preferred the ten men it replaced!”
3. Statistical Studies
If the previous modes of inquiry help foster the layperson's view that the machine is a threat to jobs, when the statistical link between innovative activity and employment is carefully observed a different picture emerges, at the micro-economic level at least. Four 'micro' levels of aggregation will be considered: process, plant, firm and industry in turn.
Studies at the process level take a particular micro-electronic technique and get engineering estimates of job losses. For example. Fleck (1984) calculated that 1.4 jobs would be lost for each robot used in an 'average' plant. The equivalent figure was 0.6 employees in Japan (Watanbe, 1987) and one for one in the United States (Hunt and Hunt, 1983). Early forecasters used these estimates to get extremely gloomy predictions about the fate of some manufacturing branches, but it is easy to see that optimists could point to the relatively slow progress of robots to forecast minimal estimates of job loss. These predictions are bogus in any case as first, they focus on sub-processes that are not applicable across all industries and secondly, the output expansion effect is ignored entirely.
Studies at the establishment level are more carefully documented. The 1984 Workplace Industrial Relations Survey (WIRS 2) and the PSI studies have data on both current and past employment, enabling researchers to look at innovation and employment changes. The PSI study shows that plants which used microelectronics experienced a mean employment growth of 1.9% between
1985 and 1987 whereas those who did not actually had a decline of 9.4%. Daniel’s (1987) analysis of WIRS2, notes that although new technology was generally associated with declines in employment, there were larger falls in manual employment for non-users of micro-electronics in private manufacturing over 1980-1984 than for users (Table IX.18). Since the 1980s were a decade of dramatic change for British manufacturing, these simple correlations are unlikely to be convincing. Machin and Wadhwani (1991b) present employment equations from WIRS2 to see whether the correlations stand up to controls for unionisation, other types of change, industry and regional dummies, etc. The controls seem to strengthen the positive relationship and, taken at face value, imply a 2.7% employment growth effect associated with advanced
technical change across the entire private sector^.
Although an improvement, employment equations from WIRS must be treated with caution for, as Machin and Wadhwani (1991a) point out, there are no adequate controls for capital, sales or fixed effects. Firm level studies are more suited to this purpose but unfortunately there is a large gap in the literature here. An exception is Entorf and Pohlmeier (1991) who examine a sample of West German manufacturing firms in 1984. They construct a simultaneous probit model for innovation, employment and export share and estimate that product (but not process) innovations significantly increase employment even when treated as endogenous. The effects are small (just over two Jobs per innovation) but this is probably due to a very broad definition of what constitutes a new product (57% of all firms realised at least one). Unfortunately, as the authors admit in their conclusion, the use of only one cross section prevents them from allowing for dynamics which are known to be
^Note that the Blanchflower et al (1991) results reported in their appendix from a larger sample are practically identical.
vitally important in employment equations and the longer run effects of 7
technical change .
By focusing on industrial data, economists have been able to combine cross sectional with time series analysis. Salter (1960) pioneered the way, reporting a long run positive effect of Total Factor Productivity (TFP) on employment. The relationship between the Solow residual and industry employment appears to have weakened since the Second World War; indeed it seemed to practically disappear in the 1970s (Wragg and Robertson,1978; Ball and Skeoch ,1981). More recently, Nickell and Kong (1989) have ambitiously estimated a four equation (production, pricing, demand and wages) imperfect competition model across nine manufacturing industries 1974-82. They use the Solow residual to proxy technology and recover the structural parameters, in particular the industry price elasticity of demand. Since the writers call their estimates of these demand elasticities ’ the least reliable', they are calculated in various ways to give a plausible range of values. Only in Bricks and Glass and Textiles is technical change calculated to have a negative effect on employment.
The use of TFP has problems in these studies as part of it may be related to factors which do not reflect innovation (e.g. market power, higher worker effort, etc). Freeman and Soete (1987) have used the head count of
innovations measure utilised in this and earlier chapters to distinguish different manufacturing sectors according to technological criteria. In the 1980s they uncovered a close relationship between the innovative
7
For example, Osterman (1986) studied the effects of increased computer power across 40 2 digit U.S. industries between 1972-1978. Overall there were decreases in employment for white collar workers, but these were concentrated very heavily in the first years after adoption. In subsequent years there was an increase in the industry’s labour force.
g
’pervasiveness’ of a sector and employment growth. The credibility of their exercise depends on whether the definition of ’pervasiveness’ was not influenced by the successful growth of the industry in question. Even if it was not, it is still essentially a correlation exercise merely whetting one’s appetite for a closer analysis.
In summary, the evidence at the micro-level does not support the hypothesis that innovation is a major cause of unemployment. It must be stressed once again that the general equilibrium effect may well be different, even though the disaggregated behavioural studies reviewed here suggests that job opportunities have been enhanced by innovation. This conclusion must be tempered by the lack of existing studies which combine output measures of innovation with dynamic employment determination. The former seem available mainly at the firm level, and the latter at the industry level. Our empirical work seeks to fill this gap.
III.
Employment and Technical Change: TheoryIn this section we present some simple bargaining models and examine the employment effects of labour saving process innovations. Appendix 5.1 looks at the more general case of non-neutral technical change, so the reader is directed there for a more general treatment. Additionally, we do not look at product innovations are not looked at directly as it is obvious that at the level of the firm (although not at the industry or economy level) that new products will expand employment if labour intensity of new production is no lower than in the existing product-range. If they do have a lower intensity
g
By pervasiveness they mean (i) those who produce more innovations than they use and (ii) supply more than 20 downstream three-digit industries. The computing, electronics and plastics industries fall within this sector.
then the analysis will proceed on the same lines as a process innovation. Finally, we are observing the effects of innovation net of any change in the elasticity of labour demand induced by new techniques.
Recall from Chapter 2 that the structural wage and employment equations can be written for the three main union bargaining models as:
Labour Demand Models W(|3,X^,X^), N(W,X^) (5.0a,5.Ob)
Efficient Bargaining W(|3,X^,X^), N(W,X^,X^) (5.0c,5.0d)
General Bargaining Model WO'^,X^,X^) N (W, X^ , X^) (5.0e,5.0f)
Where ^ = firm's power over employment, f3 = firm’s power over the wage, X^= shifters in union utility. The primary concern in this section is to model the effects of one of the exogenous influences on profits (X^) namely, technological innovations. A secondary concern will be the correct interpretation of one of the influences on one of the exogenous determinants of union utility (X^) namely, the alternative wage (W). Each of the three union models are analysed in turn, then models of worker effort and product market power are brought in.
1. Labour Demand Models
The fundamental premise of Labour Demand models is that the wage is predetermined and the employer unilaterally sets employment. This could be the case under perfect competition in the labour market or a Right to Manage model. The usual first order condition for employment holds: the wage is set to equal the marginal revenue product of labour. It is convenient to write this as (subscripts denote partial derivatives):
w(N,W,X^) = R^ = W
sign of {3N/9A} is the same as the sign of {3w/3A}. Assume that technical change is labour augmenting so that the revenue function takes the standard multiplicative form R = R(AN,K,X^) where A = a productivity index which will depend on the state of technology. Using Young’s Theorem we can write
R = 3R /3A = (N/A)R + R /A
NA N NN N
(since R = (N/A)R ). This can also be written as
Where t} = -NR /R , the elasticity of the labour demand curve. The
NW NN N
fundamental result (which holds across many models) is:
Proposition In Labour Demand Models, an innovât ion will increase employment if the elasticity of the labour demand is greater than unity.
A second proposition particular to the current model is
Proposition 2. In Labour Demand Models, an innovât ion will decrease employment if the elast icity of labour demand is less than unity
This is illustrated for the simple case of linear demand in Figure 5.1. An innovation causes a clockwise shift in the demand curve about the point where the elasticity is unity. For high wage levels (W^), we are on an elastic part of the demand curve and employment increases, the opposite happens at lower wage levels (W^).
The intuition behind this result is straightforward. The effect of innovation on employment will depend on the relative profitability of increasing output in response to a decline in labour costs. If this is large increases in output will outweigh the fact that the same amount of production is feasible at a lower level of labour input. The output expansion effect will be greater the more elastic is product demand, the larger labour is in total costs and the easier it is to substitute labour for other factors.
These are the elements which make up the elasticity of labour demand through the Marshallian Rules.
For a revenue function with more general technology R = R(A,N,K,X^), condition (5.1) becomes
(5.2) R = (R /AR)(1 - 0 /(T )
NA N N AN
where 0 = WN/R, labour’s share in revenue and <r = R R /RR , the
N AN A N AN
elasticity of substitution between labour and A, ’technical capital’, (the meaning of A is discussed more closely later). A fuller analysis of this is given in Appendix 5.1, but the basic premise that higher elasticities are more likely to generate positive employment effects still holds.
2. Contract Curve Models
More structure needs to be placed on the model in order to analyse efficient contracts as employment is now an object of the bargain. Let the union’s contribution to the Nash Bargain be U = (W - W)N^ and the firm’s the
Nash Bargains over employment and wages respectively yield the following first order conditions (assuming an interior solution): r ./I 1 f R ^ R usual n = R(AN,K,X^) - WN (5.3) W = (5.4) W = r 0 1R/N + r Ii ]
I 0
+ ^J
I 0 + ^ J
r
1 1 R/N + r ^ I 1 ^ J [1
+ 13J
WSolving out for own wages gives the reduced form for employment under the Contract Curve Model:
(5.5) w(N,W,X^,W,^) = R^ +
1 + /3
As before, sign(3w/9A) = sign(9N/9A), and ' i/j - 1 ]
R/N W = 0
(5.6) du)/dA = R^/A
, 1 + p
risk aversion, i// > 1, is the rule), employment is more likely to increase under efficient bargaining than in labour demand models. This is quite reasonable: the union will take some of the rents from innovation in the form of higher employment if it is able to do this through bargaining. Positive effects are more likely the more the union values jobs (higher 0), and the more power it has (lower 13). In particular, although Proposition 1 holds under efficient contracting proposition 2 can be relaxed
Proposition 3 In Contract Carve Models, an innovât ion will increase employment if and only if
( i/j - 1 ]
The usual theoretical restriction on the sign of 3N/3W = sign{9w/3W> ^ 0 from (5.5) still holds. The various problems with using this as the test between the two models were discussed extensively in Chapter 2. Two of the main objections were that the tests are non-nested and that efficiency wage considerations generate the same predictions as efficient bargains viz the employment equation. Both of these objections are tackled in the following two sub-sections.