPBR+INITIV Low El -0.02 +0.16 -0.02 -0.12 Med EI -0.02 +0.19 +0.01 -0.18 High EI -0.01 +0.16 -0.04 -0.12 Notes: The table records changes in fitted probabilities based on coefficients in model 2 in Table 6 applied to mean scores for non-EI independent variables for three sorts of small firm establishment Low EI who scored less than 2 on EICOUNT; Medium EI who scored 2-3 on EICOUNT; and High EI who scored 4+ on EICOUNT. Each row in the table records the financial performance impact of changing to the EI regimein bold type for each of these three types of establishment. The fitted probabilities for the whole small firm sample under model 2 in Table 6 were: a lot/little below average: 0.08; about average: 0.41; a little above average: 0.27; a lot above average: 0.24.

Adding upward problem-solving to the direct communication/management chain combination (panel 3) would bring substantial performance improvements to small firms, but they do not compare to the benefits of the direct communication/management chain package. This is because the negative effect of upward problem-solving in isolation, and in conjunction with the management chain, outweighs its positive effect in combination with direct communication.

The most EI-intensive regime (panel 6) would have a negative effect on performance akin to switching to a 'No EI' regime, emphasising the need for small firms to think carefully before adopting or dropping El practices.

Concluding remarks

This paper has investigated the relationship between El and financial performance using establishment-level data from the 1990 WIRS. Although we have not formally tackled the possibility that El is endogenous with respect to financial performance, our findings are consistent with theories predicting a differential impact of El on small-firm establishments and large-firm establishments. We have shown that it is not possible to generalise about the sorts of El practices and combinations which work for firms in general: what works for small-firm establishments is very different from what works for large-firm establishments. The least bureaucratic and least costly methods of EI - direct communication and systematic use of the management chain - have the potential to benefit small firms most. But whether they actually do so depends on the array of other EI and non-EI practices in operation: the wrong configuration can have a negative effect on performance, even if it includes direct communication and the management chain.

This research was supported by the Economic and Social Research Council (grant reference R000222294). The author thanks Neil Millward, his collaborator in the study from which this paper emanates. I also thank Michael White and Richard Dorsett at the Policy Studies Institute, and Chris Skinner and Ray Chambers at the University of Southampton, for useful comments. Copies of the data and computer programs used to generate these results are available from the author by e-mailing him at [email protected].

NOTES

1 Levine and Tyson (1990) argue that the maintenance of employee support for EI depends on linking it to financial and non-financial gains, like job security. These supporting mechanisms foster long-term commitment to employment and group cohesiveness which is necessary for EI to embed itself in an organisation. Furthermore, without managerial assurances of job security, workers may feel that there is no guarantee that productivity gains achieved through EI will not translate into job losses.

2 This is consistent with evidence that worker selection into small-firm and large-firm sectors has a significant effect on employer-size wage differentials (Idson and Feaster, 1990). See Curran and Stanworth (1979) for a critique of Ingham (1970).

3 However, the overall net effect of worker representation on performance is likely to vary with the strength of that representation (Fernie and Metcalf, 1995). For example, worker bodies with a consultative role may be able to voice worker concerns, but will have little opportunity to press for worker interests. Strong recognised unions with negotiating rights will be more able to pursue such aims vigorously.

4 Where management fail to communicate the objectives of a scheme, or incentive systems of reward are considered unfair, the introduction of a scheme can worsen labour-management relations (Mitchell et al., 1990). It is possible that such effects will feed through to firm performance.

5 An overview of the survey results is given in Millward et al., 1992.

6 It is representative of establishments with 25-199 employees belonging to firms with 25-199 employees, rather than small firms with 25-199 employees.

7 It is clear from Table 3 that EI incidence is much lower in small firms than in large firms, but if EI helps small firms grow, whereas non-El small firms are more likely to remain small, EI will be less prevalent among the stock of small firms than among 'successful' small firms.

8 However, our measures are not identical. We have incorporated establishment- and firm-wide payments-by-results in our measure of financial participation, and we have separated systematic use of the management chain from other forms of direct communication.

9 WIRS asks: "Has management here made any change in the last 3 years (or since you began operating here if less than 3 years) with the aim of increasing employees' involvement in the operation of the establishment?" If the respondent said "yes" they were asked "What changes have there been?" The most common initiatives involved extending or introducing direct communication.

10 There is no imputation.

11 With 32 degrees of freedom, chi2 = 44 The critical values are 43 for 95 per cent confidence, and 46 for 90 per cent confidence. Presenting separate models creates some expositional difficulties because the 'best fitting' model for small firms is not a well fitting model for large firms, and vice versa. We have chosen to present models which are best fitting for small firms since they are the focus of our paper. However, in Table 7, which is appended, we present a better fitting model for large firms to indicate what sorts of EI interactions appear to have significant effects for large firms.

12 Differential sampling fractions can result in standard estimator biases (Skinner, 1997). The weights account for all variation in sampling probabilities, thus eliminating differential sampling probability as a possible source of estimation bias.

13 This produces consistent standard errors in the presence of heteroscedasticity. The F statistic reported for each model is a Wald test based on the robustly estimated variance matrix.

14 The variable used, LMKTSHAR, is the log of the number of employees in the establishment divided by total employment at the 4-digit SIC level in September 1989. The latter are taken from the Employment Gazette (May 1991). It is only a loose approximation for market power, and is likely to be a particularly poor proxy in capital-intensive production.

15 The square root of EICOUNT2 was used as the dependent variable so as to attach greater weight to increases in El at the lower end of the distribution. This reflects a belief that a movement from no El to one El practice, or from one El practice to two, is a more significant change in the degree of El practised in an establishment than an increase from, say, nine to ten practices.

16 The positive effects of financial participation on performance are largely confined to profit-related payments (McNabb and Whitfield, 1996; Blanchflower and Oswald, 1988; Fernie and Metcalf, 1995). We disaggregated FINPART1 in models not reported here: none of these profit-related pay or share schemes proved significant for small firms' financial performance.

17 However, a JCC effect is discernible for large firms with a better fitting large firm model, as discussed below.

18 The net effect of having DIRCOM3 and UPS is obtained by adding together the coefficients for their main effects and interaction effect (0.53 + 1.28 - 1.00 = 0.81).

19 This is generally so whether profit-related pay schemes and share schemes are grouped together as in the models reported, or specified separately. However, there was a significant positive interaction between profit-related pay and EI initiatives, suggesting that combining 'control' and 'return' rights may benefit small firms financially, at least when introducing or amending EI schemes.

20 We interacted EI with financial information provision to see whether it enhanced or explained any particular EI effect. All interaction terms were non-significant.

21 This finding is robust to variations in the set of identifying variables in the first stage model estimating SQRTEIC2.

22 It is worth recalling that our dependent variable is financial performance measured in 1990. It is likely that most of the EI practices identified by management respondents would have pre-dated 1990 by quite some time (the variable INITIV identifies EI changes in the previous three years). This means that productivity change in the previous three years might not capture the conditions under which most EI practices were introduced. But it also reduces the probability that EI is endogenous, and supports the inclusion of EI variables on the right-hand side of the equation.

23 As was the case in the small-firm models, the predicted El score is positive and non-significant (Model 3, Table 7), and only one of the EI determinants has a significant effect on performance (Model 4, Table 7). In neither case do the extra variables have any real bearing on the EI coefficients. 24 Although individual performance-related payments are not regarded as a practice for involving employees, they do interact strongly with direct communication strategies, making it necessary to assess their impact.

REFERENCES

Airey, C., Tremlett, N. and Hamilton, R. (1992), The Workplace Industrial Relations Survey (1990): Technical Report (Main and Panel Surveys), London: Social and Community Planning Research.

Arthur, M. B. and Hendry, C. (1990), 'Human resource management and the emergent strategy of small to medium sized business units', The International Journal of Human Resource Management, 1, 3, pp. 233-50.

Atkinson, J. and Meager, N. (1994), 'Running to stand still: the small firm in the labour market' in Atkinson, J. and Storey, D. (eds), Employment, the Small Firm and the Labour Market, London: Routledge, pp. 28-102.

Baird, L. and Meshoulam, I. (1988), 'Managing two fits of strategic human resource management', Academy of Management Review, 13, pp. 116-28.

Ben-ner, A. and Jones, J. C. (1995), 'Employee participation, ownership and productivity: a theoretical framework', Industrial Relations, 34, 4, pp. 532-54

Bishop, J. (1987), 'The recognition and reward of employee performance', Journal of Labor Economics, 5, 4, s36-56.

Blanchflower, D. G. and Oswald, A.J. (1988), 'Profit-related pay: prose discovered?', The Economic Journal, 98, pp. 720-30.

Bryson, A. and Millward, N. (1997), Employee Involvement in Small Firms: A Review of the Literature, London: Policy Studies Institute.

Cooke, W. N. (1994), 'Employee participation programs, group-based incentives and company performance: a union-nonunion comparison', Industrial and Labor Relations Review, 47, 4, pp. 594-609.

Curran, J., Kitching, J. Abbott, B. and Mills, V. (1993), Employment and Employment Relations in the Small Service Sector Enterprise, Kingston, Surrey: Kingston Business School.

Curran, J. and Stanworth, J. (1979), 'Worker involvement and social relations in the small firm', Sociological Review, 27, pp. 317-42.

Defourney, J., Estrin, S. and Jones, D. C. (1985), 'The effects of workers' participation on enterprise performance: empirical evidence from French cooperatives', International Journal of Industrial Organisation, 3, pp. 197-217.

Fernie, S. and Metcalf, D. (1995), 'Participation, contingent pay, representation and workplace performance: evidence from Great Britain', British Journal of Industrial Relations, 33, 3, pp. 379-415.

Freeman, R. B. and Medoff, J. L. (1984), What Do Unions Do?, New York: Basic Books.

Goss, D. M. (1988), 'Social harmony and the small firm: a reappraisal', Sociological Review, 36, pp. 114-32.

Green, F. and McIntosh, S. (1996), 'Workers' effort in unionised workplaces: cost of job loss and human resource management', unpublished mimeo, School of Business and Economic Studies, University of Leeds, and Centre for Economic Performance, London School of Economics.

Hammer, T. H. (1988), 'New developments in profit sharing, gainsharing, and employee ownership', in Campbell, J.P. and Campbell, R. J. (eds), Productivity in Organisations, San Fransisco: Jossey-Bass.

Holl, P. and Dassiou, X. (1993), 'The determinants of performance in British establishments', City University, Applied Econometrics Research Unit Discussion Paper, No. 32.

Holl, P. and Pickering, J.F. (1991), 'Takeovers and other influences on economic performance: a plant level analysis', Applied Economics, 23, pp. 1779-88.

Huselid, M.A. (1995), 'The impact of human resource management practices on turnover, productivity and corporate financial performance', Academy of Management Journal, 38, 3, pp. 635-72.

Huselid, M.A. and Becker, B.E. (1996), 'Methodological issues in cross-sectional and panel estimates of the human resource-firm performance link', Industrial Relations, 35, 3, pp. 400-22.

Ichniowski, C., Kochan, T. A., Levine, D., Olson, C. and Strauss, G. (1996), 'What works at work: overview and assessment', Industrial Relations, 35, 3, pp. 299-333.

Ichniowski, C., Shaw, K. and Prennushi, G. (1997), 'The effects of human resource management practices on productivity: a study of steel finishing lines', American Economic Review, 87, 3, pp. 291-313.

Idson, T.L. and Feaster, D.J (1990), 'A selectivity model of employer-size wage differentials', Journal of Labor Economics, 8, 3, pp. 99-122.

Ingham, G. K. (1970), Size of Industrial Organisation and Worker Behaviour, Cambridge: Cambridge University Press.

Kessler, I. and Purcell, J. (1992), 'Performance related pay: objectives and application', Human Resource Management Journal, 2, 3, pp. 16-33.

Kruse, D. L. (1993), 'Does profit sharing affect productivity?', NBER Working Paper No. 4542, Cambridge, Mass.: National Bureau of Economic Research.

Lazear, E. P. and Rosen, S. (1981), 'Rank-order tournaments as optimum labor contracts', Journal of Political Economy, 100(4), pp. 841-64.

Levine, D.I., and Tyson, L.D. (1990), 'Participation, productivity and the firm's environment', in Blinder, A. S. (ed.), Paying for

Productivity: Looking at the Evidence, Washington D.C.: Brookings Institute.

McNabb, R. and Whitfield, K. (1996), 'The impact of financial participation and employee involvement on financial performance', Scottish Journal of Political Economy, 45, 2, May, pp. 171-87.

McNabb, R. and Whitfield, K. (1997), 'Unions, flexibility, team working and financial performance', Organisation Studies, 18, 5, pp. 821-38.

Machin, S. and Stewart, M. (1990), 'Unions and the financial performance of British private sector establishments', Journal of Applied Econometrics, 5, pp. 327-50.

-----(1996), 'Trade unions and financial performance', Oxford Economic Papers, 48, 2, pp. 213-41.

Maddala, G. S. (1983), Limited-Dependent and Qualitative Variables in Econometrics, Cambridge: Econometric Society Monographs, Cambridge University Press.

Marchington, M., Goodman, J., Wilkinson, A and Ackers, P. (1992), New Developments in Employee Involvement, London: Employment Department, Research Series No.2.

Marchington, M., Wilkinson, A., Ackers, P. and Goodman, J. (1993), 'The influence of managerial relations on waves of employee involvement', British Journal of Industrial Relations, 31, 4, pp. 553-76.

Marsden, D. and Momigliano, S. (1995), 'Economic theory of incentives and its implications for the design of public service performance related pay schemes', unpublished mimeo, London School of Economics.

Millward, N., Stevens, M., Smart, D. and Hawes, W. (1992), Workplace Industrial Relations in Transition: the ED/ESRC/PSI/ACAS Surveys, Aldershot: Dartmouth.

Mitchell, D.J.B., Lewin, D., and Lawler, E.E. III. (1990), 'Alternative pay systems, firm performance and productivity', in Blinder, A.S. (ed.), Paying for Productivity. A Look at the Evidence, Washington D.C.: Brookings Institute.

Rainnie, A. (1989), Industrial Relations in Small Firms: Small Isn't Beautiful, London: Routledge.

Scott, M., Roberts, I., Holroyd, G. and Sawbridge, D. (1989), Management and Industrial Relations in Small Firms, London: Department of Employment Research Paper No. 70.

Skinner, C. (1997), 'The use of sample weights in the regression analysis of WIRS data', unpublished paper, University of Southampton.

Weitzman, M.L. and Kruse, D.L. (1990), 'Profit sharing and productivity', in Blinder, A.S. (ed.), Paying for Productivity. A Look at the Evidence, Washington D.C.: Brookings Institute.

Alex Bryson is a Principal Research Fellow at the Policy Studies Institute, London.