A Critical Review of Craig Olson’s Article ‘Do Workers Accept Lower Wages in Exchange for Health Benefits?’ (2002)
This essay will study the work of Craig Olson in Do Workers Accept Lower Wages in Exchange for Health Benefits? (2002). The article investigates the trade-off between wages and health insurance in the US labour market. The author analyses data on married women’s wages and health insurance coverage, dependent on their husband’s health insurance coverage.
Olson estimates that women accept a wage reduction of 20% (or $4000) to move from a job without health insurance coverage to a job that provides health benefits. This estimate is important as it allows firms to accurately place value on health insurance when assessing worker’s wages and fringe benefits. The following essay will question the validity of the experiment, reasons for the study, transferability of the results and the instrumental variables. Previous literature and widely accepted labour economic theories will be compared to this article in order to assess the argument presented.
Instrumental variables are used by Olson in order to estimate the effect that health insurance has on wages. The author recognises that regressing health insurance directly onto wages will produce bias, arising from the heterogeneity of workers. This is illustrated in Figure 1 of the article where it is illustrated that preferences and ability level will determine workers’ wages and whether they accept health insurance. To eliminate unobserved heterogeneity, Olson firstly considers using the wives’ coverage on their husband’s health insurance as a dummy variable, as this correlates with the wives’ choice to take health insurance through her job. However, this instrumental variable would bias the results, due to the ‘positive assertive mating’ in the population. In this particular context, this refers to the positive correlation between husbands’ and wives’ salaries, and therefore the positive correlation of availability of healthcare insurance at their respective jobs. Another reason for this variable to produce a bias is simultaneity of the regression; Husbands are likely to choose employment without healthcare benefits if they are already covered by their wives’ package, which directly contrasts with the relationship Olson is studying. Therefore, Olson chooses to represent the husband’s health insurance through two exogenous variables; the size of the husband’s firm and his union status. These are both likely to impact whether the wife has health insurance, without correlating with the error term.
The relevant data for 22,332 households was used from the Current Population Surveys and outgoing rotations group subsamples for years 1990-1993. Once the calculations are complete, Olson produces a final statistic that a 20% wage reduction is accepted by women to move from a job without health benefits to one with health benefits.
On introduction of the article, Olson states that previous literature has failed to obtain empirical evidence to support the hedonic wage theory in the context of healthcare insurance. Often studies would observe results indicating the ‘wrong-signed trade-off’ (Simon, 2001), or insignificant estimates. Brown (1980) discusses the difficulties when attempting to estimate ‘prices for non-wage characteristics.’ It is obvious that researchers encounter problems, not only in estimating the cost of health insurance to employers, but in a variety of fringe benefits. His conclusions offer five possible explanations for why it is so hard to prove the theory, including unmeasured characteristics and too strong assumptions in the labour market, but he argues that none are fully convincing. Currie and Medrain (1999) cite the lack of suitable data as the underlying issue when previous academics have sought this empirical evidence, not the actual work itself.
Clearly, the motive of this research is to fill the gap in literature at that time in proving the compensating wage differentials theory with empirical evidence. Olson calculated the desired results, which so many had previously tried and failed to do. This may seem a success at first glance, however, since there have been numerous failed attempts for years prior to Olson’s research, the dataset comes into question. Did Olson go looking for data that would support his argument and manipulate it accordingly? Or did he conduct the experiment honestly and perhaps better than others before him? If the former is true, then the results are not useful as they only apply to a specific group of selected participants. If the latter is true, Olson should discuss the pitfalls of previous literature and their methods, to improve the validity of his own method and allow it to be repeated in the future. Brown’s (1980) theories are not mentioned in Olson’s research, despite it being a widely respected piece of literature. To justify his work further, Olson should disprove or at least discuss the conclusions outlined by Brown.
Hedonic wage theory suggests ‘wage heterogeneity can be explained by the heterogeneity of worker preferences and working conditions,’ (Kaya, 2019a). Regarding health insurance coverage, theory indicates workers will accept a lower wage in exchange for insurance benefits. This means that workers place different values on health insurance against wages, resulting in them accepting different benefits. Olson takes this theory and tries to empirically measure the value that workers place on health insurance coverage. Through his research, he finds that the average woman will accept a 20% reduction in her wages, which translates to $4000 for the average woman in his data. This comparison is not particularly useful, as it is only true for a worker who earns $20,000 per annum. The proportion of each worker’s salary will be different, so it is preferred that the implicit value is used to measure the value that workers place on health insurance. This will also make it a useful statistic to firms when assigning a value to the fringe benefits they provide.
A useful comparison that Olson investigates is how his findings compare to the average cost of health insurance in the US. Previous estimates suggest that the average cost of health insurance for a family of four per annum is $4,489 (Olson and Whittaker, 1998). This improves the credibility of his Olson’s estimate as these values are very similar. However, the assumption that all women are from a family or four, and their healthcare insurance would cover a family of this size, is strong. This variable could be accounted for if further research was to take place and would be a better comparison for Olson’s estimates.
A key flaw in Olson’s data is that he has sampled a distinct group of people, making it inappropriate to apply the findings to other contexts. One restriction he imposes on the data is only involving women. This is necessary to conduct the method accurately, however, there are key differences between how men and women approach the labour market (Kaya, 2019b). Historically, women would interrupt their career to take care of their children, whilst men focused on their jobs and earned money for the household. Therefore, women may settle for low paid jobs or not attend higher education, as they anticipate they will exit the labour market in the near future. These attitudes are slowly becoming outdated, but we still observe these patterns. The pattern of male and female work can differ, with more women taking up part-time work to make time for her children. Olson has considered this variable, as he only takes data for women who work over 34 hours per week, but this does not account for the different attitudes to the labour market. These key differences between male and female labour market approaches make it difficult to apply Olson’s conclusions to the male labour force. To further the research in this paper, the experiment could be conducted to investigate men’s choice between accepting a lower wage or health insurance dependent on their wife’s health insurance coverage.
The other restriction the author places on his research is that he investigates the US healthcare market. Although this is an important country to investigate, it is unique in the way the healthcare market operates. This can be said for most countries, as healthcare is a controversial topic for governments to tackle. The experiment may be difficult to repeat in other countries such as the UK, where public healthcare is provided. Another method could be adopted to identify patterns nationally and globally.
One assumption that Olson makes throughout his paper is that workers have an explicit choice between health insurance or an increased salary. For many low earners, this is not the case. Entry-level workers often do not have the confidence or authority to demand fringe benefits or a higher salary in their first job. This is one of the limitations Brown (1980) presents, stating that ‘testing the hypothesis on a sample in the early and mid-twenties in inappropriate.’ He is not able to produce a reason for the insufficient data from this age range and questions whether it is entirely true, but nonetheless this should be considered in Olson’s study. Olson imposes and upper bound on the age of his participants (must be under 64 years old) but no lower bound, only that they are married. Arguably, this will increase the age of the women in the sample by default but does not completely eliminate young workers. There is a restriction placed on the hourly wage of workers of $2, albeit low, Olson has considered this downfall.
Another assumption placed on this study is that health insurance is a preferred benefit by all workers. This may not be the case for some in the US, as they want a more comprehensive health insurance package than is offered by their employer so they can receive high quality healthcare. Although this is a rare attitude, it is entirely possible in the US due to the diversity of health insurance available.
This essay has argued that the results of Olson in Do Workers Accept Lower Wages in Exchange for Health Benefits? (2002) are important in the field of labour economics, but the validity of the results, assumptions in the method and applicability across contexts are all questionable. The search to find empirical evidence for this theory has been lengthy, and there is an abundance of literature that Olson’s study can be compared to. Hedonic wage theory has been applied to this context and measured again Olson’s results to assess the how his findings relate to modern labour economics.
It is important to approach this article with scepticism, as Olson’s experiment is the first to obtain empirical evidence in its field. The results should not be taken as true without debating the reliability of the method and assumptions to assess how significant the study is. It is possible that the results are an anomaly, so careful analysis should be done to ensure they are justified.
Following the findings from this article, a similar experiment based on men’s decision of whether to take up health insurance or a higher wage. Clearly the experiment produces results that authenticate the theory when analysing women’s decisions, so repeating the study and focusing on men will test the reproducibility of the results. A similar study can also be repeated across different countries. It may be appropriate to adjust the method and assumptions depending on the structure of the country’s healthcare system, and any other difference in working culture and attitudes.
- Brown, C. (1980) Equalizing Differences in the Labour Market, The Quarterly Journal of Economics, Vol. 94, Issue 1, pp. 113-134
- Currie, J. and Madrain, B.C. (1999) Health, Health Insurance and the Labour Market, Handbook of Labour Economics, Vol. 3, Part C, pp. 3309-3416
- Kaya, E. (2019a) Compensating Wage Differentials, Labour Economics BS3558, Cardiff University, 7th October 2019
- Kaya, E. (2019b) Group Differences and Discrimination: UK Labour Market Gender Facts, Labour Economics BS3558, Cardiff University, 18th November 2019
- Olson, C. and Whittaker, J. (1998) The Changing Distribution of Wages and Employer Provided Health Insurance among Full-Time Married Males, Paper presented at the 1998 winter meeting of the Econometric Society, Chicago, January 1998.
- Olson, C. (2002) Do Workers Accept Lower Wages in Exchange for Health Benefits? Journal of Labour Economics, Vol. 20, No. S2, pp. S91-S114
- Simon, K. (2001) Displaced Workers and Employer-Provided Health Insurance: Evidence of a Wage/Fringe Benefit Tradeoff? International Journal of Health Care Finance and Economics, Vol. 1, Issue 3-4, pp. 249-271