Annualised hours contracts: the way forward in labour market flexibility?

Bell, David N.F. ; Hart, Robert A.

Under annualised hours' contracts (AHCs), workers and management agree to the length and scheduling of working hours over a 12-month period. Such contracts have been widely seen as a potentially important way of achieving greater labour market flexibility and enhanced efficiency in work organisation. There exists very little empirical work on these contracts and this study is intended to provide insights into their British labour market role and potential. Especially for workers who are not in management or a profession, the costs of switching to AHCs are substantial. The enterprises that are likely to gain from the switch are those that (a) experience significant fluctuations in output/service demand and (b) desire to utilise plant and space more intensively over the calendar year. In this latter respect, plants incorporating complex shift operations are particularly associated with AHCs.

Introduction

The standard contract of employment of most full-time workers (excluding managers and professionals) in Britain and Europe stipulates a common pattern of weekly working hours. Individuals are required to work a five-day week and each working day is comprised of a specified number of basic hours. Where hours are insufficient to meet work requirements, additional labour service is usually provided through the mechanism of paid-for overtime working. Contracts often cover explicitly the terms and conditions of overtime working. Of course, the standard contract is not all pervading. There exists a spectrum of hours' contracts that range from relatively small variations on the standard contract to quite radical work arrangements designed significantly to increase work flexibility. One of the biggest departures--and the one highlighted in this paper--is the so-called annualised hours' contract (henceforward, AHC). Here, workers and management agree to the length and scheduling of working hours with respect to a 12-month period. In recent years, several well-known British firms have switched from standard weekly/monthly hours contracts to AHCs. There is strong evidence that a significant reason for these contract changes is that AHCs offer a higher degree of hours' flexibility both with respect to inter-temporal demand changes and the achievement of greater annual plant utilisation. But is this the start of an important trend in working time arrangements? We attempt to assess the costs and benefits of AHC arrangements and hence their prospects for growth.

The British history of AHC agreements is relatively short. They were not established until the late 1980s, following much earlier developments among several French, German and Scandinavian companies (Gall, 1996). By 2001, AHCs covered about 4 per cent of both full-time female and full-time male British workers. One of the key reasons for recent interest in AHCs derives from the fact that a number of well-known manufacturing, service and public sector organisations have adopted them, at least in respect of parts of their workforces (Incomes Data Services 1996 and 1999). BP Chemicals, Britvic Soft Drinks, Blue Circle Cement, ICI Chlor Chemicals, Grampian University Hospitals NHS Trust, Manchester Airport, United Distillers, Samsung Electronics, Tesco Stores and Zeneca are among the companies that have negotiated AHCs. Most are in the private sector, although local government also has strong representation. Between 1997 and 2001, over 65 per cent of AHCs covering male workers were in the private sector and 15 per cent were in local government. The respective figures for females are 42 and 33 per cent.

The advent of AHCs can be viewed in the context of the debate on labour market flexibility. Vickery and Wurzburg (1996) argue that in response to competitive forces, many large firms have adopted new forms of working that emphasise more efficient workplace organisation, individualisation of rewards and greater skills. This has resulted in a blurring of the distinction between functional and numerical labour market flexibility. Though AHCs change the structure of working time and therefore could be described as a form of numerical flexibility, they are rarely implemented without contingent functional reorganisation.

Where AHCs are introduced, they generally replace the standard contract. A typical AHC is one where management and workers agree on a basic number of annual, or rostered, hours. Leaving aside holidays, rostered hours may be distributed evenly over working weeks or they may be higher than average for parts of the year and lower at other times. They may also refer to rostered shift systems: we later give an example of a company with 1872 annual rostered hours designed to cover 156 12-hour shifts. As well as rostered hours, it is not uncommon for the parties to agree an annual number of reserve hours. These additional hours may be directed towards enhancing production and service flexibility beyond that achieved through rostered work scheduling. For example, they can be used as a buffer to meet unforeseen events such as short-run upward surges in demand or abnormal absenteeism due to sickness. They can also be used to provide time to cover planned events that are supportive of, but not directly tied-up with, production and service activities. Examples include training courses, planning meetings and company briefings. The relative proportion of these reserve hours varies greatly from company to company. While exact requirements are usually not known a priori, it is important that the firm is able to formulate reasonably precise estimates of the length of reserve hours. This is because agreements often guarantee payment whether such hours are worked or not. Under AHCs, wages are paid weekly or monthly in the same way as under the standard contract.

What are the perceived benefits of AHCs? Bell and Hart (2002) list a wide range of potential benefits accruing to workers and firms operating under AHCs. Among these, we highlight the following. On the workers' side, risk averse individuals may perceive benefits from guaranteed annual hours since these provide less income variability compared with, say, weekly hours that include significant overtime working. They would also be expected to receive a higher hourly wage rate as compensation for agreeing to provide abnormal work scheduling within rostered hours and/or reserve hours at short call. Workers' marginal utility of leisure is likely to vary over the day/week/year. Annualised contracts may better match these preferences, thereby enhancing job satisfaction and commitment. For example, workers may be happy to work relatively long hours in autumn/ winter/spring thereby concentrating more leisure time in the summer period. On the firm's side, annual hours may be better geared to matching the timing of specific production requirements over the work year (e.g. implementing a continuous production process, minimising inventories of intermediate inputs and finished goods, meeting regular seasonal patterns of demand). AHCs may permit a better annual utilisation of plant--with associated gains in capital and organisational efficiencies--by extending the number of working days in a given 12-month period. They may also facilitate the use of complicated shift working systems since annual hours can be scheduled to accommodate complex working time rotations. Reserve hours may be used to facilitate systematic employee training programmes, thereby enhancing human capital. Finally, through the use of AHCs, firms may be able to eliminate a culture in which 'guaranteed', or institutionalised, overtime is habitually worked.

In assessing the potential contribution of AHCs, we evaluate their main costs and benefits. Our principal data source is the British Labour Force Survey (LFS). Details are given in the Appendix together with an explanation of why, in large part, we use information obtained only from direct (and not proxy) LFS respondents. Apart from standard contracts, we concentrate on four additional types of contract for comparative purposes. The first two of these are structurally very similar to the standard contract. The four and a half day week, is a simple variant of the standard contract that usually involves finishing work early on Fridays. Closely related is the nine-day fortnight that involves workers having one day off every other week. (The actual day may vary but the alternation of a four-day week following a five-day week does not.) Term-time working typically involves school and college teachers with work time designed to cover students' prescribed terms or semesters. Under flexible working hours' contracts, workers can vary their daily start and finish times. Working time is subdivided into accounting periods--usually four weeks or calendar months--and debit or credit hours can be carried over from one period to another.

When discussing the incidence of AHCs in the following section, we deal with both males and females. For space reasons, we show results only for male workers in the analysis presented in subsequent sections. We report briefly on equivalent female results.

Incidence and job characteristics of AHCs

Charts 1-4 deal with the incidence of AHCs and, so as not to underestimate their relative importance, we include LFS responses from both direct and proxy respondents.

[GRAPHICS OMITTED]

Charts 1a and 1b show, respectively, the proportions of the female and male workforce under the main working agreements for the period 1994Q1 to 2001Q1. We divide standard contract workers into those who report working zero overtime and those working paid overtime. The standard contract is clearly dominant, accounting for over 80 per cent of male and 70 per cent of female workers throughout the period. By contrast, AHCs accounted for less than 5 per cent of both groups of workers. The proportion of workers with AHCs fell during the mid-to-late 1990s but has risen slightly in recent years.

Due to an LFS reclassification of occupations in 2001, the remaining charts and tables on the incidence of AHCs--together with most of our analytical work--concentrate on seven quarters of data (observed at twice-yearly intervals--i.e. quarters 1 and 3) from 1997Q1-2000Q1. Over all industries and occupations during this period, 3.5 per cent of females and 3.7 per cent of males worked under AHCs. The respective percentages in the manufacturing sector are 2.7 and 4.1 per cent. By contrast, 74 per cent of females and 83 per cent of males worked under a standard contract over this time.

Chart 2 displays the occupational breakdowns of females and males working annualised hours. The highest incidence is within professional occupations, where nearly 10 per cent of females and 6 per cent of males work on AHCs. Elsewhere, there is a reasonably equal incidence of between 2 and 3 per cent across all occupations. As for industries in chart 3, the highest incidence of annualised hours for males is in the extraction industry at 6.5 per cent and for females in the industry labelled 'other' at slightly less than 5 per cent. The latter industry is dominated by educational and health services. Note also from chart 3 that, as with occupations, there is a reasonably strong incidence of annualised hours across industrial sectors, covering at least 2 per cent of workers in virtually all industries. As for the sectoral incidence, chart 4 reveals that local government and university sectors lead the way in percentages of both genders under annual contracts. However, these charts are not weighted by industry size. In fact, 67 per cent of males and 43 per cent of females under annual contracts work in the private sector.

Table 1 provides information on hourly pay, average hours and degree of unionisation of (a) all occupations and (b) occupations excluding managerial/professional workers, who are often explicitly excluded from AHC agreements. A problem arises when comparing hourly pay and hours of work between AHC and other workers. Each individual in the LFS is observed over five consecutive quarterly waves but data on pay are collected only in wave 5. While weekly or monthly remuneration of AHC workers tends to be paid in equal increments throughout the year, agreed actual hours worked may be unevenly distributed. Thus, unlike standard contracts, recorded per-period pay may not fluctuate in line with hours worked. In order to counter this potential bias, we limited our sample to individuals classified as AHC workers for the entire 5-wave period. Their hours and hourly pay were then calculated with respect to hours averaged over three (and sometimes five) quarters. This allows us to 'average-out' variations in hours over the work year due to uneven work scheduling. For consistency, we also followed the same averaging procedure with respect to unpaid overtime of AHC workers.

Average hourly pay for workers under annualised hours is higher than under standard contracts. The wage differentials are particularly large for female occupations including managers/professionals and for males excluding managers/professionals. For the latter group, hourly wages are about 26 per cent higher than in the standard contract without overtime and about 15 per cent higher than the standard contract including overtime. Note that this male differential is considerably narrowed when managers and professionals are added. There are two reasons for this. First, average wages for male managers/professionals under standard contracts do not differ greatly from their AHC equivalents. Second, managers/professionals under AHCs work significant unpaid overtime compared to these groups under standard contracts (without overtime). Clearly, unpaid hours serve to deflate the average hourly wage by entering the denominator but not the numerator of the hourly wage calculation (see Bell and Hart, 1999).

Only the 'nine day fortnight' and flexitime contracts offer hourly rates comparable to AHCs. The data do not allow us to distinguish between actual and effective hourly rates under annualised hours. Since, on average, some reserve hours are paid for but not worked, the effective rate per hour worked is almost certainly higher than shown under the annual contract. The degree of underestimation is probably slight, however. Of course, these statistics almost certainly reflect differences in occupation, skill and industry, factors we control for subsequently.

Weekly hours excluding overtime under AHCs are roughly comparable to those of standard contracts that involve no overtime. Note, however, that excluding managers and professionals, male total AHC hours are more than male standard hours under standard contracts with overtime hours, the latter involve considerably more weekly overtime. Total weekly hours (i.e. standard plus overtime) under the standard contract with overtime exceed total AHC hours (44.5 compared to 41.2). The latter observation is also true for comparable females but the gap is much narrower.

The switch from a standard to an annual hours contract invariably requires that management and workers involve themselves in complex negotiations over time and work organisation together with appropriate pay scales. It would be expected that such agreements are more easily undertaken if the workforce is unionised. If a firm perceives advantages in adopting an AHC, then a necessary condition for it to alter its working time practices is the presence of a unified bargaining mechanism with its workforce. The complexity of agreeing to particularly changes in work schedules under AHCs requires that each party possesses the ability to deliver agreement on radical departures from standard practices on behalf of its constituency. We might expect that the more unionised the workforce, the better the chances of the parties to effect the required changes. From table 1 we see that there is considerably more union representation of males with annual contracts than with standard and most other contracts. For females, unionisation under AHCs is also higher than under standard contracts while equal to or lower than other contracts. We note from later results (see table 4), that the probability of undertaking an AHC is increased if the firm is unionised.

We also conducted three probit regressions (Bell and Hart, 2002) into the effects of type of contract and other variables on the length of employment, job search and job status. They indicate that, relative to a standard contract, jobs held under AHCs are characterised by longevity (with significantly more employees in the same job for five years or more), stability (relatively insignificant numbers looking for a new job), and permanency (significantly more workers with permanent job status). Further, we undertook a multinomial logit regression analysis designed to examine non-permanent job status--covering seasonal work, contract length, agency temping, casual work and other temporary employment. We found that AHCs are relatively weakly negatively associated with working time agreements that represented limited contract duration and casual employment.

Costs

Apart from the transitional costs of switching contracts, an AHC in the steady state would be expected to involve relatively high hourly wage costs. Anticipating later findings, suppose that an AHC is designed to provide greater flexibility in the scheduling of hours over the working year. This may be attained, for example, by splitting hours into a regular core component and a flexible reserve component. In this event, it can be shown formally that--within the context of an efficient bargain between management and workers--agreement will be reached in which the hourly wage is higher and per-period hours shorter under an AHC compared to a standard contract (Bell and Hart, 2002). In effect, a union increases its share of the rent by giving management more degrees of freedom to schedule hours outside the typical workweek arrangement under the standard contract. This rent extraction is predicted to take two forms. Not only would the average hourly wage rate be expected to rise but also the total number of per-worker hours in the working year would be expected to fall in an AHC compared to a standard contract. Combining these two effects, average hourly wage earnings would rise.

In order to test more systematically for hourly earnings differences, we estimated Mincer-type wage equations in which the dependent variable is average hourly earnings. Hours in the wage calculation include basic hours, paid overtime and unpaid overtime. We differentiate between occupations including and not including managers and professionals. There are two reasons for this. First, while the incidence of AHCs is relatively high among professionals and managers (see chart 2), many AHC agreements are struck only with respect to workers outside of these two groups. In fact, the great majority of the case studies presented by Incomes Data Services (1999 and 2002) exclude managers and professionals. Second, although subject to collective bargaining agreements, AHCs under consideration here concern hours' scheduling determined by the employer. Other annual hours' configurations may entail significant elements of self-determined hours scheduling (albeit within boundaries laid down by the employer). In particular, managers and professionals may have substantial degrees of freedom over how and when they allocate their hours. To the extent that such arrangements are coded as AHCs in the LFS questionnaires, they may serve to introduce measurement error into the analysis.

Results are shown in table 2. Data refer to full-time male workers. The omitted category in the dummies that describe contract type is the standard contract with zero overtime. Other variables are familiar to this type of equation. They include (a) the human capital variables tenure (length of time in the current job) and experience (current age minus age when completed full-time education) as well as pre-work educational attainment, (b) union status, (c) company descriptors (size, sector), (d) household characteristics (marital status, children), (e) other controls (industry and year).

Taking all occupations together, the average hourly wage under an AHC is estimated to be (statistically) the same as under a standard contract without overtime. However, excluding managers and professionals, the average hourly wage is found to be significantly higher--in fact, 12.9 per cent higher--than the standard contract without overtime. Reasons for the differences with and without managers/professionals are discussed in relation to table 1. When managers and professionals are excluded, the wage under the standard contract with overtime is 1.6 per cent higher than the excluded contract. So, excluding managers/professionals, the average hourly wage under an AHC arrangement is estimated to be around 11 per cent higher than the standard contract with overtime.

There exists little comparative evidence for these wage estimates. Interestingly, however, we know that our estimates that exclude managers and professionals are representative of at least one specific case study (Incomes Data Services, 1999). ICI Chlor Chemicals introduced AHCs for all employees in 1993. Employees who were previously weekly-paid received a 14 per cent increase in basic salary in return for agreeing to more flexible work practices under the AHC agreement.

In general, our remaining results in table 2 conform to findings of other studies into wage determination. Experience, tenure and size of company exert a positive influence on the wages. Given that the excluded education dummy refers to degree-level qualifications, our education dummies are significantly negative. Single males earn less than their married, separated or divorced counterparts. Finally, hourly earnings of males with no children or children over the age of five are lower than for males with young children under five.

We also estimated equivalent regressions for females. In general terms, results are quite similar to males. In respect of estimated wage effects, females wages under AHCs were estimated to by 10 per cent higher than under the (excluded) standard contract with zero overtime including managers/professionals--and 20 per cent higher excluding managers/professionals--but with relatively low t-ratios (1.57 and 1.12, respectively).

Benefits

In the introduction, we listed a number of potential benefits to management of workers converting to annual contracts. For some of these, such as the provision of more training programmes and the elimination of guaranteed overtime, Bell and Hart (2002) find little supporting evidence for substantial benefits. In the latter case, guaranteed overtime is clearly an important phenomenon (see also Bell and Hart, 2003). Moreover, Incomes Data Services provides case studies supporting the notion that some firms have regarded AHCs as providing the means to move away from an overtime culture. But, this explanation alone is insufficient inducement for making a switch in contract. Paid overtime is rarely available to managers and professional workers. While the high marginal costs of overtime can be eliminated, expected average hourly wages are significantly higher for these workers under AHCs (see the discussion in relation to table 2). In other words, there would be an expected total payroll increase if overtime workers under a standard contract were to switch to an AHC.

It is in two other areas--dealing with demand variation and better plant utilisation--that real benefits are most apparent. We concentrate on these.

(1) Demand variation

One possible factor associated with the adoption of annual contracts is that of demand fluctuation. We might expect that AHCs are associated with companies that experience a relatively unpredictable flow of demand for their products. The argument is relatively simple. Consider a manufacturing firm that carries out make-to-stock production. It faces a well-established and systematic demand pattern throughout the working year and gears its production activity to maintain an inventory of finished goods that is adequate to meet the regular flow of products leaving the factory. A priori, this type of production experience would be expected to be associated with a standard contract (with little or no overtime), or one of its close derivatives. There would be no reason to have recourse to a compensation system that emphasised premium payments in order to meet irregular work demands. Alternatively, consider a company with production scheduling that is geared to make-to-order. Again, if the flow of orders is predictable, exhibiting relatively low variation, it may be able to meet demand at lowest cost via a standard contract. However, if unforeseen fluctuations in demand occur quite commonly then it may have to pay for additional hours on a regular basis. This could involve the use of paid overtime and/or the employment of temporary workers and/or the use of subcontractors. Each method involves additional costs. If unexpected demand variations are particularly large and persistent then the firm may seek to find a permanent solution through an AHC that incorporates significant year-by-year reserve hours. A further cost consideration involved with standard overtime contracts concerns the voluntary nature of paid overtime. Under standard contracts, shortfalls in labour supply may occur if the proportions of workers willing to work extra hours do not satisfy given short-term demand increases. Under AHCs, reserve hours apply to all contracted workers. For example, Matsushita Electric UK (Cardiff plant) cited as one of their reasons for implementing an AHC in 1992 as having sufficient working hours to meet peak levels of production (Incomes Data Services, 1999).

Demand fluctuations associated with the adoption of AHCs may not necessarily be unanticipated, however. Where firms face uneven product demand due to seasonal fluctuations, they may also wish to seek AHC agreements. Reserve hours would be expected to play a particularly important role in this instance. Workers may well be happy to work longer hours for part of the year and shorter elsewhere if these fit better than even-length workweeks with their leisure preferences. For example, the firm General Domestic Appliances (Peterborough) implemented an AHC in 1992--involving over one thousand manufacturing employees (60 per cent of the workforce)--and gave as its main reason for switching to annual hours that it could devise a system that reflected its seasonal demand pattern (Incomes Data Services, 1999).

Here, we discuss two approaches towards studying relationships between output fluctuations and types of contracts. First, we establish whether or not AHC contracts are associated with relatively large changes in contractual working hours over the working year. Second, we attempt to ascertain whether AHCs are associated, specifically, with unanticipated output fluctuations.

The first part of the investigation takes advantage of the limited panel aspect of the LFS data. If AHC firms were more prone to output shocks and/or to seasonal demand fluctuations, we would expect workers in such firms to experience greater than average per-period hours' changes over the working year. Respondents answer working time and other questions for up to five successive quarters. The LFS consists of overlapping cohorts of interviewees with each cohort surviving up to five quarters. The information on contract type is gathered in quarters one, three and five while working time questions are asked in every quarter. If an individual reported that he/she worked on, say, an AHC in quarters one and three, we infer that an AHC also applied in quarter two and so we used the working time information for this middle quarter. The same procedure was applied to quarter four if the individual reported working under the same contractual arrangement in quarters three and five. Two types of LFS working time questions were used to measure hours' variation. First, we used quarter-on-quarter differences in actual standard hours worked. Second, we took advantage of the LFS distinction between, in any given quarter, actual standard hours and usual standard hours worked.

Based on male direct respondents, we then constructed four measures of hours' changes for each individual for the time observed on the same contract. These were:

(i) the mean absolute difference in actual hours worked from one quarter to the next;

(ii) the maximum absolute difference between successive quarters;

(iii) the mean absolute difference between actual standard hours and usual standard hours within the same quarter;

(iv) the maximum absolute difference between actual and usual standard hours.

Table 3 shows the results of OLS regressions of each of the above measures of hours' changes on contract type as well as controls for industry and company size. The coefficients on the three contracts shown are expressed relative to the excluded standard contract (with and without overtime). Firms with AHC contracts are significantly more associated with (all four measures of) hours' variation than the standard contract while flexitime and the four-and-a-half day week display a significantly weaker association. To the extent that hours' variations are caused by output variations then this evidence points firmly towards this influence on the formation of AHC agreements. But does the hours' variation stem from anticipated or unanticipated output movements?

Establishing that AHCs are associated with relatively strong contractual annual hours' fluctuations tells us nothing about whether these derive from anticipated or unanticipated economic events. In order to investigate the role of the latter, we examined the associations of contract type and 'surprise' output fluctuations within manufacturing industries. We obtained gross value added for manufacturing industries from 1986 to 1997--supplied by the Office for National Statistics--at the level of 3-digit 1992 industry codes. We used data on 228 industries. Output variation was measured as the standard deviation of the deviation from trend of deflated gross value added calculated over the period 1986-97. Individuals were matched to the relevant industry code. For regressions not including a union variable, the data relate to seven quarters (i.e. Q1 and Q3 of each year). We tested for union effects in the manufacturing using Q3 data--the one quarter for which union information is available.

Manufacturing industry results, based on multinomial logit regressions, are shown in table 4. The first three columns deal with our full sample: that is, they cover seven quarters (although they exclude the union variable). Relative to the excluded standard contract (with and without overtime), output variation has a significant impact on the probability of working under an AHC. By contrast, and unsurprisingly, output fluctuation significantly reduces the probability of adopting 4 1/2 day week contracts relative to a standard contract. It has no differential effect on flexitime relative to standard contracts.

In the reduced sample, shown in the last three columns of table 4, we find that unionisation significantly increases the probability of working under an AHC. Note, however, that the output variable is no longer significant in the AHC column using this reduced sample.

We summarise briefly the remaining key findings in tables 4. We find that the probability of working an annualised hours contract, relative to the base contract, is

(a) positively related to company size;

(b) positively associated with male employees;

(c) higher for workers with lower educational attainments.

We also conducted similar regressions--though without output fluctuation--for service industries (Bell and Hart, 2002). The contributions of unionisation and company size in the service sector were very similar to the results reported in table 4.

(2) Workplace utilisation

Association between AHCs and output variation is one aspect of hours' flexibility. More important, we believe, is the fact that AHCs offer enhanced workplace utilisation throughout the working year. In this dimension, hours' flexibility stems from the fact that AHCs more readily accommodate unconventional work patterns that are designed to provide enhanced utilisation of plant, machinery and space. Some firms are willing to trade-off the higher fixed costs of guaranteed rostered and reserve hours for the ability to achieve improved plant utilisation (e.g. complex shift systems) or continuous service delivery (e.g. hospitals). The claim has been made by several companies that planning operations over a 12-month period permits detailed organisation of complex shift working schedules designed to (a) ultilise plant efficiently and (b) maximise individuals' leisure intervals between shifts. Three examples are among the case studies covered by Incomes Data Services (1999). First, Britvic Soft Drinks (Widford Factory) introduced annual hours in 1997. Total annual hours (running from April to April) are 2,100 and are made up of 1872 rostered hours and 228 reserve hours. Annual hours were selected on the basis of annual production time, with an allowance of 15 per cent of this time to cover changeovers and basic maintenance. The rostered hours cover 156 12-hour shifts. Workers are 'rostered off' for eleven weeks with a further 'floating' week deducted from the roster for a selected week. Second, Devro-Teepak (biomedical and food processing applications), introduced an AHC in 1995. It was designed to extend the working year from 338 to 355 days by working through the annual summer shutdown and to moving from a four-shift to a five-shift continental system. Third, Holliday Dyes and Chemicals implemented an AHC in 1999 in order to reduce the number of workers on site, introduce teamwork and a move from five to seven-day production.

Our analysis of the determinants of shiftworking is based on probit regressions and is shown in table 5. Since no earnings variables are included, the sample size is substantially enhanced. The results indicate that, relative to a standard contract (with and without overtime), shiftworking is significantly positively associated with AHC arrangements. As we would expect, flexitime and 4 1/2-day week contracts also display significantly less shiftworking. In a more detailed breakdown of shiftworking, Bell and Hart (2002) show that the AHC contract is particularly associated with the more complex continental shift system, a finding consistent with several of the case studies reported by Income Data Services (1999 and 2002). Continental shifts denote three-shift systems that rotate rapidly--e.g. three mornings, then two afternoons, then two nights. Such shifts are especially designed to maximise the utilisation of plant on a near-continuous basis. (We also carried out these shift work regressions for females, but found no significant association with AHCs. One problem here is that relatively few females work shifts and we had to deal with relatively very small samples.)

The picture that emerges concerning shift working is that AHCs appear to be suited to accommodating complicated male shift patterns that involve constantly changing configurations of work attendance through time, interspersed with compensating long (and, often, uneven) breaks from work. To the extent that this is true, we might expect that workers under AHCs would be more likely to experience work weeks in which they worked clusters of shift work days that are of significantly less duration than continual weekly work periods experienced on standard and other contracts. This expectation is confirmed by the multinomial logit regressions presented in table 6. Compared to standard contracts, AHC workers display a significantly greater propensity to work between 1 and 4 different days per week. (Of the other contracts, only job-share workers display this tendency.) By contrast, their propensity to work on six or seven different days does not differ from standard contract workers. Many workers under AHCs are more likely to work complicated rotating shift patterns with each shift applying over a relatively short span of days, interspersed with relatively long rest periods.

The attempt to utilise plant, equipment and space more intensively over the work year may also be expected to reflect AHC agreements that involve greater work cover during statutory holidays and weekends. Evidence that AHC contracts help firms to utilise more working days in the year in these directions is provided in two probit regressions shown in table 7. Compared to the standard contract, AHC workers are significantly more likely to work during bank holidays and on Sundays. Note that these findings are independent of shift working effects since the latter variable is one of the controls in these regressions. (As in the case of shift working, the female-equivalent regressions to those in table 7 revealed no significant associations with AHCs.)

Assessment

Annual hours' arrangements have drawn increasing interest in the general debates concerning labour market flexibility and working time. There is undoubtedly a link between AHCs and the achievement of hours' flexibility. The fact that a number of major manufacturing and service companies as well as public sector enterprises have introduced AHCs in recent years has served to stimulate interest in their potential for achieving a more adaptable labour market. However, the take-up of such schemes has been relatively modest, although there is some indication of recent trend increases (see charts 1a and 1b). The evidence points to a limited likelihood of future growth. Undoubtedly, the costs of switching from standard contracts to AHCs may be quite substantial, especially (given the findings in table 2) for the set of workers that excludes managers and professionals. Costs include:

(a) time and money costs of evaluating the most suitable AHC arrangements;

(b) detailed employer-worker negotiations--generally relying on formal union bargaining agreements--over implementation of the chosen scheme;

(c) a mark-up in average hourly pay for workers, (excluding managers and professionals) that is probably in excess of 10 per cent;

(d) likely increases in fixed employment costs since, in the majority of negotiated contracts, payment for annual rostered and reserve hours is guaranteed.

What type of firm is willing to incur such costs? The pursuit of greater hours' flexibility almost certainly lies at the heart of many decisions to move to annual hours' settlements. We identify three types of enhanced hours' flexibility offered by AHCs.

(i) Where product or service demand varies systematically over the working year, as within firms prone to seasonal demand fluctuations, AHCs may be used to provide hours' supply that better matches the variations in hours' demand.

(ii) Through holding reserve hours that can be called on at short notice, AHCs offer firms the ability to respond more effectively to unexpected fluctuations in product and service demand. The inbuilt advantage compared to conventional overtime working is that the supply of reserve hours are guaranteed across all workers and not subject to the whims of individual supply decisions.

(iii) Planning hours over the entire working year offers firms the ability to design and plan complex shift and other working time patterns that facilitate a fuller year-round utilisation of plant and other work facilities.

Future moves towards AHCs will be concentrated among organisations that perceive the benefits of one or more of the above advantages to outweigh the high costs of implementation. In such cases, AHCs offer great potential towards achieving enhanced operational efficiency through greater labour flexibility. Where future switches towards AHCs occur, they are likely to involve relatively large companies (see the results in table 4). It is here, especially, where scale economies and organisational efficiencies are most likely to derive from the introduction of complicated work schedules.

Appendix: Labour Force Survey (LFS)

This is a major survey carried out by the Social Survey Division of the Office for National Statistics (ONS) and provided by the ESRC Data Archive. It is a household survey. Since spring 1992 the survey has been carried out quarterly, each quarter containing data on 150,000 individuals. Of these, 80,000 are in employment and 60,000 are in full-time employment. Information on earnings is also available for 8000 people in 1995, this number rising to 15,000 in 2002. Data are collected quarterly from March to May (Spring), June to August (Summer) September to November (Autumn) and December to February (Winter). Each quarter consists of five waves. Each wave is interviewed in five successive quarters, such that in any one quarter, one wave will be receiving their first interview, one wave their second and so on. This gives an 80 per cent overlap in samples from one quarter to the next. Also it is possible for any one individual to be observed in five successive quarters. Earnings data have been available from waves 1 and 5 since 1997Q1. We restrict the earnings information to wave 5 to maintain compatibility with earlier data and to provide independent observations for the analysis.

The data used here fall into three groups.

(a) 1994Q1-2001Q1 (see charts 1a and 1b): Because the information on contract type is only available in the Spring and Autumn quarters, these data are from the Spring of 1994 to the Spring of 2001--fifteen quarters in all. To avoid duplicating individuals, only wave S is used from each quarter.

(b) 1997Q1-2000Q1 (all charts and tables excluding (a) and (c)): in these data, seven quarters were used, starting in Spring 1997 through to Spring 2000. Again only wave S data were used.

(c) 1997Q1-2000Q1 panel data (table 3): all the quarters from Spring 1997 to Spring 2000 were used. However the data were confined to people who had not changed their job during this period and who had the same contract type throughout the period. By doing this it was possible to use the information on hours of work from the two intervening quarters i.e. summer and winter. In the majority of cases this meant hours of work information in three successive quarters and for a much smaller number, five successive quarters.

Proxy respondents and the Labour Force Survey

An important feature of the LFS is that it allows interviewers to record answers to questions by proxy if a respondent is unavailable. Proxies account for about 30 per cent of LFS responses. A proxy response is usually from another related adult who is a member of the same household, although there are exceptions to this rule: (i) a young person, of the same household, may translate for a non-English speaking relative; (ii) a carer, of the elderly or infirm, although not related, may answer for someone in their care; (iii) anyone can respond by proxy with the personal permission of the head of household or spouse.

It may well be the case that, in general, the quality of information provided by proxy respondents is inferior to that given by direct respondents. Accordingly, we have taken the risk-averse strategy, for most of the analytical work reported here, of incorporating only the data provided by direct respondents. We indicate where the exceptions occur; for example, in the early descriptive material.

Measuring hours of AHC workers

Hourly pay for AHCs is calculated using hours averaged over at least three quarters, up to a maximum of five quarters. If an individual was AHC in quarter 1 and quarter 3 and did not change job, it is assumed that they were also AHC in quarter 2 and therefore we know their hours in that quarter. For all other contract types, the hours are those as given in wave 5, i.e. the same quarter that information on pay is recorded. We excluded all AHC respondents observed for only one quarter.

Table 1. Wages, hours and unionisation by working time
contracts, 1997Q 1-2000Q1

 Annualised Standard Standard
 hours contract contract
 without with
 overtime overtime

Males (including managers and professionals and excluding proxy

Hourly pay 10.02 9.42 9.31
Weekly hours (excl. overtime) 40.20 41.79 38.36
Paid overtime hours 1.30 0.00 3.93
Unpaid overtime hours 2.48 0.00 3.08
% union representation 73% 24% 34%

Females (including managers and professionals and excluding proxy
respondents)

Hourly pay 9.91 6.60 7.75
Weekly hours(excl, overtime) 38.10 37.77 36.90
Paid overtime hours 0.58 0.00 1.73
Unpaid overtime hours 6.55 0.00 3.31
% union representation 70% 25% 33%

Males (excluding managers and professionals and excluding proxy
respondents)

Hourly pay 8.62 6.90 7.38
Weekly hours(excluding overtime) 41.51 40.84 38.79
Paid overtime hours 1.51 0.00 5.71
Unpaid overtime hours 0.343 0.00 0.95
% union representation 72% 27% 40%

Males (excluding managers and professionals and excluding proxy
respondents)

Hourly pay 6.21 5.63 6.16
Weekly hours(excluding overtime) 38.15 37.26 36.92
Paid overtime hours 1.74 0.00 2.38
Unpaid overtime hours 0.77 0.00 1.48
% union representation 38% 20% 26%

 Four and a Nine day
 half day fortnight
 week

Males (including managers and professionals is and excluding proxy

Hourly pay 8.56 10.51
Weekly hours (excl. overtime) 37.50 37.13
Paid overtime hours 3.11 3.64
Unpaid overtime hours 0.99 0.47
% union representation 45% 70%

Females (including managers and professionals and excluding proxy
respondents)

Hourly pay 6.18 8.77
Weekly hours(excl, overtime) 37.27 36.90
Paid overtime hours 1.40 1.15
Unpaid overtime hours 0.47 1.47
% union representation 40% 61%

Males (excluding managers and professionals and excluding proxy
Hourly pay 7.31 9.87
Weekly hours(excluding overtime) 37.60 37.50
Paid overtime hours 3.51 3.92
Unpaid overtime hours 0.17 0.21
% union representation 51% 76%

Males (excluding managers and professionals and excluding proxy
respondents)

Hourly pay 5.69 6.75
Weekly hours(excluding overtime) 37.39 36.39
Paid overtime hours 1.37 1.56
Unpaid overtime hours 0.12 0.33
% union representation 38% 48%

 Term-time Flexitime
 working

Males (including managers and professionals is and excluding proxy

Hourly pay 9.44 10.37
Weekly hours (excl. overtime) 37.92 37.73
Paid overtime hours 0.51 1.41
Unpaid overtime hours 8.01 1.98
% union representation 78% 45%

Females (including managers and professionals and excluding proxy
respondents)

Hourly pay 8.04 7.89
Weekly hours(excl, overtime) 36.32 36.78
Paid overtime hours 0.12 0.80
Unpaid overtime hours 6.82 1.11
% union representation 69% 47%

Males (excluding managers and professionals and excluding proxy

Hourly pay 5.21 7.82
Weekly hours(excluding overtime) 37.64 37.98
Paid overtime hours 1.21 2.10
Unpaid overtime hours 0.71 0.61
% union representation 31% 44%

Males (excluding managers and professionals and excluding proxy
respondents)

Hourly pay 5.56 6.58
Weekly hours(excluding overtime) 33.80 36.77
Paid overtime hours 0.23 0.80
Unpaid overtime hours 1.02 0.42
% union representation 43% 44%

Note: See Appendix for calculation of AHC pay and hours.

Table 2. Determinants of hourly wages (full time
males): LFS 1997Q1-2000Q1

Dependent variable: log hourly earnings.
Excluding proxy respondents

 With managers
 and professionals
 Coefficient t-stat.

Union -0.049 -3.62
Tenure 0.025 13.75
[Tenure.sup.2] 0.000 -8.14
Experience 0.006 9.57
[Experience.sup.2] 0.000 -3.39
Company size >50 0.153 12.63
Private sector -0.015 -0.83
Annualised hours 0.038 0.76
4 1/2 day week -0.010 -2.95
Flexitime -0.018 -0.91
Standard (with o/time) -0.014 -1.10
Other higher
 education -0.253 -13.90
A-level, Highers -0.346 -16.82
O-levels -0.397 -24.25
YT certificate -0.470 -19.59
Any other prof. quals. -0.504 -23.66
Married 0.1454 9.66
Separated 0.189 5.88
Divorced O.120 4.91
Widowed -0.002 -0.03
Youngest child
 aged 5-11 -0.039 -1.83
Youngest child
 aged 12-18 -0.042 -1.84
No children -0.070 -3.98
Sample size 5323
F statistic 80.8
Adjusted [R.sup.2] 0.34
Other controls industry, year

 Without managers
 and professionals
 Coefficient t-stat.

Union 0.051 2.29
Tenure 0.022 10.48
[Tenure.sup.2] 0.000 -5.55
Experience 0.003 3.71
[Experience.sup.2] 0.000 -3.12
Company size >50 0.080 5.52
Private sector -0.009 -0.21
Annualised hours 0.121 2.20
4 1/2 day week -0.077 -1.92
Flexitime 0.042 2.01
Standard (with o/time) 0.016 2.64
Other higher
 education -0.102 -3.67
A-level, Highers -0.230 -7.65
O-levels -0.210 -7.76
YT certificate -0.215 -7.02
Any other prof. quals. -0.291 -9.64
Married 0.139 7.71
Separated 0.174 4.55
Divorced 0.114 4.26
Widowed 0.071 1.01
Youngest child
 aged 5-11 -0.035 -1.36
Youngest child
 aged 12-18 -0.049 -1.83
No children -0.086 -4.12
Sample size 2816
F statistic 32.1
Adjusted [R.sup.2] 0.27
Other controls Industry,year

Notes: The excluded variables are non-union, company size<50, public
sector, standard contract with zero overtime, degree, single,
youngest <5.

Table 3. Difference in hours regressions (panel data,
males) LFS 1997Q1-2000Q1 (OLS regressions)(a)

 Mean quarter on Maximum difference
 quarter difference in in quarter on quarter
 basic hours basic hours

 Coefficient t Coefficient t

Annualised hours 1.113 263 1.153 1.78
Flexitime -0.406 -1.91 -0.744 -2.29
4 1/2 day week -0.763 -2.42 -1.379 -2.87
Sample size 4848 4848
[R.sup.2] 0.0202 0.0114
Other controls Industry, company size Industry, company size

 Mean difference between Maximum difference
 usual and actual basic between usual and actual
 basic hours basic hours

 Coefficient t Coefficient t

Annualised hours 0.989 3.78 1.820 3.09
Flexitime -0.267 -2.03 -0.639 -2.16
4 1/2 day week -0.380 -1.95 -0.989 -2.25
Sample size 4848 4848
[R.sup.2] 0.0120 0.0142
Other controls Industry, company size Industry, company size

Note: (a) Excluding proxy respondents and managers/professionals.
Coefficients are expressed relative to the excluded standard
contract (with and without overtime).

Table 4. Determinants of working time contracts: multinomial logit
regressions, LFS 1997Q1-2000Q1 (a)

Manufacturing Annualised Flexitime
industries hours

 RRR t-stat RRR t-stat

Output variation 2.340 2.48 0.788 -0.76
Private 0.889 -1.07 0.283 -24.03
Company size 1.570 6.39 1.535 8.52
Union
Sex 1.395 4.34 0.657 -8.91
Year 1998 0.835 -2.34 0.984 -0.30
Year 1999 0.855 -2.02 0.987 -0.24
Year 2000 1.184 1.91 1.014 0.21
Other higher
 education
 qualification 1.713 4.13 0.643 -7.39
A-level, higher 2.047 5.13 0.401 -11.09
O-level, SCE 1.961 5.76 0.353 -18.31
YT certificate 2.088 5.55 0.337 -13.82
Other professional/
 vocational
 qualifications 2.112 5.81 0.313 -14.66
Energy & Water 1.332 2.07 3.205 13.58
Minerals, ores,
 metals 1.761 6.57 2.189 10.74
Metal goods,
 engineering &
 vehicles 0.846 -2.22 1.817 9.79
Construction 0.720 -3.11 2.190 10.74

Other details Sample size = 36199
 Log likelihood = -21799.1
 Pseudo [R.sup.2] = 0.073

Manufacturing 4 1/2 day week Annualised
industries hours

 RRR t-stat RRR t-stat

Output variation 0.314 -3.24 2.088 1.38
Private 0.800 -2.44 0.890 -0.70
Company size 2.157 13.17 1.404 2.92
Union 1.867 6.28
Sex 0.888 -2.26 1.095 0.79
Year 1998 0.912 -1.62 0.890 -1.04
Year 1999 0.906 -1.70 1.132 1.15
Year 2000 0.844 -2.26
Other higher
 education
 qualification 1.604 5.16 1.332 1.42
A-level, higher 1.499 3.83 1.617 2.22
O-level, SCE 1.576 5.45 1.582 2.57
YT certificate 1.911 6.81 1.622 2.37
Other professional/
 vocational
 qualifications 1.634 5.18 2.003 3.58
Energy & Water 0.133 -6.57 1.166 0.73
Minerals, ores,
 metals 0.539 -5.89 1.822 4.50
Metal goods,
 engineering &
 vehicles 2.211 14.96 0.899 -0.90
Construction 0.240 -11.41 0.725 -1.90

Other details Sample Sample size - 15714
 size = 36199 Log likelihood = -9397.1
 Log Pseudo [R.sup.2] 0.078
 likelihood
 = -21799.1
 Pseudo
 [R.sup.2]
 = 0.073

Manufacturing Flexitime 4 1/2 day
industries

 RRR t-stat RRR t-stat

Output variation 0.429 -1.68 0.292 -2.29
Private 0.278 -15.87 0.910 -0.67
Company size 1.415 4.51 1.820 6.63
Union 1.180 2.29 1.799 8.09
Sex 0.645 -6.06 0.791 -2.96
Year 1998 1.097 1.24 1.017 0.21
Year 1999 1.074 0.94 1.049 0.59
Year 2000
Other higher
 education
 qualification 0.618 -5.22 1.425 2.55
A-level, higher 0.422 -6.87 1.330 1.79
O-level, SCE 0.361 -11.63 1.391 2.61
YT certificate 0.323 -9.33 1.621 3.35
Other professional/
 vocational
 qualifications 0.318 -9.43 1.455 2.63
Energy & Water 2.777 7.72 0.107 -4.92
Minerals, ores,
 metals 2.304 7.47 0.518 -4.22
Metal goods,
 engineering &
 vehicles 1.885 6.74 2.138 9.60
Construction 2.231 7.20 0.248 -7.57

Other details Sample size = 15714
 Log likelihood = -9397.
 I Pseudo [R.sup.2] = 0.078

Notes: (a) Excluding proxy respondents. Coefficients are expressed
relative to the excluded standard contract (with and without
overtime). (b) The left side of the table excludes the union
variable and is based on quarters 1 and 3 from the Labour Force
Survey. The right side includes union and refers only to quarter 3.
The dependent variable is type of contract and the base category is
the standard contract. The excluded variables are: public companies;
company size<S0; non union; female; 1997; degree; Other manufacturing.

Table 5. Determinants of shiftworking (males
excluding managers/professionals): probit regression
LFS 1997Q1-2000Q1(a)

 Including proxy Excluding proxy
 respondents respondents
 Coefficient t-stat. Coefficient t-star.

Annualised 0.246 4.41 0.300 4.18
Flexitime -0.979 -15.91 -1.022 -13.65
Term-time -0.651 -1.56 -0.987 -1.71
Job share 1.550 1.71
9 day fortnight -0.107 -0.76 -0.054 -0.30
4 1/2 day week -0.657 -9.54 -0.737 -8.24
Zero hours 0.117 0.79 0.055 0.27
Agriculture -1.026 -6.65 -1.146 -4.76
Extraction 0.273 3.28 0.305 2.81
Engin & Veh 0.000 0.00 0.004 0.04
Other Manuf 0.254 3.36 0.262 2.64
Construction -1.217 -13.54 -1.245 -10.20
Distribution -0.142 -1.88 -0.150 -1.50
Trans & Comms 0.242 3.18 0.304 3.05
Finance -0.183 -2.23 -0.245 -2.28
Other 0.136 1.74 0.163 1.61
Year 1998 0.037 1.26 0.039 1.00
Year 1999 0.020 0.70 0.048 1.25
Year 2000 0.055 1.88 0.057 1.47
Company size>50 0.703 30.60 0.702 22.83
Private sector 0.080 2.37 0.050 1.16
Constant -1.053 -13.97 -0.976 -9.86

Sample size 19781 10696
Log likelihood -9845.4 -5619.3

Note: (a) The excluded variables are: standard contract (with
and without overtime); energy & water, 1997; company size<50;
public companies.

Table 6. Number of different days per week
worked (males excluding managers/professionals)--multinomial
logit regressions, LFS 1997Q1-2000Q1
(excluding proxy respondents) (a)

 Work 1-4 Work 6 or 7
 different days different days

 Coefficient t-stat. Coefficient t-stat.

Annualised 1.623 3.15 0.979 -0.18
Flexitime 0.593 -2.30 0.558 -5.34
Term-time 0.837 -0.18 0.417 -1.15
Zero hours 0.492 -0.96 0.886 -0.40
Agriculture 0.540 -0.81 2.539 4.38
Extraction 1.607 1.64 1.162 0.85
Engin & Veh 1.308 1.00 0.851 -1.02
Other Manuf 1.573 1.69 0.836 -1.10
Construction 0.565 -1.65 0.931 -0.43
Distribution 0.816 -0.71 1.126 0.75
Trans & Comms 0.861 -0.54 1.697 3.33
Finance 0.996 -0.01 0.854 -0.93
Other 1.184 0.61 1.086 0.51
Year 1998 1.105 0.92 1.080 1.29
Year 1999 1.084 0.73 0.981 -0.31
Year 2000 1.301 2.48 0.962 -0.63
Company size>50 1.219 2.05 0.764 -5.61
Private sector 1.137 1.06 0.850 -2.34
Shift dummy 5.813 20.18 1.818 12.07

Other details Sample size = 12415
 Log likelihood = -9138.5317
 Pseudo [R.sup.2] = 0.0575

Note: (a)The excluded variables are: standard contract (with and
without overtime); energy & water; 1997; company size <50; private
companies. The shift dummy had to be imputed. This limited the number
of observations since the pattern of response necessary to form this
variable required information for all five waves in which an
individual might be surveyed.

Table 7. Bank holiday and weekend working (males--excluding managers/
professionals)--probit regressions: 1997Q1-2000Q1 (excluding proxy
respondents)

 Worked bank holidays Ever work on a Saturday

 Coefficient t-stat Coefficient t-stat

Annualised 0.516 2.33 0.076 0.99
Flexitime -0.510 -2.50 -0.217 -4.19
Term-time 0.250 0.28 -1.108 -3.30
Job sharing -- -- -0.750 -0.76
9 day fortnight -0.113 -0.18 -0.015 -0.09
4 1/2 day week 0.010 0.04 -0.007 -0.01
Zero hours -- -- -0.154 -0.81
Agriculture 0.680 1.51 0.345 2.13
Extraction -0.296 -0.77 -0.132 -1.18
Engin & Veh -0.974 -2.87 -0.259 -2.63
Other Manuf -0.469 -1.36 -0.421 -4.23
Construction -0.401 -1.08 -0.100 -0.97
Distribution -0.134 -0.39 -0.213 -2.13
Trans & Comms -0.019 -0.06 -0.158 -1.55
Finance -0.320 -0.89 -0.530 -5.13
Other -0.089 -0.25 -0.180 -1.77
Year 1998 -- (6) -- -0.036 -1.04
Year 1999 -- -- -0.473 -12.60
Year 2000 -- -- -0.407 -10.79
Company size>50 0.125 1.12 -0.082 2.81
Private sector 0.114 0.73 -0.040 -0.93
Shift dummy 0.951 7.84 0.515 15.06
Constant -0.338 -1.03 0.960 8.82

Sample size 784 10584
Log likelihood -9214.4 -7162.7

 Ever work on a Sunday

 Coefficient t-stat

Annualised 0.164 2.34
Flexitime -0.228 -4.44
Term-time -0.894 -2.49
Job sharing -0.403 -0.40
9 day fortnight 0.031 0.20
4 1/2 day week 0.055 0.82
Zero hours -0.163 -0.90
Agriculture 0.320 2.33
Extraction -0.470 -4.54
Engin & Veh -0.592 -6.47
Other Manuf -0.716 -7.67
Construction -0.210 -2.22
Distribution -0.617 -6.64
Trans & Comms -0.453 -4.82
Finance -0.605 -6.21
Other -0.132 -1.39
Year 1998 -0.021 -0.65
Year 1999 -0.226 -6.49
Year 2000 -0.240 -6.85
Company size>50 0.073 2.68
Private sector -0.007 -0.19
Shift dummy 0.872 28.17
Constant 0.222 2.44

Sample size 11416
Log likelihood -7790.1

Notes: The excluded variables are: standard contract (with and
without overtime); energy & water; 1997; company size<50; public
sector. Bank holiday details are only available for one quarter
(2000Q1); the first regression, therefore, has no time dummies. None
of the regressions above required hours information, resulting in a
substantial increase in the number of observations. However, bank
holiday and shift questions were asked in different quarters and,
given that the shift dummy was again imputed as in table 5, the number
of observations in the first regression above was limited.

REFERENCES

Bell, D.N.F. and Hart, R.A. (1999), 'Unpaid work', Economica, 66, pp. 271-90.

--(2002), 'Annualised hours' contracts', ESRC main report, under Grant R 000 22 3442.

--(2003), 'How important is guaranteed or institutionalised overtime?', (mimeo) Department of Economics, University of Stirling.

Gall, G. (1996), 'All year round: the growth of annual hours in Britain', Personnel Review, 25, pp. 35-52.

Incomes Data Services (1996), Annual Hours, Study 604, London.

--(1999) Annual Hours, Study 674, London.

--(2002), Annual Hours, Study 721, London.

Vickery and Wurzburg (1996), 'Flexible firms, skills and employment', OECD Observer 202, Oct-Nov, pp. 17-21.

Robert A. Hart, Department of Economics, University of Stirling. e-mails: d.n.fbell@stir.ac.uk or r.a.hart@stir.ac.uk. This project was funded by the ESRC (ESRC grant R 000 22 3442). Elizabeth Roberts provided excellent research assistance. We thank a referee for a number of very helpful comments. We are also grateful to Julian Dowsell (Office for National Statistics) for supplying us with detailed manufacturing output data. This paper is based on a fuller report entitled 'Annualised hours contracts' (Bell and Hart, 2002). The authors would be happy to supply a copy on request.