RDP 9705: The Response of the Current Account to Terms of Trade Shocks: A Panel-data Study Appendix B: Data

The data sources used in this paper include: the IMF's International Financial Statistics (March and June 1996 CDROM versions); the World Bank's databases (both the 1994 and the 1995 CDROM versions); the Summers and Heston database in the Mark 5.6 version of the Penn World Tables; and the OECD National Accounts and Yearbook of Employment Statistics. For each series, I provide a brief discussion of the main issues and outline the rule used to construct the preferred measure from the various data sources. It is beyond the scope of this paper to provide details on every component of the data (see the original sources for more information). There were exceptions to these rules when data was lacking from the preferred source for individual countries; it was often necessary to splice together data from different sources to get extensive coverage across countries and across time. A detailed list of the exceptions to the rules, and the reasons for these exceptions, is available on request. Also, many obvious errors in original data were identified and corrected using graphical techniques.

Current account balance

As a rule I used the IFS (line ‘78ald’) for the current account balance (recorded in US dollars). When this data was missing, I used World Bank data (1995 CDROM, BN CAB FUND CD, current account balance after official transfers, US dollars, balance of payments basis).

Terms of trade

There are two main sources for the terms of trade – the IFS (by combining unit values ‘74…’ and ‘75…’ or price indices ‘74..d’ and ‘75..d’) and the 1995 World Bank CDROM (terms of trade index, 1987=100, TT PRI MRCH XD). Prior to 1995, the World Bank published terms of trade series that had been estimated in house. In 1995, they began to use the UNCTAD's database for the terms of trade.

The World Bank essentially uses the IFS data for the developed countries. However, prior to the 1995 CDROM, they constructed their own estimates for developing countries based mainly on commodity prices. The most obvious difference is the extra weight given to the price of oil for many of these countries because of the exclusion of the prices of manufactures and services. The oil shocks are more apparent in the World Bank 1994 database than in the World Bank 1995 or the IFS databases.

For the terms of trade I tried to use the WB95 data where possible. This was the preferred source because it generally provided better coverage historically and across a broader range of countries. The WB95 data were similar to the IFS for developed countries. The WB94 database was not so useful because it will no longer be available in the future and for developing countries it was based mainly on commodity prices, ignoring manufactures and services.

Trade shares

Individual country trade shares were used to weight the terms of trade of each country, although these results are not reported in the paper. I used the measure of openness provided in the PWT5.6: variable 25, Openness, which is the ratio of the sum of imports and exports to GDP in national currencies. In order to avoid introducing extra noise from short term variation in openness, I took a nine-year centred moving average of the trade shares before applying them to the terms of trade. (The moving average was truncated at the end points of the sample.) This still allowed for long term variation in countries' openness.

Nominal GDP

Nominal GDP was needed to construct current account and government fiscal balances as a ratio to GDP.

Nominal GDP in local currency

The two sources used were the IFS (line ‘99b..’ or ‘99b.c’) and the World Bank (1995 CD ROM – GDP at market prices, local currency, NY GDP MKTP CN). Much of the WB95 database, especially for developed countries, comes from the IFS. As a rule the IFS was the preferred source, providing the best coverage historically and most reliable in terms of revisions. (Although, the two sources are very close for most countries.)

Exchange rates

The current account balance is recorded in US dollars. In order to construct the ratio of the current account to GDP, nominal GDP in local currencies had to be converted using the exchange rate. In some cases the official exchange rates published by the IMF in the IFS (line ‘rf’) do not reflect the same international prices that are automatically incorporated into the current account data (almost all countries for which this was a problem are invoiced in US dollars for international transactions). To account for this, I used the World Bank's ‘Atlas’ exchange rate series. These rates are generally based on the official IFS reported rates. Exceptions arise when multiple exchange rates are averaged or when official exchange rates are fixed far from free market levels (for more information refer to the World Tables publication).

I used the IFS exchange rates when the WB95 did not provide the data – primarily in the most recent years in the sample. The only problematic country was Uganda. The official exchange rate (from the IFS) was clearly inappropriate for the late 1970s and early 1980s because it implies implausibly large GDP growth rates (in US dollar terms). So for Uganda, I used the IFS official exchange rate, except for the years 1977–81 for which I took a linear interpolation of nominal GDP in US dollars.

Real GDP

Real GDP was taken mostly from the PWT5.6: variable 2, RGDPCH, real GDP per capita in constant dollars (Chain index, expressed in international prices, base 1985). This was converted to real GDP using the population variable. This data was supplemented by the World Bank 1995 (NY GDP MKTP KD) and the IFS data (line ‘99b.r’ or ‘99b.p’) – mostly to make series as current as possible. As a rule, I used the PWT5.6 data supplemented by the WB95 data (notice that WB95 data is equivalent to IFS data for the developed countries).

Government fiscal balance

The primary data source for the government fiscal balance was the IFS. I had to supplement this with data from the World Bank 1995 and 1994 CDROMs[38] and the hard copy of the IMF's Government Financial Statistics (GFS, 1995).

Solow residuals for OECD countries

Solow residuals were constructed from a Cobb-Douglas production function:

where, Inline Equation is the Solow residual for country i at time t, Y is real GDP, L is employment, K is the capital stock and δ is the share of labour in output. A ‘hat’ over a variable represents the percentage change on the previous year.

Employment data was kindly provided by Aart Kray and comes from Kray and Ventura (1996). The employment data comes originally from the OECD Yearbook of Employment Statistics and is the total civilian employment. The labour share of output data was originally from the OECD National Accounts (it was constructed as the ratio of compensation of employees, MOCOM, to gross domestic expenditure, MOGDPE, both in current local currency units).

Data on the capital stock was taken from the PWT5.6 – variable number 20, KAPW, non-residential capital stock per worker (1985 international prices). This was converted to the level of the capital stock by constructing a series of the number of workers from the real GDP per worker series (number 19, RGDPW) and real GDP (see above).


The WB94 database contains some data on the government fiscal balance that is not in the WB95 database; otherwise, they are very similar. The WB94 CDROM was used more often than the WB95 CDROM for this variable – but it should make little difference. [38]