Machines that go ‘ping’: Medical Technology and Health Expenditures in OECD Countries
修正了此前编程错误，重新估计18个OECD国家1981-2012年卫生支出面板模型，确认技术是驱动卫生支出增长的关键因素，收入、生活方式（BMI）和技术的影响显著，激进型与增量型医疗创新的差异效应也得到验证。
The published version of our paper contained an error in the programming code. This note describes the mistake and reports results based on corrected data. The revised results are qualitatively similar to those published and confirm the important role of technology as a driver of aggregate health spending. Willemé and Dumont estimate a panel model of aggregate health expenditures for 18 Organisation for Economic Co-operation and Development countries over the period 1981–2012 in a paper published in the August 2015 issue of Health Economics (first published online in 2014). When the model was recently updated, a programming error was discovered that resulted in the values of the dependent variable not being properly deflated (the series were converted to current US$ purchasing power parity units instead of constant purchasing power parities). Consequently, the historical growth of health spending (but not gross domestic product) was measured in current instead of constant prices and was therefore too high. 1 At the same time, it became clear that the estimation results were also seriously affected by the inclusion of the share of out-of-pocket in total spending (%OOP), a variable for which only relatively recent data are available for most countries. The inclusion of the variable substantially shortens the length of the time dimension of the panel in the extended model and thereby affects the estimated coefficients and the computed contributions of the model variables to the explained historical growth of real per capita health expenditures. Because the contribution of the %OOP variable to historical growth is negligible, we have excluded the variable in our revised results, which are reported in the updated Tables IV and V. An additional advantage of excluding the out-of-pocket variable is that enough observations are available to compute the Im–Pesaran–Shin test of residual stationarity (not reported in the original paper). The updated results are qualitatively similar to those originally published, but there are some obvious and inevitable quantitative differences. Specifically, the estimated coefficients of all variables except those related to the share of the elderly in the total population remain highly significant and confirm the effects of income, lifestyle (BMI) and technology on spending. The differential effects of ‘radical’ (NME and PMA) and ‘incremental’ (NDA and PMN) medical innovation are also confirmed. In terms of the contributions to (explained) historical growth, the contribution of income is revised upward, while the effect of technology is somewhat lower but remains very substantial and similar to the range reported in Smith et al. (2009).