RDP 2001-05: Understanding OECD Output Correlations Appendix B: Data

We consider 17 OECD countries; these and their associated two-digit ISO codes are listed in Table B1. These countries are selected on the basis of data availability across all of the various bilateral variables we consider.

Output

We use quarterly real GDP data to construct four-quarter-ended growth rates (constructed as log fourth differences). The sources for the quarterly GDP data are given in Table B1. To calculate the correlations for the various samples discussed in the text, it is necessary in some instances to take account of the fact that we do not have data for all quarters of the sample. This arises because of limited data, particularly for the earlier years, and also because we use a moving average filter to seasonally adjust some of the series. Where there is missing data, we exclude those observations from the calculation of the bilateral correlations only on a pair wise basis; in other words, for each bilateral correlation, we use the maximum data available.

Bilateral trade in goods and services

The index of bilateral trade intensity we use is the natural logarithm of the following measure:

Where X_ijt is total nominal exports from country i to country j in period t, M_ijt is total nominal imports of country i from country j in period t, and Y_it is nominal GDP for country i in period t. All series are measured in USD.

The bilateral trade flows data are from the IMF International Direction of Trade Statistics. For the period 1960–1993, we use the annual data collected by Frankel and Rose (1998) available at <URL:http://www.haas.berkeley.edu/group/eap/>. For the 1993–2000 period, we use the monthly IMF International Direction of Trade Statistics available on Datastream. The monthly data are aggregated to annual data. All series are reported in USD.

Nominal annual GDP is taken from the OECD Economic Outlook, December 2000. The data are converted to USD using year averages of quarterly nominal USD exchange rates for each country. The exchange rate data are from the IMF International Financial Statistics available on Datastream; the codes are reported in Table B2.

Bilateral foreign direct investment positions

The index of bilateral trade intensity we use is the natural logarithm of the following measure:

Where FI_ijt is the inward forward investment position in year t of country i from country j; FO_ijt is the outward foreign investment position in year t for country i in country j. As before, Y_it is nominal GDP for country i in period t and all series are measured in USD. The foreign investment positions are taken from OECD Direct Investment Statistics Yearbook 1999. In most instances, the foreign investment series are reported in local currencies; these are converted to USD using the exchange rate series discussed above.

In theory the numerators of the two indices should be the same, however because they are obtained (by the OECD) from different statistical agencies there can be discrepancies between the figures. By taking the maximum we get a measure of the greatest degree of exposure that two countries might have with each other; it is also consistent with the measure of bilateral trade intensities discussed above.

Real interest rates

To calculate real interest rates, we use quarterly nominal interest rates and quarterly consumer price indices. From these, we calculate year average interest rates and price levels. Real interest rates are calculated as an ex post measure:

Where i_it is the nominal year average interest rate and P_it is the year average consumer price index. For long interest rates, we use long-term government bonds (usually of 10-year maturity). For short interest rates, we use short-term rates (usually of 3-month maturity). For the price level, we use consumer price indices. All series are available on Datastream. The bilateral series used in estimation, for both short and long rates, is the natural logarithm of the standard deviation of the real interest rate spread:

Real equity returns

To calculate real equity returns we use monthly nominal stock market indices and quarterly consumer price indices. From these, we calculate year average levels and use these to calculate real returns over the year:

Where S_it is the value of the stock index in country i in year t and P_it is the consumer price index described above. The stock market indices are monthly and taken, via Datastream, from either the OECD Economic Outlook or IMF International Financial Statistics. The bilateral series used in estimation is the natural logarithm of the standard deviation of the spread on real equity return:

Exchange rate volatility

We use the standard deviation of the quarterly difference, in logarithms, of the nominal bilateral exchange rates. The data is from Datastream; the codes are identified in Table B2.

Bilateral characteristic variables

We use a number of variables in the empirical models, either as instruments or regressors, which are based upon specific economic or geographic characteristics. The variables typically measure whether or not two countries share a particular characteristic. These variables can be grouped as follows: geographic and language, economic size, industry structure, corporate governance, structural economic reform, and the take-up of new technology.

Geographic and language variables

The geographic and language variables we use are bilateral distance, adjacency, and language; these are taken from Frankel and Rose (1998). The adjacency and linguistic dummies are reported in Table B3; the bilateral distance measure is reported in Table B4.

Economic size

To measure economic size, we use median nominal annual GDP measured in USD for the period 1980–2000. These are constructed from the nominal GDP series identified in Table B2 and converted to USD using the exchange rate series identified in Table B2. For any two countries, we use the natural logarithm of the product of the median GDP measures for both countries as our bilateral index.

Industry structure

This is a measure of the difference between the industry structures of two economies and is based on Imbs (2000); it follows from previous work by Krugman (1991). The index is constructed as follows:

The variable l_ik,t denotes employment in sector k as a share of total employment in country i. M is the number of sectors and t is the period for which the index is computed. For the empirical work in this paper, we use 1989 data disaggregated to the one digit sector level, which corresponds to the middle of the 1980–2000 sample. The data is from the OECD Labour Force Statistics, various issues.

Corporate governance

We use three variables related to corporate governance taken from La Porta et al (1998). These are country of legal origin, accounting standards and firm ownership concentration. The first identifies the origin of the legal system used by a particularly country; there are four possible groups: English, French, German and Scandinavian. Accounting standards rates each country's accounting standards on a scale of 0–100, with high values indicating good quality accounting standards. Firm ownership measures the concentration of ownership of the top ten non-financial firms for each country. These variables are reported in Table B5.

For country of legal origin, we construct a dummy variable that is 1 if a country pair shares a common country of legal origin and 0 otherwise. There are 28 such pairs. We also construct dummy variables specific to each possible origin. For accounting standards, we use as an index the natural logarithm of the product of the two countries accounting standards. We use a similar bilateral index for firm ownership concentration.

Structural economic reform

The structural economic reform bilateral index is constructed from an index prepared by Lehmann Brothers and reported in The Economist (2001). On the basis of a large number of indicators, it rates countries on a scale of 0 to 10 on their structural economic policies (0, worst; 10, best). The measure of structural economic reform for each country is reported in Table B5. The bilateral index is constructed as the natural logarithm of the product of the measures for each country.

Technology adoption

This measures the speed with which a country adopts new technology. We use data on the following three new technologies: mobile phones – number per 1,000 people in 1995; personal computers – number per 1,000 people in 1995 and Internet hosts – number per 10,000 people in 1996. Data on these variables are taken from the World Bank Social Indicators. Countries are ranked according to their adoption of each of the new technologies: those in the top third receive a score of 3, those in the middle third a score of 2 and those in the lowest third a score of 1. The scores are then added across the three technologies to give a total score out of 9. These are reported in Table B6.

An index of difference in speed of adoption is then constructed by taking the absolute difference between the scores of two economies divided by nine. This index then describes how similar two countries are in their adoption of technology.