Anderson, T.W., and Cheng Hsiao, 1982, “Formulation and Estimation of Dynamic Models Using Panel Data,” in Journal of Econometrics, No. 18, pp. 47-82.
Berg, Andrew, Eduardo Borensztein, Sahay Ratna, and Joromin Zettelmeyer , 1999 “The Evolution of Output in Transition Economies: Explaining the Differences,” IMF Working Paper 99/73 (Washington: International Monetary Fund).
Havrylyshyn Oleh, Ivailo Izvorski, and Ron van Rooden, 1998, “Recovery and Growth in Transition Economies 1990-97: A Stylized Regression Analysis,” IMF Working Paper 98/141 (Washington: International Monetary Fund).
Judson, Ruth A. and Ann L. Owen , 1996 “Estimating Dynamic Panel Data Models: A Practical Guide for Macroeconomists,” Federal Reserve Board of Governors (January).
Kikeri, Sunita, 1998, “Privatization and Labor: What Happens to Workers When Governments Divest?,” World Bank Technical Paper No. 396 (Washington: World Bank).
Sevestre, Patrick, and Alain Trognon, 1996, “Dynamic Linear Models,” in The Econometrics of Panel Data: A Handbook of the Theory with Applications, edited by Mátyás and Sevestre (Dordrecht: Kluwer Academic Publishers).
White, Halbert J., 1980, “A Heteroskedasticity-Consistent Covariance Matrix Estimator and A Direct Test for Heteroskedasticity,” in Econometrica, 48:4, pp. 817-838 (May).
This research was carried out as part of a study on the fiscal and macroeconomic impact of privatization, and a condensed version appears as an Appendix in Davis, et al. (2000). I am especially grateful to Jeffrey Davis, Rolando Ossowski, and Tom Richardson for their valuable insights and comments, Will Riordan for his research assistance, and Heather Huckstep and Erik Freas for their administrative assistance. The numerous comments from different departments within the Fund are also greatly appreciated. The author takes sole responsibility for any errors.
An abridged version appears as an appendix in Davis, et al. (2000).
The noise in the debt to GDP series is due to: (1) movements, possibly large, in nominal GDP growth rates that cause significant changes in the debt to GDP ratio; and (2) financing operations, such as the assumption of previously non-budgetary debt, that affect the debt stock without impacting the recorded deficit.
Sevestre and Trognon (1996) discuss the general econometric difficulties and potential solutions. Judson and Owen (1996) address issues pertinent to macroeconomic panels and recommend the use of the Anderson-Hsiao (1982) instrumental variable technique that is employed below.
These are: Argentina, Bolivia, Cote d’Ivoire, Czech Republic, Egypt, Estonia, Hungary, Kazakhstan, Mexico, Mongolia, Morocco, Mozambique, Peru, Philippines, Russia, Uganda, Ukraine, and Vietnam.
A formal test of the proposition that budgetary privatization proceeds are only used to reduce domestic financing when there is a Fund program is rejected. However, there are limited observations without a Fund program, and in some such cases a program may have been under discussion.
The intuition behind this sample is that the contemporaneous impact of privatization proceeds is likely to be most pronounced when there has been a significant change in their magnitude.
Unless otherwise noted, statistical significance refers to a two-sided t-test evaluated at the 10 percent significance level.
It would be difficult to formally establish the directions) of causality between privatization, total revenue, and expenditure. Repeating the expenditure regression with total revenue included as an explanatory variable wipes out the significance of the coefficients on privatization, whereas the reverse (including expenditure in the total revenue regression) does not. While this may provide some indication that it is revenue and not privatization fueling the change in expenditure, it falls well short of formally settling the issue.
To illustrate this point, suppose the public firm paid no taxes, the purchase price was 1 percent of GDP and the private owner earns an annual taxable return of 10 percent (of the purchase price). With a corporate tax rate of 50 percent, this would yield additional corporate tax revenue of only 0.05 percent of GDP.
Given possible shortcomings in each of these techniques, both estimation methods are used. Specifically, the least squares dummy variable (LSDV) estimates are biased. However, the bias on the coefficients for privatization could be quite small, and the Anderson-Hsiao (1982) technique may not perform well when the time-dimension is short (see Judson and Owen, 1996). It is reassuring, therefore, that the LSDV and Anderson-Hsiao (1982) estimates yield the same qualitative results.
The impact in t+1 is calculated as: (0.55*0.13) + .35.