The Discrepancies in GDP Growth Estimates for Sub-Saharan Africa
Xiao Ye
Whyare different GDP growth rates cited for Sub-Saharan Africa (Africa thereafter) even for the same year? There are three main reasons: the data source, the aggregation methodology, and the weighting variable. The basic unit for reporting GDP and its growth is at the national level in the current local currency. Both the country economists from the World Bank and from the IMF provide GDP growth estimates, which may differ from each other. In addition, when the growth rates to be calculated at a more aggregated level, say, the Africa GDP growth in 2004, it is necessary to aggregate all individual country growth ratesinto a regional one. When aggregating, one must choose an aggregationmethod and a weighting variable. The World Bank and the IMF have chosen different aggregation methods andweighting variables. Below is a discussion of how the choices of the data source, the aggregation method and the weighting variable create the GDP growth discrepancies between the two institutions.
I. The Discrepancy due to Different Data Sources
There are three major databases supplying GDP growth rates: the World Bank Africa Region Live Database (AFRLDB), the World Bank World Development Indicator (WDI) database and the IMF World Economic Outlook (WEO) database. The most recent estimates by these two institutions are shown in Figure 1. Historically, IMF has a tendency over predictingSSA GDP growth by one percentage point (See Appendix I). The 2004 growth rates will be likely revised again in September 2005. It is still too soon to tell the GDP growth discrepancies in 2004among the three databases. For historical data, however, discrepancies do exist as in shown in Figure 1.
Figure 1 GDP growth by difference data sources
Data sources: WB WDI, AFRLDB and IMF WEO.
The two databases from the World Bank use exactly the same data sources, but the IMF uses its separate data sources. While the AFRLDB and the World Bank WDI databases take their data from the World Bank country economists,the IMF WEO database takes its data from the IMF country economists. The IMF country economists make their own calculations based on extensive consultation with the relevant government officials and their own observations. The World Bank country economists often use the IMF estimates, but with their own adjustments, especially when concerning the predictions. There, however, is no formal mechanism for the Bank and the IMF country economists to consolidate their growth estimates before reporting to their respective databases. After the reported country level data are entered into the databases, different cleaning procedures and consistency checks may apply. There seems to be no agreement on a set of standard cleaning procedures among the three databases.
II. The Discrepancy Due to the Aggregation Method
The World Bank databases use the same aggregation method, while the IMF uses a different one. GDP growth is essentially a ratio between the current and the last year’s GDP, taking into account of the inflation effect. The essence of the difference between the two aggregation methods is to aggregate GDP growth rates by the ratio of the sum (The Bank’s method) or by the sum of ratios (The Fund’s method).
The World Bank aggregation method (the ratio of the sum)
First, one converts the individual country GDP estimates of the current local currency into the base-year local currency based on the country specific GDP deflators. Second, the GDP of the base-year local currency is converted to the GDP of the base-year $US, using the base-year exchange rate. Third, the GDP estimates of the base-year $US from all Africa countries are added up to obtain a series of the total Africa GDP in the base-year $US (the sum). The Africa GDP growth is then calculated based on this series (the ratio of sum).
The IMF aggregation methodology (the sum of the ratios)
First, one converts the individual country GDP estimates of the current local currency into the base-year local currency based on the country specific GDP deflator. Second, the GDP growth rate is calculated for each country based on thecountry series (the ratio). Third, the growth rates from all counties are summed up using the country GDP of the current $PPP as the weighting variable (the sum of ratios).
III. The Discrepancy Due to the Weighting Variable
The World Bank weighting variable
Bycalculating the aggregated growth rate using the series of Africa GDP in base-year $US value, the World Bank has essentially chosen the GDP series ofthe base-year $US as its weighting variable. Because the conversion is based on the base-year exchange rate, the weight can change for the same country of the same year when the base year changes. The two databases from the World Bank always use the same base year for their aggregated GDP calculation.
The IMFweighting variable
The IMF weighting variable is the country GDP valued in current $PPP. The choice of a base year does not affect the weight since the conversion is based on the current year exchange rate.
IV. The Contribution to the GDP Growth Discrepancy by Data Sources, the Aggregation Method and the Weighting variable
In summary, the Africa GDP growth discrepancy between the World Bank and the IMF are due to three factors: the different data sources, aggregation methods, and weighting variables. Figure 2 shows the percent contribution of these three factors in 2002 and 2004. In 2002, the majority discrepancy is due to the different aggregation methods and weighting variables. In 2004, the data sources contribute a large proportion of the discrepancy, but this is likely to decrease by September 2005 when the WDI and the WEO revise their 2004 growth rates again. As shown in Figure 2, by agreeing on the same aggregation method and the same weighting variable, the three databases can eliminate a large proportion of the discrepancy.
Figure 2The IMF vs. the World Bank: the Contribution of the Weighting Variable and the Growth Rates to the Total Discrepancy of Africa GDP growth estimates
Data sources: see Appendix II.
V. Additional Information: Different Weighting Variables for Different Purposes
In addition to the GDP-weighted GDP growth, researchers also use the population-weighted or unweighted GDP growth for different purposes. The GDP-weighted growth rates are perhaps the most often cited ones due to their ready availability in all databases. This method is especially appropriate for monitoringregional or global GDP growth.
The GDP weighted growth, however, could be inappropriate for other purposes. In the context of Africa, where South Africa and Nigeria account for 50 percent of the total regional output, the GDP-weighted GDP growth at the regional level reflects mainly the growth performances of these two countries. Therefore, some researchersprefer the population-weighted GDP growth, which they arguerepresentative of the growth experienced by a typical African (Collier and O’Connell, 2005). In some cases, the growth experience of each country counts. This is especially true when one wants to examine the growth experience by a group of countries based on the average values of country variables such as country policy performances or natural resource endowments.
The GDP-weighted GDP growth, the population-weighted GDP growth and the unweighted GDP growth can yield quite different growth rates from the same data series. Figure 3 presents the GDP growth rates estimated by each of these three weighting variables, using IMF WEO series.
Figure 3 GDP growth rates by different weighting schemes
Data sources: IMF WEO and staff calculation.
The accumulated effect of the different weighting scheme could be even more pronounced for the GDP per capita growth, which involves more complex weighting scheme than that is used for the GDP growth, due to the added population variable. Figure 4 presents such effects. The “Conventional calculation” in the graph is the method used by WDI database. Based on the GDP and the population weighted GDP growth, Africa region has experienced a negative GDP per capita growth since 1980. This is probably due to the poor performance of a few large countries, includingNigeria, South Africa and The Democratic Republic of Congo. Figure 4 shows that the unweighted GDP per capita growth tells a different story, with more countries having made a good growth performance than not. But these better performing countries weigh less in terms of the population or GDP in the region.
Figure 4GDP per capita growth for SSA, by weighting scheme
Data sources: World Bank WDI and GDF and IMF WEO.
The aggregated GDP growth rates are likely to vary depending on aggregation methods and data sources. One should use one data source as much as possible, but some time it is necessary to combine data from different sources. All three weighting schemes presented above are the correct ways to aggregate GDP growth, depending on the purpose of the research. The best one can do is to provide a consistent series in the same study and to clearly state the data sources and the aggregation methods used.
Appendix I
Systematic overshooting in predicting near future GDP growth and
difficulties in predicting near future export
Each May, IMF puts out predicted values for selected macro economic indicators in its “World Economic Outlook”. The forecasts are 18 months or 6 months ahead, respectively. The information below shows the difficulty in accurately predicting even the current year macro economic indicators, which may influence the decision on lending. Table 1 and figure 1 show that the predicted GDP growth is consistently higher than the realized ones. Table 2 and figure 2 show the volatility of exports. After 1997, the export growth fluctuates a great deal, the predicted values miss the realized values by a large margin. The actual average export (goods only) growth rate between 1997 and 2000 is 4 percent, but the average current year prediction is 8 percent.
Table 1 Forecasts of annual real GDP growth for Sub-Saharan Africa
1995 / 1996 / 1997 / 1998 / 1999 / 2000 / 2001 / 2002Actual real GDP growth rate / 3.8 / 5.2 / 3.5 / 2.6 / 2.2 / 3.1 / 3.9
6 month forecast / 5.2 / 5.4 / 4.4 / 4.1 / 2.9 / 4.2 / 3.9
18 month forecast / 5.1 / 5.4 / 5.2 / 5 / 4.8 / 5.2 / 4.6 / 4.6
Data source: IMF World Economic Outlook, May 1994, 95, 96, 97, 98, 99, 2000, 2001.
Figure 1 Forecasts of annual real GDP growth for Sub-Saharan Africa
Table 2 Sub-Saharan Africa: forecasts of annual growth rate for total exports in goods
1995 / 1996 / 1997 / 1998 / 1999 / 2000 / 2001 / 2002One year forecast / 14.2 / 7.9 / 6.6 / -3.2 / 1.8 / 21.4 / 1.4 / 3.1
Current year forecast / 12.3 / 9.4 / 6.1 / 6.6 / 9.1 / 11.9 / -0.1
Actual export growth / 18.5 / 11.0 / 1.5 / -13.8 / 5.6 / 22.0
Data source: IMF World Economic Outlook, May 1994, 95, 96, 97, 98, 99, 2000, 2001.
Table 2 Sub-Saharan Africa: forecasts of annual growth rate for total exports in goods
Appendix II
World Bank and IMF estimates of country growth rates and weights
GDP growth estimates / Weights / GDP growth estimates / Weights2002 / 2002 / 2002 / 2002 / 2004 / 2004 / 2004/ / 2004
IMF / WDI / IMF / WDI / IMF / WDI / IMF / WDI
Benin / 6.0 / 6.0 / 0.6 / 0.8 / 3.0 / 3.0 / 0.6 / 0.8
Botswana / 5.0 / 4.4 / 1.2 / 1.8 / 5.2 / 3.8 / 1.2 / 1.8
Burkina Faso / 5.2 / 4.4 / 1.1 / 0.9 / 4.8 / 3.9 / 1.2 / 0.9
Burundi / 4.5 / 4.5 / 0.4 / 0.2 / 5.5 / 5.5 / 0.4 / 0.2
Cameroon / 6.5 / 4.2 / 2.8 / 3.0 / 4.3 / 5.0 / 2.8 / 3.0
Cape Verde / 5.0 / 4.6 / 0.2 / 0.2 / 4.0 / 5.5 / 0.2 / 0.2
CAR / -0.6 / -0.8 / 0.4 / 0.3 / 0.9 / 5.8 / 0.3 / 0.3
Chad / 9.9 / 9.9 / 0.7 / 0.5 / 30.5 / 31.0 / 1.0 / 0.7
Comoros / 2.3 / 2.5 / 0.1 / 0.1 / 1.9 / 1.8 / 0.1 / 0.1
Congo, Dem. R. / 3.5 / 3.5 / 2.7 / 1.3 / 6.8 / 6.3 / 2.8 / 1.4
Congo, Rep. / 5.4 / 3.5 / 0.3 / 1.1 / 4.0 / 4.0 / 0.3 / 1.0
Cote d'Ivoire / -1.5 / -1.6 / 2.2 / 3.2 / -0.9 / 1.8 / 2.0 / 2.9
Equatorial Guinea / 9.6 / 17.6 / 1.0 / 0.5 / 34.2 / 10.0 / 1.2 / 0.6
Eritrea / 0.6 / 0.7 / 0.3 / 0.2 / 1.8 / 1.8 / 0.3 / 0.2
Ethiopia / 1.6 / 2.7 / 4.4 / 2.2 / 11.6 / 11.9 / 4.3 / 2.2
Gabon / 0.0 / 0.0 / 0.7 / 1.5 / 1.9 / 2.0 / 0.7 / 1.5
Gambia, The / -3.2 / -3.2 / 0.2 / 0.1 / 7.7 / 8.3 / 0.2 / 0.1
Ghana / 4.5 / 4.5 / 3.7 / 1.7 / 5.5 / 5.2 / 3.8 / 1.7
Guinea / 4.2 / 4.2 / 1.4 / 1.0 / 2.5 / 2.6 / 1.3 / 1.0
Guinea-Bissau / -7.2 / -7.2 / 0.1 / 0.1 / 4.3 / 1.7 / 0.1 / 0.1
Kenya / 1.1 / 1.1 / 2.7 / 3.3 / 3.1 / 2.4 / 2.6 / 3.1
Lesotho / 4.5 / 3.8 / 0.4 / 0.3 / 2.3 / 3.0 / 0.4 / 0.3
Madagascar / -12.7 / -12.7 / 1.0 / 1.1 / 5.2 / 5.3 / 1.1 / 1.2
Malawi / 2.1 / 1.8 / 0.5 / 0.5 / 4.3 / 3.6 / 0.5 / 0.5
Mali / 4.3 / 4.4 / 0.9 / 0.9 / 2.2 / 4.7 / 0.9 / 0.9
Mauritania / 4.1 / 3.3 / 0.5 / 0.3 / 5.2 / 5.0 / 0.5 / 0.3
Mauritius / 3.4 / 4.4 / 1.1 / 1.5 / 4.4 / 5.0 / 1.1 / 1.5
Mozambique / 7.4 / 7.4 / 1.7 / 1.4 / 7.8 / 8.4 / 1.8 / 1.5
Namibia / 2.5 / 2.5 / 1.0 / 1.1 / 4.4 / 3.8 / 1.0 / 1.1
Niger / 3.0 / 3.0 / 0.8 / 0.6 / 0.9 / 4.1 / 0.8 / 0.6
Nigeria / 1.5 / 1.5 / 11.2 / 13.4 / 3.5 / 4.1 / 11.9 / 14.3
Rwanda / 9.4 / 9.4 / 0.9 / 0.6 / 4.0 / 5.9 / 0.9 / 0.7
Sao Tome and P. / 5.0 / 4.1 / 0.0 / 0.0 / 6.0 / 6.5 / 0.0 / 0.0
Senegal / 1.1 / 1.1 / 1.4 / 1.4 / 6.0 / 6.0 / 1.4 / 1.5
Seychelles / 1.3 / 0.3 / 0.1 / 0.2 / -2.0 / -2.0 / 0.1 / 0.2
Sierra Leone / 27.5 / 6.3 / 0.3 / 0.2 / 7.4 / 7.7 / 0.3 / 0.2
South Africa / 3.6 / 3.6 / 38.8 / 41.4 / 3.7 / 2.6 / 37.4 / 40.0
Sudan / 6.0 / 6.0 / 5.5 / 4.2 / 7.3 / 6.0 / 5.8 / 4.3
Swaziland / 2.8 / 3.4 / 0.4 / 0.4 / 2.1 / 1.7 / 0.4 / 0.4
Tanzania / 7.2 / 7.2 / 1.8 / 3.1 / 6.3 / 6.3 / 1.8 / 3.3
Togo / 4.5 / 4.1 / 0.6 / 0.4 / 2.9 / 3.0 / 0.6 / 0.4
Uganda / 6.8 / 6.8 / 3.3 / 2.0 / 5.9 / 5.7 / 3.3 / 2.1
Zambia / 3.3 / 3.3 / 0.7 / 1.1 / 5.0 / 3.5 / 0.7 / 1.1
Data sources: World Bank WDI, AFRLDB and IMF WEO databases.
1
For further questions please contact Xiao Ye at .
This appendix was authored by Gelb and Ye in 2001.