More on the Penn World Tables Data

Should G/Y and I/Y be Measured in Local or International Prices?

In the Penn World Tables data set, C, I, G, and Y/P (income per capita) are measured in "international prices". That is, rather than simply measuring these variables in US dollars, the data are adjusted to take account of the fact that the purchasing power of a US dollar differs across countries (typically a US dollar buys more in a low-income country than in a high-income country). This provides a very useful way of comparing incomes per capita across countries. However, I argue that the methodology employed can give misleading measures of G/Y and I/Y. 

The bias is caused because the construction of the data set involves setting relative prices in all countries equal to an international set of relative prices. This has an interesting implication. The price of labour, relative to capital, in poor countries is lower than in rich countries and the government sector tends to be labour intensive in poor countries. G/Y therefore tends to be quite low. However, conversion to international prices, articifially increases the value of G/Y (effectively what the conversion does is ask what G/Y would be in poor countries if workers were paid rich country wages). For high-income countries, conversion to international prices tends to reduce G/Y. For I/Y, conversion to international prices tends to reduce I/Y for low-income countries and increase it for high-income countries. 

Further Thoughts on the Measurement of G/Y and I/Y

Since this paper has been published, I have come round to the view that it may be appropriate to measure I/Y using international prices, but remain convinced that it is inappropriate to measure G/Y in international prices. This change in view with regard to I/Y is due largely to discussions with Professor Norman Gemmell from the University of Nottingham.

The issue of whether to use local or international prices to measure G/Y and I/Y when making comparisons across countries can be related to the debate over whether to measure G/Y in current (nominal) or constant (real) prices when making time-series comparisions. Initially, it may appear that it makes no difference, but if efficiency gains are lower in the government sector (eg in the provision of health care) than the non-government sector, it does. The cost of producing health care as a share of Y is increasing faster than in other sectors of the economy, even though the value of health care may not be. To capture the increase in cost requires using current prices; to capture the fact that the value of the output is not increasing requires the use of constant prices. 

Relating this back to comparisons of G/Y across countries, if we want to focus on the cost of paying for goverment consumption, we should use local prices (analagous to current prices), but if we wanted to focus on the value of the output (which seems less likely if we want to look at the effect on growth) we should use international prices (analagous to constant prices). 

To continue to use international prices would suggest that the government of Nepal needs to raise 47.5% of GDP in taxes to pay for government consumption (if version 5 of the PWT  is used), when in fact government consumption only accounts for 8.3% of GDP when measured in local prices.

With I/Y it is valid to use international prices if we are interested in the value of I/Y, rather than the cost.

^ Top of page | Contact Us | Disclaimer | Email economics@otago.ac.nz