Labour costs and employment policy.

Hart, Robert A. ; Ruffell, Robin J.

Introduction

Western European governments have tried to tackle problems of high and persistent unemployment in recent years against a background of severe economic limits to available policy menus. Under the dominance of German macroeconomic management, tight monetary and fiscal regimes in the main European economies have severely limited active interventionist employment policies since the late 1980s. More recently, room for manoeuvre has been further restricted by efforts to meet the Maastricht criteria for entry into European Monetary Union. In the labour market, those in employment have not displayed a great enthusiasm for work-sharing, especially if this requires significant degrees of pay restraint. So can governments realistically intervene in this type of climate to price workers into jobs? The most feasible policies would seem to require the achievement of two joint goals. These are (a) reduced labour costs and (b) revenue neutrality.

A strategy of targeting reductions in non-wage labour costs would appear to be a useful approach towards meeting the first goal. Indeed, the European Commission (1994) has advised member governments to attempt to reduce such costs by the equivalent of 1-2 per cent of GDP. In this respect, employers' payroll taxes offer the most important potential intervention. Statutory welfare contributions by employers constitute one of the most significant items of non-wage labour costs in Europe and elsewhere (Hart, 1984) and, unlike most other non-wage costs, they are directly under government control. The difficulty is that reducing payroll taxes in absolute terms will not in itself be revenue neutral. One simple way to achieve neutrality would be to offset reductions in employer contributions by increases in employee contributions. Employers' cost reductions may be short-lived, however, since existing workers' subsequent wage demands may well be aimed at restoring real take-home pay. Such a reaction on the part of skilled labour, for example, may well be successful within the current relatively tight labour market conditions. An alternative strategy would be to reduce employers' payroll tax contributions in respect of workers whose job descriptions match most closely with the unemployed and impose offsetting employer payroll tax increases in respect of other workers. For example, contribution reductions for unskilled workers might be offset by contribution increases for managerial and professional workers. This may raise distributional problems related to tax incidence among industries with differing skill levels but it is a strategy that comes near to achieving the two stated goals as well as targeting the most acutely affected unemployment groups. Interestingly, it is a policy that has already been pursued - by the British government in 1985.

Payroll taxes in Britain refer to employers' National Insurance Contributions (NICs). They represent about 7 per cent of total labour costs in the 'typical' firm. There was a radical reform of the structure of NICs in the 1985 Budget that involved explicit employment creation goals. As stated by Mr. Lawson, the Chancellor of the Exchequer at the time, "the centrepiece of the 1985 Budget... was a reform of National Insurance contributions designed 'to cut the costs of employing the young and unskilled and to sharpen their own incentive to work at wages which employers can afford to pay'" (Lawson, 1992). The changes effectively reduced contribution rates for lower paid workers and increased them for the highest paid workers. Its attractiveness to government was that the measures could be revenue neutral while producing a net positive increase in employment. A 1 per cent increase in the contributions payable on the higher paid funded a 5 per cent reduction in contributions payable on the low-paid (Dicks and Robinson, 1985).

An interesting feature of non-wage labour cost reductions in general, and payroll taxes in particular, is that they may involve changes in both fixed and variable labour costs. The first are costs that are independent of hours of work while the latter vary directly with working time. As is well known in the literature (Hart, 1984; Hamermesh, 1993), changes in fixed and variable labour costs not only affect firms' levels of employment demand but also the relative use of workers and average hours per worker within total labour services. Stated simply, understanding the employment consequences of changes in payroll tax rates necessitates examining both the effects on workforce size and the work utilisation of each worker. In evaluating the employment implications of the 1985 changes, a comparative advantage of this study is that we accommodate these important and interrelated work dimensions. However, our framework concentrates purely on the demand side of the labour market. We trade-off capturing workers-hours reactions against attempts to model the supply side of the labour market and the wage bargaining process. Concentrating on the worker-stock dimension, Pissarides (1997) models and simulates the effect of employment tax cuts in competitive, union bargaining, efficiency wage and search theoretic settings. As we point out later, our estimates of employment gains due to the 1985 measures should be treated as upper bounds.

Lessons learned from analysing the 1985 measures have relevance to current policy initiatives.(1) We concentrate attention on the New Deal job scheme that affects young people. An employer who provides a part-time or full-time job to an 18-24 year old who has been employed for at least 6 months will receive a per-capita weekly fixed sum for 6 months plus a payment towards the cost of training.(2) This labour market intervention involves changes in relative fixed and variable labour costs and so links to our analysis of the 1985 measures. A further association between the New Deal proposals and the 1985 NICs changes is that the former also represent an employment policy approach of reducing employment costs while achieving revenue neutrality. (13) In one important respect, however, the New Deal and the 1985 interventions differ. The New Deal reductions represent marginal employment cost changes in that they apply only to additions to the existing workforce. NICs restructuring, by contrast, affects general labour costs since employer payroll contribution rate changes apply to both existing and new workers.

In Section 1, we describe the implications for the NICs rate structure of Chancellor Lawson's budget measures. We then discuss briefly the relative importance of NICs within total labour costs (i.e. wage and non-wage), highlighting the distinction between fixed and variable costs. Our labour demand framework is outlined in Section 3 while Section 4 presents estimated employment-hours repercussions of the 1985 budgetary changes. Section 5 advances reasons for exercising caution over our estimated employment impacts; several of these apply generally to employment policy analysis. In the light of the 1985 analysis, we present a few empirical insights in Section 6 into the employment effects of the 1998 New Deal proposals. Section 7 provides brief concluding remarks.

1. The rate structure of NICs before and after October 1985

Employed persons whose earnings are above the Lower Earnings Limit (LEL) pay so-called Class 1 primary NICs; their employers pay earnings-related Class 1 secondary contributions. We concentrate here on these latter contributions.(4) Before October 1985, employers had to pay contributions on earnings up to the Upper Earnings Limit (UEL); since October 1985, employers' contributions have also been paid on earnings above the UEL. The change in the rate structure in October 1985 is presented in Table 1. For simplicity we here concentrate on the structure for those 'not contracted out'. For those contracted out, the marginal rates were lower but the same earnings limits applied.

The rate structure prior to October 1985 had three key features:

1. Those whose earnings were below the Lower Earnings Limit paid a zero contribution.

2. Those whose earnings were between the Lower and Upper Earnings Limits paid a contribution calculated as 10.45 per cent of earnings. Extraordinarily, unlike income tax rates, this rate applied to all earnings. Thus someone whose earnings increased by [pounds]1 from [pounds]35 to [pounds]36 per week faced a rise in contribution from [pounds]0 to [pounds]3.76.

[TABULAR DATA FOR TABLE 1 OMITTED]

3. Those whose earnings were above the Upper Earnings Limit paid a fixed amount equal to 10.45 per cent of the Upper Earnings Limit.

From October 1985, the main changes were that:

1. The single rate of 10.45 per cent between the Lower and Upper Earnings Limits was replaced by a sequence of four rates rising from 5 per cent to 10.45 per cent. The applicability of the rates to all earnings was retained so that there were marginal rates over 100 per cent at each rate change.

2. Although the name 'Upper Earnings Limit' was retained, the highest rate of 10.45 per cent now applied to all earnings of those above the upper limit, i.e. their contribution changes from fixed to variable.

Thus the only workers now with a fixed contribution were those below the Lower Earnings Limit, who still paid nothing.

2. NICs in relation to total labour costs

Our labour cost data are derived from the Labour Cost Survey and our time points are conditioned by its collection periods. The two Surveys that straddle 1985 are those conducted in April of 1984 and 1988. Our cost definitions for this and for later developments are presented in Table 2. This breakdown is available by industry (at the 2-digit level) and by establishment size.(5)

Following Hart (1984), two types of cost dichotomy are identified in Table 2. First, we distinguish between [TABULAR DATA FOR TABLE 2 OMITTED] variable costs (w, bx, b) - or costs that vary directly with hours worked - and fixed costs (z, zx) - or costs that are hours independent. Second, we identify costs that are statutorily imposed (bx and zx) and costs that are privately agreed within firms themselves (w, b, z). The latter costs may be determined either through unilateral firm decision making or through collective agreement. From Table 2, we note that NICs accounted for 7.4 per cent of total labour costs in 1984 (divided into 6.3 per cent variable and 1.1 per cent fixed costs) and 7 per cent of total costs in 1988 (purely variable costs). As previously indicated, the positive fixed cost element of NICs was removed by the 1985 Budget changes.

3. NICs, labour demand and fixed costs

In this and the following sections, we examine the effects of changes in NICs on employment and hours of work. A particularly useful framework for achieving this objective is workers-hours demand models (Ehrenberg, 1971; Hart, 1984; Hamermesh, 1993). A critical feature of these models is that they accommodate the distinction between fixed and variable labour costs combined with the closely related concepts of labour demand at the intensive and extensive margins of the firm's operations. As we have seen, NICs can have both fixed and variable elements, and their labour market analysis falls naturally into this modelling framework.(6)

[TABULAR DATA FOR TABLE 3 OMITTED]

(a) Fixity and NICs

Table 3 shows, for 1984 and 1988, the percentage of NICs paid in respect of employees whose contributions were 'fixed', i.e. whose pay was such that a marginal change in pay would not alter their contribution. The figures given for 1984 are estimates obtained using the distribution of gross pay for each industry(7) and take account of variations in the proportion of workers contracted out.(8) Table 3 also shows the percentages of employees below the Lower Earnings Limit and above the Upper Earnings Limit.(9)

The inter-industry variation shown in Table 3 is due largely to the variation in the centrality and dispersion of the pay distribution and, to a much smaller extent, to differences in the proportion contracted out. The highest degree of fixity is in mineral oil processing, an industry with high pay and a high proportion contracted out; the lowest (zero) is in leather and leather goods, which has low pay and a low proportion contracted out.

(b) Labour demand framework

With reference to the cost definitions in Table 2, we define fixed costs f = z + zx and variable costs v = w + b + bx and let p = f/v. Following the analysis of the previous sub-section, NICs are split between bx and zx. Workers-hours demand equations can be derived from the cost minimisation problem:

min L(N,h,K, [Lambda]) = fN + vNh + rK+ [Lambda][Q - G(N,h,K)] (1)

where N is the number of employees, h is average hours per employee, Q is output, K is capital and r is capital cost, and [Lambda] is a Lagrangian multiplier. Minimisation produces the equilibrium demand equations

[N.sup.*] = [N.sup.*](p, Q, K) (2)

[h.sup.*] = [h.sup.*] (p, Q, K). (3)

Hart (1984) provides full derivations of this model (see also, Hamermesh, 1993 for discussion). In general, the responses of [N.sup.*] and [h.sup.*] to the factor price ratio, p, cannot be signed unambiguously in this model. More insight is gained from the analysis holding the capital input fixed. In this situation, a rise in p - i.e. an increase in the ratio of fixed to variable costs - produces an hours-worker substitution effect. At the margin, firms offset increased fixed costs attached to new hires by utilising existing workers more intensively (i.e. by increasing h). A rise in output has a positive scale effect on employment and a positive or zero scale impact on hours.(10) Note that, in relation to NICs, p can be changed through changes either in tax rates or in the LEL and UEL.

4. The employment and hours effects of restructuring NICs in the 1985 Budget

(a) Estimation

As detailed in Section 1 (see Table 1), the 1985 Budget reduced employers' tax liability in respect of low-paid employees by up to 5.45 per cent while raising tax rates from zero to 10.45 per cent on earnings above the UEL. The government's stated aim of these measures was job creation, especially among the low paid. Here, we estimate the employment effects of these changes via the demand system outlined in equations (2) and (3).

Table 4 shows the industry-level percentage changes in total NICs (for given levels of employment and earnings distributions) as well as in their fixed and variable components resulting from the 1985 budgetary measures. Large net changes in contributions affected relatively few industries. Three industries experienced [TABULAR DATA FOR TABLE 4 OMITTED] percentage NICs decreases of more than 4 per cent and 6 industries increases of more than 4 per cent. The former group of industries accounted for 11.5 per cent of the total workforce while the latter comprised 27.6 per cent. A 'typical' industry, therefore, experienced falls in contribution payments with respect to low-paid workers offset, to a large extent, by rises to (albeit fewer numbers of) high-paid workers.

We estimated logarithmic versions of the workers-hours equations given in (2) and (3) for 1984, our nearest cross-section to the Budget changes. Our key variables are given in (2) and (3). We included two further variables that are commonly included in the relevant literature (e.g. Ehrenberg, 1971): standard weekly hours, s,(11) and union density, u (trade union or staff association members as a percentage of all workers). We treat the variables p, Q and K as potentially simultaneously determined with N and b and therefore use a two-stage least squares (2SLS) estimator. Results are shown in Table 5 where we also list our choice of instrumental variables.

The factor price ratio exerts a significant positive influence on h and a significant negative influence on N. These are the expected outcomes in a model with K assumed to be fixed. The positive effect on h is obtained for the United States labour market by Ehrenberg (1971) and Ehrenberg and Schumann (1982). Output exerts an expected significantly positive influence on N though it displays an insignificant association with h. K is an insignificant addition to both equations. Taken together, the lack of association of h with either Q or K supports the notion of scale-independence of hours.(12) Since the pioneering work of Ehrenberg (1971), this assumption is commonly adopted in the workers-hours literature. Standard hours are significantly positively related to h - in line with underlying theory - but an insignificant addition in the N-equation. Finally, the union variable exerts a significantly negative effect on hours of work and significantly positive effect on workforce size. Both these latter outcomes might have been anticipated, a priori.

Table 5. 2SLS estimates of hours and workers equations 1984

 Hours equation Worker equation
 (dependent variable: In h) (dependent variable: In N)

Constant -2.30 (1.07) -9.63 (9.18)
ln p 0.048 (0.012) -0.35 (0.11)
ln s 1.67 (0.30) 2.86 (2.55)
ln u -0.0094 (0.0042) 0.18 (0.04)
ln Q -0.0079 (0.0164) 0.99 (0.14)
ln K -0.00022 (0.01544) -0.12 (0.13)
[R.sup.2] 0.40 0.94
n 130 130

Notes:

The figures in brackets are standard errors. Parameters in bold
are significantly different from zero at the 5 per cent level. n
is the number of observations, 2-digit industries. Instruments
variables are ln(bx) and tn(zx) (see Table 2) as well as the
1988 values of ln(p), ln(Q) and ln(K).

Data:

h The number of hours worked per week, from Labour Cost Survey 1984.

N The number of workers per establishment, from the Census of
Production 1984.

p Ratio of fixed to variable non-wage labour costs as defined
in the text, from Labour Cost Survey 1984.

s Standard hours, from New Earnings Survey 1984.

u Union density, trade union or staff association members as
a percentage of all workers, from Workplace Industrial
Relations Survey - unpublished data provided by courtesy
of the Policy Studies Institute.

Q Gross value added at factor cost, [pounds] per employee
per week, from the Census of Production 1984.

K Net capital expenditure, [pounds] per employee per week,
from the Census of Production 1984. (Expenditure rather than
stock had to be used as there is no source of stock figures by
establishment size for 2-digit industries.)

The estimated p elasticity from the worker equation is -0.35. From Table 4, we know that the ratio f/v (= p) declined 2 per cent with respect to all manufacturing industry following the 1985 tax changes. Combining these figures produces an estimate of a 0.7 per cent increase in employment. If we assume that these calculations are relevant to all industries then, using the UK all industry total employment figure of 21,423,000 in 1985 (Economic Trends Annual Supplement), we arrive at an estimate of 154,000 employment gain due to the tax changes.

This estimate is remarkably similar to the 'ball-park' forecasts of the time.(13) For example, the London Business School (Dicks and Robinson, 1985) produced a forecast of 150,000 new jobs created over the 5 years following the changes. Arguments were based on the aggregate workforce. Given an assumed employment/real wage elasticity of unity, it was calculated that 250,000 low paid jobs would be created, with a loss of 100,000 jobs among the higher paid.

5. Notes of caution

We should treat the foregoing employment estimates with some caution. We draw attention here to five problems. It is worth adding that several of the limitations we outline are common to the majority of related employment policy studies.

First, as can be inferred from Table 4, most industries experienced NICs increases for high paid employees that offset NICs reductions for the low paid. If positive employment changes were to take place, then these industries would have had to substitute low-wage for high-wage employees. But this may have meant substituting low skilled for high skilled workers and/or more labour intensity for less capital intensity.(14) Considerations of the net cost of such readjustments may serve significantly to offset the unconstrained aggregate employment creation calculations. A complete assessment involves the evaluation of other relevant elasticities, entailing estimation of the degree of factor substitutability and complementarity (i.e. workers, hours and capital).(15)

Second, as the Government intended, several low paid industries - for example, footwear and clothing and textiles - experienced very significant NICs reductions. Interestingly, Table 4 reveals that, for many firms in these industries, the f/v ratio change would have been positive. From the estimates in Table 5, this is predicted to result in offsetting employment reductions due to hours substitution. In other words, those low-pay industries that were targeted to achieve the most beneficial employment scale impacts of total NICs reductions were also the industries most prone to adverse employment substitution effects of changes in the ratio of fixed-to-variable labour costs. There is reason to believe, however, that these hours effects would have been very small. The elasticity of hours worked with respect to the ratio of fixed to variable costs in Table 5 is low (0.048). The predicted hours changes are in the order of -0.1 per cent in all manufacturing industries, ranging from -0.3 per cent in office machinery manufacture to 0.04 per cent in leather and leather goods.

Third, tax discontinuities with respect to earnings schedules may have encouraged employers to concentrate on the establishment of permanent low-paid jobs. For example, where practicable, there is a clear incentive for businesses to keep low-paid workers below the LEL not only because payrolls are zero tax-rated but also because the lowest tax rate applies to all earnings if the LEL is exceeded.(16)

Fourth, the dramatic increase in marginal tax rates paid by employers in respect of high-paid workers provides incentives, for both employers and employees, to evolve alternative compensation systems in order to avoid tax payments. Examples include a greater emphasis on fringe benefit payments(17) and a higher degree of profit sharing.(18)

Fifth, the analysis omits supply-side labour market considerations. From a supply-side perspective, a declared intention of the Government's 1985 NICs changes was to improve the incentive to work among young and unskilled workers. As summarised by Davis and Dilnot (1985), the changes would have acted to "alleviate the problem of the 'unemployment trap' for low-paid workers at the cost of higher marginal tax rates and increased distortion to the labour market". The supply-side 'pricing-into-jobs' incentives were intended to be reinforced on the demand side by the employment creating effects of reduced labour costs. A poor response by low-paid workers to these incentives - that is, an inelastic labour supply - would serve to reduce our demand-based employment effects. In the extreme polar case, a completely inelastic supply curve would result in no employment effect, with changes in NICs rates being reflected purely through wage changes.(19) In general, our analysis ignores offsetting wage responses.

6. Related employment measures under the New Deal

The NICs measures in the 1985 Budget were aimed at pricing young and unskilled unemployed into jobs. They were employer-based contributions which, as well as altering the occupational and industrial incidences of payroll taxes, affected ratios of fixed-to-variable labour costs. The present government's New Deal initiatives are also aimed at helping the young unemployed through cost reductions to employers. Moreover, they serve to alter firms' relative fixed and variable labour costs. Under the New Deal, employers are given [pounds]60 per week for 6 months and a [pounds]750 training grant for taking on an 18-24 year old who has been unemployed for at least 6 months and is receiving a Jobseeker's Allowance.(20) These subsidies serve to reduce fixed costs. The scheme is designed to help about 120,000 unemployed young persons. The key difference between the 1985 NICs measures and the current New Deal is that the former introduced cost changes with respect to all workers employed while cost reductions under the latter are realised only in respect of additional new hires who are eligible for subsidies.

Using the existing analytical framework, we attempted a very tentative set of calculations on the possible employment impact of the New Deal. A critical aspect of any evaluation concerns the expected length of jobs created under the scheme. As a minimum, we assume that all new jobs last for just 6 months, that is the length of time for which the weekly subsidy is available. The New Deal does not constrain employers to offer jobs for longer than 6 months. Beyond this, we assumed a range of expected lengths of new jobs up to 24 months in length.

The key variable here is the ratio of fixed-to-variable labour costs, denoted by p = f/v. Clearly, a change in employment due to take-up of the New Deal subsidy will affect p. The subsidy serves to reduce average fixed costs. Further, average variable costs will also be reduced since young trainees would be expected to earn lower wages and other variable non-wages compared to existing employees. Let N be the initial level of employment and [Delta]N the change in employment due to the New Deal. As before f is the fixed cost per existing worker while we denote the fixed costs associated with each new hire under the New Deal as [f.sub.y]. If s is the subsidy per new hire under the New Deal, then the net fixed cost of a new hire is [f.sub.y - s]. Further, let v and [v.sub.y] be the respective variable costs per head of existing workers and new hires.

The new value of p after a change in employment due to the New Deal is given by

[Mathematical Expression Omitted].

After some manipulation we can derive the proportional change in p, [Delta]p/p, in terms of the proportional change of employment, [Delta]N/N and the labour cost variables.(21) From earlier, b is the elasticity of N with respect to p. Developing the expression in footnote 21, we can then derive an expression for [Delta]N/N in terms of this elasticity and the labour cost variables. This is given by

[Mathematical Expression Omitted] (4)

using the approximation b = [Delta]N/[Delta]p [multiplied by] p/N. Given our estimate of b (= -0.35) from the previous analysis, we are then able to compute values of [Delta]N/N associated with given values of f, [f.sub.y], s, v and [v.sub.y].

The 1992 Labour Costs Survey gives figures for the breakdown of total labour costs for all manufacturing. To make these comparable with the New Deal figures for 1998, we adjusted them by the average earnings index, projecting to November 1998 using the underlying rate of increase in the series at November 1997 (= 4.75 per cent). This gives an increase of 33 per cent on 1992. The evidence of the Labour Cost Surveys is that the long-term nominal fixed costs have risen in line with nominal wages and salaries (the ratio being 10.3 per cent in both 1984 and 1992 for manufacturing). This process gave us estimates of v and f. For [f.sub.y], fixed costs pertaining to new hires, we removed the redundancy payments element from f since, under our assumptions about the expected length of new jobs, redundancy payments would not apply. We also removed so-called 'services to employees' from the new recruits' fixed costs since these do not appear to be relevant.(22) As for [v.sub.y], we assumed that new recruits would obtain 47 per cent of v, a figure derived from the pay of apprentices and trainees in the 1992 Labour Cost Survey relative to fully trained workers. The New Deal gives us an s value for full time workers of 26 x [pounds]60 = [pounds]1560 plus up to [pounds]750 for training giving a possible value of [pounds]2310.

The impacts of s, [f.sub.y] and [v.sub.y] on the value of p and hence on N depends on the lengths of the newly created jobs. The subsidy is a once-over cost reduction and its size depends strictly on the number of hires of eligible unemployed persons. Fixed and variable costs of existing workers are unaffected by the subsidy. Clearly, the maximum subsidy effects on employment will be realised if the expected length of newly created jobs is 6 months. For longer expected lengths, the relative importance of the fixed subsidy within total labour cost calculations reduces as the length of the employment period extends.

In Autumn 1997, the number of employees in employment in Britain was 22,749,000 which we use as the value of N in our calculations.(23) On the basis of (4), we can calculate the magnitude of the subsidy which is needed to employ 120,000 New Deal workers. Since the subsidy is paid for only six months the answer depends on whether the jobs are expected to last for.those six months or for rather longer. If the duration of each job is expected to be limited to the duration of the subsidy, then it turns out that the New Deal is excessively generous. We estimate that 120,000 new jobs could be created with a total subsidy of [pounds]1,680 per capita rather than the [pounds]2,310 on offer. In other words, the payroll subsidy could run at under [pounds]40 per week rather than the [pounds]60 on offer. On this basis the existing proposal is more generous than it needs to be.

However, the relationship is not proportional and the structure of our equation implies that, if the jobs are viewed as lasting for one year or longer, then the impact of the subsidy on employment is likely to be extremely small. Technically this arises because, in relation to equation (4), spreading the subsidy over a longer period very rapidly changes the fixed costs of new workers to the fixed costs of existing workers. In fact, the ratio changes by 55 per cent for a move of 6 to 12 months.

These calculations suggest that a succession of 120,000 6-month jobs could be created by a continuing subsidy of [pounds]1,680 per 6-month job and suggest that the cost of a permanent reduction to unemployment of this size is likely to be more expensive than the government anticipates.

On face value, the New Deal is a potent means of creating short-run jobs but a less convincing instrument for longer-term job creation. The intuitive reasoning is quite easily understood. In terms of jobs with an expected 6-month duration period, the subsidies to new workers have relatively big measurable impacts on fixed-to-variable costs averaged over all employees. As the job duration is increased - i.e. outside the period eligible for subsidy - the marginal employment subsidies quickly have negligible effects on average cost ratios and, therefore, on associated employment changes.

We stress that these are very tentative calculations using a critical parameter, b - the elasticity of N with respect to p - derived from a study covering a different time period and a limited economic structure. Further, as in the analysis of NICs changes, we omit such considerations as job-match effects, factor (labour-capital) substitution and supply-side influences. We also disregard the secondary effects on employment of enhanced worker quality and productivity due to the training programmes under the New Deal.

At least, the exercise illustrates an important point. If it is anticipated under the New Deal that the jobs created will be of short duration - that is, lasting for the 6 months 'subsidy-on' period - then the employment impact could be considerable. It is likely, however, that the government hopes that jobs of longer average duration will be created. The subsidies do not appear to offer significant inducements in this latter event.

7. Conclusions

The payments to employers under the New Deal measures constitute marginal employment subsidies to firms; their receipt is predicated on the employment of new workers. By contrast, the 1985 NICs changes generally applied to both new and existing workers. The comparative benefits of the current proposals may lie, therefore, in the well-known employment creation advantages of marginal compared to general subsidies (Layard and Nickell, 1980).(25) There are also problems with the NICs changes highlighted in Section 4 which are either not present (e.g. the case of tax rate discontinuities) or, perhaps, less acute (e.g. unequal incidence of cost reductions across industries) under the New Deal.

The main uncertainty of the New Deal initiative is the length of job created as a result of the marginal subsidies. Within the typical firm's labour cost structure, the form of the subsidies provided is unlikely to play a major role in influencing decisions to create jobs of expected long duration. By contrast, National Insurance restructuring can be geared to produce a cost climate that helps to price younger unemployed workers into long-term jobs.

We are grateful to Dan Hamermesh, Martin Lunnon, Mike Marland and a referee foe useful comments. Helpful information was provided by Adrian Gallop, Neil Millward, Jane Mitchinson, Clive Singleton and Ann White. None of these individuals is responsible for views expressed or for errors and misconceptions.

NOTES

(1) One of these relates to the 1998 Budget proposals. A new rate structure of employers' NICs will take effect in April 1999. Flat rate contribution of 12.2 per cent will apply to wages over [pounds]81 per week. This replaces a more complicated tiered structure which we describe in Section 1. Data limitations preclude an empirical exploration of these proposed changes.

(2) For a 6 month period, employers receive [pounds]60 per week for providing a full-time job, [pounds]40 per week for a part-time job and up to [pounds]750 towards the cost of training.

(3) Although revenue neutrality has obviously been achieved through a different mechanism under the New Deal; that is by funding employment subsidies through [pounds]3..5 billion of windfall taxes receipts from private utilities.

(4) We exclude self employed persons. They pay Class 2 and Class 4 contributions, calculated under a different system.

(5) There are 23 2-digit industries (see Table 3). There are 6 size ranges, in numbers of employees per establishment; these are (1) 10-49, (2) 50-99, (3) 100-199, (4) 200-499, (5) 500-999 and (6) 1000 and over. Combining industries and size ranges determines the number of observations in the empirical analysis in Section 4.

(6) See also Hart and Kawasaki (1988) for an application to the (West) German payroll tax system.

(7) Derived from unpublished New Earnings Survey data.

(8) Contracting out, ceteris paribus, tends to lower the proportion fixed because the gap between the 'in' and 'out' average rate widens, from zero at the LEL, as pay increases.

(9) The percentage of employees below the LEL is subject to more error than the other figures since their recording is not a requirement of the New Earnings Survey. The figures are underestimates, especially in respect of smaller business establishments.

(10) The latter effect is predicated on the chosen form of production function. Hours are scale independent if the underlying production function is homothetic.

(11) From the theory, assuming a fixed capital input, h should relate positively to s, and N negatively to s; see derivations in Ehrenberg (1971) and Hart (1984).

(12) Dropping Q and K from the hours equation has little effect on the estimated coefficients of the remaining variables compared to Table 5.

(13) In answer to a House of Commons question on how many jobs the Chancellor of the, Exchequer thought would be created as a result of the NICs Budget changes, Mr Lawson replied that he expected them to lead to significant increase in jobs over the following two years" and went on to state that outside forecasters had estimated employment increases, "some of 150,000, some of 160,000 and some higher" (Hansard, 1985).

(14) Note that several of the industries with the highest falls in the f/v ratio - such as motor vehicles, chemicals and mineral oil processing - are among the most capital intensive. It would be particularly difficult for such industries to substitute low paid for high-paid employees.

(15) See Hamermesh, 1993, for a detailed discussion of these issues. A classical example of these broader considerations is found in the research of Ehrenberg and Schumann (1982) into the policy objective of increasing the cost of overtime in order to induce worker-hours substitution in the United States. They begin by making a simple calculation of the effects of the worker-hours effects of increasing the mandatory premium from 1.5-2 times the standard rate. In the absence of other. her considerations, they estimate that this would create a 1-2% increase in employment within relevant industries. They then extend their analysis to cover the interrelated questions of capital-labour substitution, job-matching, indivisibilities etc. and find that their initial employment growth estimates are almost certainly too high. In fact, taken together, the estimated impacts of the wider considerations lead them to argue against this sort of policy initiative.

(16) As a specific example, the prevalence of low-paid part-time employment in large supermarkets may, in part, stem from a desire by employers to avoid paying wages that are in excess of the LEL. Unfortunately, it is difficult to measure these effects because the New Earnings Survey does not record, in any systematic fashion, information on employees below the LEL.

(17) See Woodbury (1983) for a highly related empirical analysis.

(18) Hart and Moutos (1994) demonstrate theoretically - via a firm/union efficient bargaining model - the potential incentive effects of changes in the payroll tax system for bargaining parties to move towards profit-sharing contracts. They show that increased contributions, for either employers or employees, will encourage an increased emphasis on profit sharing. Also, they show that profit sharing, is encouraged if, for a given aggregated level of contributions, payroll tax incidence is switched towards workers. Actually, in this latter respect, the 1985 changes were in the opposite direction. Further, they discuss the effects of changes in UEL limits.

(19) For simple analytical detail, see Hamermesh (1993, pp. 18-20). Pissarides (1997) also reviews the labour market literature on the shape of the labour supply schedule which suggests inelastic responses, especially in the case of females. It is not clear that inelastic responses can be inferred for the targeted low-paid groups in the 1985 legislation, however.

(20) For simplicity, we avoid discussion of part-time jobs, other available subsidies, and the so-called Gateway guidance and support set-up.

(21) [Mathematical Expression Omitted].

(22) Thus: [f.sub.y] consisted of employers' liability insurance, voluntary social welfare payments, benefits in kind, and vocational training.

(23) We also estimate the number unemployed in the 18-24 age group as 482, 000 in Autumn 1997 (12.5 per cent), although the New Deal is targeting only those unemployed for at least 6 months.

(24) The ratio changes by 55 per cent for a move from 6 to 12 months.

(25) Although, it is important to note that the temporary nature of the subsidy is also recognised as a major weakness of marginal employment subsidies. See Whitley and Wilson (1983) for a more negative appraisal of the impact of marginal employment subsidies.

REFERENCES

Davis, E.H. and Dilnot, A.W. (1985), 'The restructuring of National Insurance contributions in the 1985 Budget', Fiscal Studies, 6, pp. 51-60.

Dicks, G. and Robinson, B. (1985), 'A Budget for Sterling and a Budget for jobs', Economic Outlook 1984-1988, London Business School, vol. 9, no. 6.

Ehrenberg, R.G. (1971), Fringe Benefits and Overtime Behavior, Massachusetts: Heath.

Ehrenberg, R.G. and Schumann, P.L. (1982), Longer Hours or More Jobs?, Cornell Studies in Industrial and Labor Relations, no. 22, New York: Cornell University.

European, Commission (1994), 'Growth, competitiveness, employment: the challenges and ways forward into the 21st Century, White Paper (Office for Official Publications of the European Community: Luxembourg.

Hamermesh, D.S. (1993), Labor Demand, Princeton NJ: Princeton University Press.

Hansard (1985), Parliamentary Debates (Oral Answers), 77, 982.

Hart R.A. (1984), The Economics of Non-Wage Labour Costs, London: Allen and Unwin.

Hart R.A. and Kawasaki, S. (1988), 'Payroll taxes and factor demand', Research in Labor Economics, 9, pp. 257-85.

Hart R.A. and Moutos, T. (1994), 'Tax structure and the choice of compensation system', European Journal of Political Economy, 10, pp. 765-81.

Lawson, N. (1992), The View from No. 11, London: Bantam Press

Layard R. and Nickell, S. (1980), 'The case for subsidising extra jobs', Economic Journal, 90, pp. 51-73.

Pissarides, C. (1997), The impact of employment tax cuts on unemployment and wages: the role of unemployment benefits and tax structure', Centre for Economic Performance, Discussion Paper no. 361.

Whitley, J.D. and Wilson, R.A. (1983), 'The macroeconomic merits of a marginal employment subsidy', Economic Journal, 93, pp. 862-80.

Woodbury, S.A. (1983), 'Substitution between wage and nonwage benefits', American Economic Review, 73, pp. 166-82.