Revision of Gdp Data for Staff Analysis1
1. Sudan’s national accounts have been plagued by significant weaknesses in quality and timeliness. This has long been pointed out by Technical Assistance (TA) experts.2 The base year is 1981, based on 1968 SNA methodology and the data are published with a 2-year lag.3 Moreover, lack of financing and insufficient staff at the CBS complicate the task at hand, slowing the progress in implementation of TA recommendations.
2. The secession of South Sudan highlighted this challenge by showing the inability of the CBS to properly capture the shock. The performance of the economy in the period 2011–12 based on official data looks implausible given the size of the shock faced by the economy. Although Sudan lost three-quarters of its oil fields (about 15 percent of GDP), a large portion of its land and an equally significant share of population, the official data shows positive GDP growth rates in both years (3.8 and 0.7 percent, respectively).
3. Given these problems, staff has for several years constructed its own GDP series using a bottom-up value-added approach from sectoral data. Sectoral data from the Ministry of Agriculture (crops), Ministry of Animals (livestock), and other official sources was used to come out with GDP figures. However, policy analysis was constrained by the lack of an accurate time series of expenditure-side GDP to facilitate the assessment of the impact of fiscal and monetary policies on economic activity.
4. With US sanctions now revoked and the authorities intending to embark on significant economic reforms, the need for adequate expenditure side GDP data is pressing. Thus, staff has conducted a review of the authorities and IMF GDP data, comparing also with other available datasets such as the BOP, fiscal, and monetary data, and prepared a revised series of expenditure side GDP to guide its analysis.
B. Compiling Revised Estimates of Expenditure-Side GDP data
5. Official data and IMF estimates of nominal and real GDP are broadly in line until 2008. Thus, the headline GDP for the years up to 2008 was taken to be the official data. However, BOP data appeared more stable and reliable than the official national accounts data for exports and imports, so these were used in place of the original national accounts data for that period. Nominal domestic demand was then estimated using the national accounts identity. Real GDP was then estimated by deflating each of the components of nominal GDP by an appropriate deflator; CPI index for final consumption, export and import price indices, and a simple average of CPI and import prices for investment). As earlier indicated, official and Fund estimates are broadly in line over this period. The average nominal GDP growth for this period was 23.6 percent based on the official data compared to 22.4 percent for the IMF estimate, with no tendency for either series to be higher or lower than the other.
Sudan: Nominal Domestic Demand
6. However, official data quality for 2009-14 appears particularly weak. For instance, the export and import series from the official national accounts suddenly shrunk in 2009 and remained at very low levels until 2014. Closer inspection of the data suggested a misplaced decimal point as the culprit. Adjusting the placement of the decimal point yields more plausible exports and imports data (dotted lines in chart), but raised other questions about the accuracy of the domestic demand data. Another concern about the quality of the official GDP series was a large revision of the 2013-2014 real GDP growth rates in early 2017. The CBS raised the growth rates from 5.3 and 1.6 percent to 6.8 and 7 percent, respectively, and the reasons for these revisions were not adequately explained by the staff of the CBS.
Sudan: Exports and Imports at Constant Prices
7. Thus, the official data for 2009-14 was significantly adjusted, building on data from other sources (notably the BOP) and empirical estimations. To calculate the nominal GDP series, BOP data and proxies that correlate with components of domestic demand (Private and public sector consumption and investment) were used. This was a challenge as there are not many alternative indicators available in Sudan.
8. The following specification was used to estimate private consumption growth:
Agriculture production was used as a proxy for household income as a large share of population depend on this economic activity. In addition, imports are another variable that should be correlated to domestic demand. As expected, estimates show that agriculture production and imports growth are positively correlated with private consumption growth.
|Dependent Variable: DLOGPRIV_CONS|
|Method: Least Squares|
|Date: 08/08/17 Time: 09:20|
|Sample (adjusted): 1997 2014|
|Included observations: 18 after adjustments|
|R-squared||0.64||Mean dependent var||0.21|
|Adjusted R-squared||0.59||S.D. dependent var||0.12|
|S.E. of regression||0.08||Akaike info criterion||−2.11|
|Sum squared resid||0.09||Schwarz criterion||−1.96|
|Log likelihood||21.98||Hannan-Quinn criter.||−2.09|
9. The following specification was used to estimate investment growth:
Here the main explanatory variables are credit to the private sector (lag 1 period) and the oil price. Both variables have the correct sign, though only credit to the private sector is statistically significant. However, and owing to the small sample, we should be mindful of the implications for such analysis.
10. Public consumption was calculated as the sum of wages and goods and services from the fiscal accounts.
|Dependent Variable: DLOGGCF2|
|Method: Least Squares|
|Date: 08/08/17 Time: 09:45|
|Sample (adjusted): 1998 2014|
|Included observations: 17 after adjustments|
|R-squared||0.42||Mean dependent var||0.19|
|Adjusted R-squared||0.34||S.D. dependent var||0.26|
|S.E. of regression||0.21||Akaike info criterion||−0.15|
|Sum squared resid||0.60||Schwarz criterion||0.00|
|Log likelihood||4.24||Hannan-Quinn criter.||−0.13|
11. The nominal GDP components were then deflated by their relevant deflators to calculate the real GDP series. The CPI was used to deflate final consumption (private and public), while a simple average of CPI and import prices was used to deflate investment (which uses imported goods).
Real GDP growth
12. The results show a real growth path that is consistently lower than official data suggests. On average, the official real GDP growth rate for 2009-2014 was 4.9 percent (median 5.5 percent) whereas our estimate shows an average of −1.1 percent (median −0.2 percent).
13. This approach was also used to estimate GDP for 2015-16, in the absence of published official data. Actual BOP data was used for exports and imports, while the estimating equations above were used to estimate the components of nominal domestic demand. Appropriate deflators were then applied to the various components to yield estimates of real GDP.
C. Projections for 2017–22
14. Projecting GDP over this period requires forecasting domestic demand, imports, and exports. The approach taken was as follows:
Since the bulk of Sudan’s exports are in a few large sectors (livestock, oil, and gold) forecasting exports is based on discussions with the key government ministries overseeing these sectors and staff’s assessment of the outlook for the external environment.
However, for domestic demand and imports, forecasting equations are estimated.
15. The following specification was used to estimate growth in domestic demand:
Explanatory variables used were agriculture production, total public expenditures, non-oil fiscal revenues, and credit to the private sector. Given the simultaneity bias between domestic demand and non-oil fiscal revenues, however, a restricted regression was estimated where the coefficient for non-oil fiscal revenue was assumed to be −0.2, broadly in line with the likely size of the fiscal revenue multiplier. Regression results indicate that the restriction is not rejected by the data. All estimated coefficients have the correct sign. Agriculture production is again the most significant variable with the largest coefficient, followed by public spending and credit.
|Dependent Variable: DLOGDD_RESTR|
|Method: Least Squares|
|Date: 08/10/17 Time: 17:52|
|Sample (adjusted): 2000 2014|
|Included observations: 15 after adjustments|
|R-squared||0.48||Mean dependent var||0.21|
|Adjusted R-squared||0.33||S.D. dependent var||0.11|
|S.E. of regression||0.09||Akaike info criterion||−1.76|
|Sum squared resid||0.09||Schwarz criterion||−1.57|
|Log likelihood||17.23||Hannan-Quinn criter.||−1.77|
|DLOGDD_RESTR= Domestic demand - 0.2*Non-oil revenues|
16. To project imports growth, we use domestic demand, credit to the private sector, nominal exchange rate and exports as explanatory variables. While some coefficients are not statistically significant, they are of plausible size and sign and help improve the quality of the forecast.
|Dependent Variable: DLOGM|
|Method: Least Squares|
|Date: 08/10/17 Time: 16:53|
|Sample: 1998 2014|
|Included observations: 17|
|R-squared||0.55||Mean dependent var||0.19|
|Adjusted R-squared||0.40||S.D. dependent var||0.19|
|S.E. of regression||0.15||Akaike info criterion||−0.77|
|Sum squared resid||0.26||Schwarz criterion||−0.52|
|Log likelihood||11.54||Hannan-Quinn criter.||−0.75|
17. The estimated set of equations are relevant inputs for medium-term growth projection in Sudan. The aim of this note was twofold. First, to reconstruct a real GDP growth series of the Sudanese economy that better represent its actual economic performance. Second, to estimate a medium-term projection of real GDP growth from the aggregate demand side, which is necessary to assess the implication of monetary and fiscal policies on growth. Nevertheless, data quality and availability requires that staff judgement be a critical part of the process of macroeconomic projections.
Prepared by Gabriel Presciuttini (MCD).
METAC has provided technical assistance to the Central Bureau of Statistics (CBS) to improve the methodology, accuracy and reliability of Sudan’s national accounts. After the 2012 multi-sector statistics mission, several TA missions on national accounts were held in January 2013, August 2014 and November 2015 and April 2017 to assist in implementing the recommendations.
The current actual GDP figures correspond to 2014.