In this paper, we present country-specific productivity cost estimates for several countries. Our human capital-based PVLE estimates for the US were slightly lower than those calculated by Grosse et al
]. Their estimated PVLE (excluding household work) for males ranged from just under $16,000 to over $1,000,000 (2007 USD) depending on age; our estimates ranged from approximately $22,000 to $829,000 across similar age groups (age groupings not reported in this paper). This is likely due to our differing methodologies in estimating annual wages. Grosse et al.
aggregated hourly wage estimates to an annual level based on estimated hours worked per week, whereas, to ensure a certain level of consistency across countries methodologically, we used average annual wages by age and sex. Estimates from our model for China were on par with the literature for the human capital-based method as well. For males between 35 and 64 for 5-year age groups, our model estimated PVLE to be between approximately 57,000 and 209,000 (2009 Chinese Yuan), while Sung et al.
estimated a range of 27,350 and 264,000 (2003 Chinese Yuan) [30
The most notable critique of the human capital approach is that it measures potential lost production instead of actual lost production, which could lead to significant overestimation of productivity costs due to premature mortality [9
]. It has been argued that the friction cost methodology is more precise than the human capital methodology in that it captures actual productivity losses versus the total potential loss in the human capital approach [9
]. However, the level of detail and extensive data required for this method (e.g., industry- or education-level data on average length of vacancy for an open position, disease-related average frequency and length of absence from work) made its application to our objective of estimating productivity costs across several countries unrealistic. We did include a sensitivity analysis using a rough approximation of the friction cost approach for the US and Brazil and found, as might be expected, substantially lower costs for lost productivity among those of working age. Further research into available sources of data to fully implement the friction cost approach would be valuable, such as length of the friction period for various occupations, availability of surplus labor, and whether positions are filled by unemployed workers or employed workers who are changing jobs (and thus generating a new friction period).
By its nature, our model is subject to certain limitations, namely surrounding data availability. It was not possible to stratify wages by age and sex in all countries, or to assess variation by occupation. Employment statistics did not distinguish between full-time and part-time work in most countries; our use of a full-time work assumption may overstate lost productivity costs. However, proportionate downward adjustments are possible to allow for data or assumptions about part-time work. Labor force participation rates and wages both vary based on general economic conditions, which would influence our estimates (e.g., higher estimates in good economic times, lower estimates in bad times). Due to this data variability, along with differences in epidemiology among diseases of interest and differences across health care systems, we recommend, in general, that researchers use the PVLE estimates to conduct single country analyses. Despite these known limitations, there has been recent interest in analysis of the transferability of economic evaluations across countries [31
]. Thus, it may be possible to cautiously relate results between countries if special attention is given to assessing the comparability of model inputs and any key differences in country characteristics.
Our model evaluated lost productivity based on averages for general populations, including persons in the paid work force and those who are not, rather than only for those with specific diseases or comorbidities. Although these may limit the direct relevance of our model inputs and calculation to specific diseases, it strengthens the generalizability of our findings regarding the societal productivity costs. We also focused only on the value of lost productivity due to premature mortality. We recognize that indirect costs also should include morbidity costs, which were beyond the scope of our analysis. Finally, it could be argued that the actual cost to society from premature mortality would be production loss minus consumption loss, which would overstate productivity losses under the human capital approach.
We expanded on the traditional human capital model by making it possible to incorporate estimates of the value of household work, which has known limitations, including, for example, the transfer of gender wage differentials in the paid labor market to the production of household services [33
]. Furthermore, we acknowledge that our use of an opportunity cost approach to estimate the value of household work rather than a replacement cost (i.e., the cost to hire someone to complete the household work) results in higher estimates of PVLE when household work is included, as exhibited by the sensitivity analysis that we conducted. As stated previously, due to data limitations we chose to use the opportunity cost approach in this model. As with any valuation based on age- and gender- specific wages, these PVLE estimates are biased against older and female populations. Despite the potential ethical considerations surrounding valuing one life more than another, it is still recommended to use age- and gender-specific wages where possible, as they are more targeted and accurate [5
Also, in order to collect some input values from source documents written in other languages, we used Google Translate, which produced English versions that were imperfect and sometimes required additional interpretation.
Our analysis highlights the need for uniform data across countries. International organizations, such as the WHO and ILO, provide useful country-specific estimates for important model inputs such as life expectancy and labor force participation. However, where there are gaps in standardized data, country-specific data sources had to be used, which introduced some methodological variability into our results. Finally, the use of broader age categories leads to more imprecision in estimates of PVLE, so we recommend that researchers use the 5-year age brackets when feasible.