Economic cycles and structural changes in the economy, brought about by the expansion of the service economy, technological innovation, globalisation and the major reorganisation of services traditionally provided by the public sector (eg privatisation or outsourcing), have led to rapid changes in workplace staffing levels. It has been suggested recently that these changes—particularly downsizing—may have adverse health effects among those who remain in work at such workplaces.1 2

It has also been suggested that the possible effects of downsizing on health result from elevated levels of work stress in the downsized workplaces. Previous evidence has shown that downsizing is associated with adverse changes in job demands and job control, and that its association with sickness absence is strongly attenuated when psychosocial and physical job exposure is controlled for.3 Downsizing may thus constitute a common threat to public health in the working population.

Most of the evidence concerning the adverse effects of downsizing on health among those remaining in the downsized workplaces comes from studies on sickness absence1^–5 and, until now, research on its effects on disease incidence and mortality has been rare.2 Moreover, few of the previous studies were based on nationally representative populations. Researchers have thus been at a disadvantage in terms of not being able to study the effects of downsizing on mortality by sex or by a broad range of causes of death, and have studied changes in employment by large employers or industries rather than being able to define exposure to downsizing at the workplace level. Furthermore, few studies put the results of the effects of downsizing in a broader perspective and inform about selection processes by comparing the health of those who remain in downsized workplaces with those who left to work elsewhere or became unemployed.

In order to overcome these limitations, we used unique prospective population registration-based Finnish data that link 85 833 individuals with data on 1-year changes in the staffing levels of workplaces, and on mortality over 8 years. Our specific research objectives were to estimate:

• whether downsizing is associated with increased mortality among those who remain in work in the same workplaces and, for comparative purposes, among those employed in other workplaces and those with experience of unemployment

• whether the possible increase in mortality after downsizing for those who remain in work in the same workplaces is similar for men and women, for different causes of death and for different durations of follow-up.

PARTICIPANTS AND STUDY DESIGN

The target population consisted of men and women born between 1924 and 1959, living in Finland between the years 1987 and 2000 and working in workplaces with five or more employees.6 Because of the data protection regulations of Statistics Finland covering both individuals and workplaces, we did not have access to the total population, as this would have been a violation of the regulations of data provision. To obtain a dataset that had sufficient power to analyse mortality by cause, the research team thus designed a three-part sample of the target population, with oversampling of deaths and those in strongly downsized workplaces: (1) a 50% random sample of people who had worked at a workplace between 1988 and 1994 and had died before the end of 2002; (2) a 20% random sample of workers whose workplace had reduced its workforce by at least one half between two adjacent years during the period 1988–1994; and (3) a 10% random sample of the whole population in the period 1988–1994. To obtain results that are representative for the population, we used sampling weights, constructed from the known sampling probabilities, in the analyses.

For the purposes of this study, we extracted those aged 35–64 years on 1 January 1994 who were working in a workplace. For this cohort, we measured workplace downsizing for the year 1994 (a time of recession and strong downsizing) and followed mortality from the beginning of 1995 to the end of 2002. In addition, because we wanted to include as many study subjects as possible in the analyses, we obtained the corresponding information from the beginning of 1993 for those who could not—because of unemployment, for example—be allocated to a workplace at the beginning of 1994. For this subpopulation, we thus acquired the equivalent information from the beginning of 1993, assessed downsizing for the year 1993 and followed these subjects for mortality for a corresponding 8-year period from the beginning of 1994 until the end of 2001. The two populations were combined in the analyses.

Measurements

Statistics Finland provided information on every sampled person from three different data sources, linked the individual records using personal identification codes and provided us with an anonymised set of data.

All the sampled people were allocated to a workplace, the total number of workplaces employing our study participants being 21 418. Information on all workplaces in Finland was taken from the Statistics Finland workplace register, which provides detailed information on turnover, production, industry and staffing levels on an annual basis, and also gives general information on the industry concerned and on the geographical location of the workplaces. For workplaces that had five or more employees, we calculated the percentage changes in the size of the workforce in the 1-year period from the baseline (e_b+1–e_b)/e_b, where e_b and e_b+1 were numbers of employees at the baseline and 1 year later. This was classified as: (1) employment secure (less than –9%); (2) moderate downsizing (−10% to −29%); (3) severe downsizing (−30% to −49%) and workplace closure (−50% to −100%; of the people in the 50–100% range of downsizing, about 70% were in workplaces that were fully closed). In the general population, about 70% had worked in secure jobs, and more than half of those who had experienced downsizing had worked in workplaces that had downsized by less than 10% (table 1). Previous analyses show that, in these data, workplace downsizing is associated with unemployment.6

We also had access to data from the Statistics Finland labour market data file, which combines basic demographic and socioeconomic information on everyone living in Finland with detailed information on labour market participation from different sources on an annual basis. It covers all employment and unemployment episodes on an individual basis. Statistics Finland acquires information on unemployment from the records compiled by the Ministry of Labour on registered unemployment spells. The data source can be considered reliable, as it forms the basis of the payment of unemployment benefits. From these data and the information in the workplace register, we determined the employment status of all our study participants. We classified them as: (1) employed in the same workplace—working at the beginning and end of the baseline year in the same workplace; (2) employed in another workplace—having changed workplace during the baseline year; and (3) having experienced unemployment—working at the beginning of the baseline year but having experienced unemployment (one or more months of unemployment) before the end of it.

The same data source provided information on the highest completed educational degree. The three categories were tertiary education, intermediate education, and basic education or less or education unknown. Four social classes were used: upper white collar, lower white collar, manual and unknown (0.15% of all participants). Marital status was classified as married, never married, divorced or widowed.

The mortality follow-up is based on the Statistics Finland register of deaths from which we extracted the date and cause of death, the cause being classified according to the 10th revision of the International Classification of Diseases and Deaths (ICD).

Methods and modelling strategy

The data were analysed by means of Cox regression analysis, and the STATA statistical package was used in fitting the models.7 The results of the Cox models are presented as hazard ratios, with one category of each explanatory variable taken as a reference group. Age (5-year categories), education, socioeconomic status and marital status were adjusted as categorical variables. In further confirmatory analyses, we also adjusted for industry (16 main groups, NACE 1.1 classification), geographical location (20 regions, NUTS3) and size of workplace (five categories) but, as these adjustments did not substantially change our analyses, we do not show these results here. Our main results were also obtained in a subset of participants employed in workplaces that had 50 or more employees (58% of all participants) and in the 10% random sample of the population.

In evaluating the study aims, we adopted the following modelling strategy. In order to establish the magnitude of the association between downsizing and mortality separately for those who remain in work in the same workplaces, those employed in other workplaces and those with experience of unemployment, we fitted an interaction term between downsizing and employment status adjusting for age and sex (table 1). To follow our main aim, all subsequent analyses only included those who remained in work in the same downsized workplaces, and evaluated whether the effects of downsizing varied by duration of follow-up and cause of death. Sex differences were evaluated in stratified sex-specific, age-adjusted models. Finally, duration of follow-up specific models and cause-specific models again combined the sexes and adjusted for age and sex.

RESULTS

Table 1 shows relative mortality rates by level of downsizing and employment status with reference mortality set at 1 for those in secure employment and working in the same workplace. When comparing the numbers in table 1 down the columns, we observe that workplace downsizing was not associated with increased all-cause mortality among those remaining in the same workplaces, those moving to another workplace or those who became unemployed. The data provide evidence of reduced mortality among those who remained employed in the most strongly downsized workplaces, with hazard ratios of 0.89 (95% CI 0.81 to 0.97), 0.90 (95% CI 0.77 to 1.05) and 0.72 (95% CI 0.61 to 0.85) for those in moderately, severely downsized and closed workplaces respectively. In the 10% random sample part of our dataset, the corresponding hazard ratios were 0.85 (95% CI 0.69 to 1.04), 0.81 (95% CI 0.54 to 1.21) and 0.91 (95% CI 0.59 to 1.40). When reading table 1 across the rows and keeping the level of downsizing constant, ie when assessing the effects of change in workplace or experience of unemployment on mortality, we see that, at lower levels of downsizing (10–29%), about 50–70% excess mortality was observed among those who left one workplace to work elsewhere (1.52/0.89) or who became unemployed (1.36/0.89) compared with those who stayed in the same place (table 1). The corresponding excess mortality was about 10–30% at more severe levels of downsizing.

Table 2 focuses on the effects of downsizing on those who remain in the same workplace. The differences between men and women were small and statistically insignificant (p = 0.98, test not shown) (table 2). Furthermore, these associations remained unchanged when we adjusted for education, social class and marital status.

In addition, the total mortality differences from downsizing among those remaining in the same workplaces were similar at different durations of follow-up (table 3). For our cause-specific analyses, we combined the sexes (table 4). No excess mortality was associated with downsizing in any of the specific causes of death studied. A protective effect was observed in the “other cancers” category particularly in the most strongly downsized workplaces. Results not presented here also showed similar associations at all durations of follow-up for the main causes of death categories used in this study.

DISCUSSION

Our main aim was to quantify the effects of workplace downsizing on cause-specific mortality among those remaining at work. Research into the health effects of downsizing contributes to the more general social epidemiological debate on the health effects of psychosocial stressors at work.8^–10 Methodologically, the study of the health effects of downsizing is appealing because the interpretation of the results of such studies are less likely to be affected by confounding or selection than other observational study designs. In particular, (1) because all those exposed to downsizing remain at work, any confounding due to changes in material/financial conditions is more limited; and (2) downsizing—particularly severe downsizing—closely resembles a natural experiment whereupon exposure is less likely to be determined by prior individual characteristics.

Our data give no evidence of any adverse effects of workplace downsizing on mortality over an 8-year follow-up period: there was no increase in all-cause or in cause-specific mortality among the men and women who remained in work in downsized workplaces. Furthermore, there was still no evidence of increased mortality even in the early years of our follow-up period or for employees in workplaces with 50 or more employees (results not shown). Overall, we observed no adverse effects of downsizing on mortality among those who remained at work.

We were able to study the effects of much more severe downsizing than has been considered in previous studies, in which the most severe has often been classified as 18% or more in 2 years1 or 18% or more in 1 year.5 We were able to refine the “above 18%” grouping to include annual changes of 18–29%, 30–50% and 50% and more. At the more extreme level of downsizing, exposure to changes in work practices and job demands might be expected to be most severe among those remaining in work. Nevertheless, we observed no excess total mortality among those remaining in the workplaces that were most strongly downsized. Overall, this result is consistent with the notions that downsizing may not be strongly and consistently associated with changes in working conditions in all sectors of the economy11 and/or changes in these mediating factors are not associated with mortality.

Contrary to expectations, our data show a possible beneficial effect of severe (more than 50%) downsizing on mortality among those remaining in downsized workplaces. Also, Westerlund and colleagues,5 in a nationally representative study of Swedes, provide some evidence that moderate, rather than severe, downsizing is associated with sickness absence. There are at least three possible explanations for the beneficial effect of severe downsizing that we observe. First, our study was carried out in the context of a severe economic recession of nearly 20% unemployment in Finland. In addition to the health-damaging effects of severe recessions, our observation may be in line with arguments concerning the possible beneficial effects of economic recessions, possibly associated with a reduction in unhealthy behaviour such as alcohol abuse.12^–16 These protective effects may be particularly evident for physical health.12 16 Second, the protective effects among those remaining in downsized workplaces may only be observed in the Finnish or Nordic context of relatively stringent legislation on working conditions, overtime work and work protection. These protective effects may be absent in societies with less comprehensive protection of workers. Third, those left in the most strongly downsized workplaces may be the owners or executive staff, or may be selected on account of their good health. The protective effect could thus simply result from a strong healthy worker effect. Overall, the lower mortality rate of those in strongly downsized workplaces needs to be confirmed in future studies on other populations and at other points in time, in particular in periods of more normal economic circumstances or in periods of rapid economic growth. Only after that can serious attempts to explain this association, for example in terms of reduction in unhealthy behaviour, be developed. This lies outside the scope of the present study.

In addition to allowing us to establish the effects of downsizing on mortality (reading table 1 down the columns), we believe that our data provide good evidence of the association between rapid labour market change and workplace downsizing and mortality among those who leave such workplaces, thus potentially providing us with insight into how selective movement out of downsized workplaces may influence mortality. At moderate levels of downsizing (10–29%), there was a 50–70% excess mortality rate among those who left or had been forced to leave a workplace—either to work elsewhere or to become unemployed—compared with those who stayed (table 1). Because this excess mortality was relatively similar in both groups of leavers, we believe it may be consistent with the idea that those who are forced to leave are selected on the grounds of prior health problems or factors that are strongly correlated with them and that employers might find it easier to get rid of workers with health problems at a time of moderate downsizing. This conclusion is further supported by the observation that the mortality levels of those who left a more severely downsized workplace appear to be more similar to the mortality levels of those who stayed, indicating that, at more extreme levels of downsizing or closure, health selection has less importance in the explanation of mortality differentials between the three groups. Overall, these results suggest that health selection may play an important role in the explanation of mortality differentials between employment status groups. On the other hand, mortality differences between those who remained in non-downsizing workplaces compared with their colleagues who became unemployed or were hired elsewhere are very small. This finding may be related to the fact that our baseline year, 1994, was close to the height of the recession of the early 1990s. Indeed, the mortality difference between the unemployed and employed who started in the group “employment secure” was very large in 1989, a year of low unemployment.6

The data used for this study are of high quality, with practically no missing data and no recall bias. They cover a large sample of the Finnish population aged 35–64 years working in workplaces with more than five employees (more than 90% of all employees) at the beginning of the follow-up. Uniquely, the sample combines individual-level data with data on workplace downsizing for 21 418 workplaces and is large enough to allow the study of mortality in a way that exploits natural experimental strategies. The fact that we strictly excluded those with experience of unemployment (more than 1 month) guarantees that the strong association between unemployment and mortality in individual-level observational studies does not confound our findings. Furthermore, because we have studied 1-year changes in workplace staffing levels, we have minimised the risk of contamination under the assumption that only the healthiest survive at work (with longer follow-up for downsizing, less healthy employees may be more likely to fall out of the labour force). However, this analytical decision might also lead to a failure to detect true associations, because it is possible that the total exposure is not measured accurately, both because single measures of exposure may lead to misclassification and because accumulated exposure is needed for adverse health consequences to emerge. No previous study has attempted to do this. More evidence is also needed that takes into account the reasons for downsizing and incorporates exact information on which workers are downsized.

Comparisons with other studies

Most of the previous analyses of the effects of workplace downsizing on the health of those remaining in work are based on regional samples with limited numbers of workplaces and/or the workplaces are largely defined on the basis of occupational groups rather than workplaces as such. All these factors may lead to bias in the estimation of the degree of downsizing in terms of representativeness, random variability and validity. We believe we have overcome these difficulties with our random sample of the Finnish population and our workplace-level assessment. We believe that analysis at the workplace level—based on location and facility (building)—is most suitable for testing hypotheses concerning the health effects of workplace downsizing.

Much of the previous research on downsizing among those remaining in downsized workplaces has concerned sickness absence. The results are somewhat conflicting in that both increased and decreased absence levels have been found, and the size of the effects varies.1 4 5 This inconsistency in results may be due to a variety of factors. For example, the type of employment contract may make a difference, with those on temporary contracts less likely to take sick leave after downsizing and those on permanent contracts more likely to do so.2 More generally, short-term changes in sickness absence do not necessarily reflect the incidence of new illnesses or the aggravation of existing ones, and may be connected with perceptions of the need to be present at work (ie sickness presenteeism)17 or with illness behaviour. However, given the results of studies on sickness absence, it remains possible that substantial downsizing may be associated with an increased incidence of mental health problems in particular.

The results of a long line of longitudinal research on the health effects of unemployment, and more recent work on job insecurity, have similarly shown a robust effect on mental health.18^–20 However, evidence of the effects of unemployment on physical health, and particularly on mortality, is less consistent, and the nature and size of the effects of unemployment and job insecurity remain difficult to determine.21 22 In showing that severe downsizing is not associated with mortality either among those remaining in work or among those becoming unemployed, the results of this study corroborate these earlier findings. Our findings are also consistent with those of a previous study conducted in Finland during the recession of the early 1990s showing that changes in mortality were similar in occupational groups in which the unemployment rates increased at different rates.23

Conclusion

Research on the effects of workplace downsizing on health among those remaining in downsized workplaces is an emerging area of study, and it is understandable that the results from different studies can be quite different. Thus far, the available evidence is based on a limited number of investigations carried out in specific national and historical contexts, and using a variety of study designs, methodologies and outcomes. We believe future research should focus particularly on comparing different ways of measuring exposure to workplace downsizing, and on identifying whether workplace reduction in staff at the same occupational rank has larger health effects than overall staff reductions. Because our analyses are limited to mortality—a health outcome of particular severity—further work should also try to identify whether the possible harmful effects of downsizing are limited to other less severe outcomes.

Overall, our results indicate that the effects of workplace downsizing on total mortality among those who remain in these workplaces are non-existent. Furthermore, among those remaining in downsized workplaces, we have observed no excess mortality for specific causes of death—suicide, alcohol-associated causes or IHD—which are most often hypothesised to react to the changes brought about by downsizing or its immediate aftermath. The implication may be that the possible changes in psychosocial and physical working conditions following the reduction in workplace personnel are not significant determinants of mortality among those remaining in downsized workplaces.