Академический Документы
Профессиональный Документы
Культура Документы
PA U L C O O K and Y U I C H I R O U C H I D A
I. INTRODUCTION
Paul Cook and Yuichiro Uchida, Centre on Regulation and Competition, Institute for
Development Policy and Management, University of Manchester, UK; correspondence: yuichiro-
uchida@man.ac.uk.
The Journal of Development Studies, Vol.39, No.6, August 2003, pp.121–154
PUBLISHED BY FRANK CASS, LONDON
396jds05.qxd 29/08/03 14:13 Page 122
I I . T H E O RY A N D E M P I R I C A L A N A LY S I S
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 123
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
study, using Probit and Tobit models, found that privatisation positively
affected GDP growth and that the effect on growth was more significant for
activities of a public goods type than for other sectors. The study concludes
that on average institutional reform increased economic growth from 0.8 per
cent to 1.5 per cent between the sub-periods 1984–88 and 1988–92.
The second study was published by the IMF in 2000 [Barnett, 2000].
This study investigated the effect of privatisation on real GDP growth,
unemployment, and investment. Only twelve developing countries were
used in this 18-country study. The rest consisted of transitional economies.
The empirical analysis from this study strongly supported the hypothesis
that privatisation was positively correlated with real GDP growth. They
found that privatisation of a one per cent of GDP was associated with an
increase in the real growth rate of 0.5 per cent in period one and 0.4 per cent
in period two. These periods vary for each country to reflect periods of
active privatisation, but the precise span of years for the study is not
specified. It has also been suggested that the privatisation variable used in
Barnett’s is likely to capture the positive impact of a general regime change
towards better economic policies [Davis, Ossowski, Richardson and
Barnett, 2000].
I I I . M E T H O D O L O G Y A N D D ATA
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 125
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
Privatisation Variable
The most crucial question is how to measure the magnitude of privatisation
in a given country. Plane [1997] has used the cumulative proceeds from
privatisation during the period 1988–92 as a share of GDP in 1990 and a
dummy variable based on the information obtained from this indicator. A
similar indicator is used in this study, although the average GDP during the
sample period is used as a weight instead of choosing a particular year. This
reduces the arbitrariness from choosing a particular year for which the
sample countries might experience a variety of external or internal shocks.
In addition, the use of a dummy variable for privatisation is dropped
since it cannot convey information on the magnitude of privatisation into
the regression.
The dataset for privatisation is summarised in Table 1. The availability
of privatisation data largely determines the number of sample countries and
the sample period. As a result, 63 developing countries and the period
1988–97 are selected. The dataset shows the highly concentrated pattern of
privatisation, despite its adoption in a wide range of countries. Latin
American countries dominate the table in terms of privatisation as a share
of GDP. The dataset and full list of other variables corresponding to these
countries are shown in Appendix 2.
To reduce the risk that the privatisation variable reflects other policy and
structural reforms at the macroeconomic or aggregate level, it is necessary
to check if the effect of this variable can be isolated. This can be achieved
by examining the relationship between some of the policy variables that
might be captured by the privatisation variable. The effects of the
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008 396jds05.qxd
TA B LE 1
C U M U L AT I V E P R O C E E D S F R O M P R I VAT I S AT I O N ( C P P ) A N D AV E R A G E G D P ( A G D P ) D U R I N G T H E P E R I O D 1 9 8 8 – 9 7
(IN US$ MILLIONS)
11 Papua New Guinea 223.6 4397.3 5.1 43 Thailand 1378.4 296925.5 0.5
12 Bolivia 884.2 17720.7 5.0 44 Tunisia 171.0 37380.8 0.5
13 Brazil 34559.4 877419.5 3.9 45 Oman 60.1 13370.7 0.4
14 Ghana 872.6 23098.0 3.8 46 Ecuador 169.4 48336.9 0.4
15 Venezuela 5914.1 168554.8 3.5 47 Mali 21.9 6015.3 0.4
16 Colombia 5685.1 215739.9 2.6 48 Costa Rica 56.7 18735.7 0.3
17 Barbados 51.0 2039.5 2.5 49 Paraguay 42.0 15767.2 0.3
18 Morocco 1846.7 77307.4 2.4 50 Cameroon 41.1 23779.9 0.2
19 South Africa 6064.4 257829.9 2.4 51 Guinea-Bissau 0.5 235.2 0.2
20 Cote d’Ivoire 476.3 20874.1 2.3 52 Malawi 10.8 5549.1 0.2
21 Sri Lanka 725.9 34640.0 2.1 53 Bangladesh 60.3 98712.6 0.1
22 Nicaragua 130.2 6520.6 2.0 54 Burkina Faso 6.4 8411.6 0.1
23 Egypt 2777.7 142925.9 1.9 55 Burundi 4.2 4332.2 0.1
24 Philippines 3810.0 204678.8 1.9 56 Guatemala 43.4 33999.6 0.1
25 Senegal 191.4 12344.0 1.6 57 Nepal 15.0 18530.0 0.1
26 Turkey 3843.4 306790.8 1.3 58 Sierra Leone 1.6 2396.3 0.1
27 Mozambique 110.6 8932.9 1.2 59 Uruguay 17.0 23925.7 0.1
28 Pakistan 1951.0 161154.2 1.2 60 Iran 18.1 209417.6 0.0
29 Chile 1484.1 129306.1 1.1 61 Mauritania 1.1 3357.2 0.0
30 Benin 56.5 5692.5 1.0 62 Vietnam 2.6 72856.0 0.0
31 Indonesia 5162.8 494440.0 1.0 63 Yemen 0.8 8544.5 0.0
32 Nigeria 763.5 87263.8 0.9
Source: Primary data from World Bank Privatization Database and Privatisation International (various issues), and World Bank Global Development
Network Growth Database.
396jds05.qxd 29/08/03 14:13 Page 128
F I GURE 1
T HE CORRE L AT I ON BE T WE E N PRIVATISATIO N A N D
GOVE RNME NT BUDGET D EFIC IT
10
9
8
7
PRIV/GDP
6
5
4
3
2
1
0
-30 -25 -20 -15 -10 -5 0 5 10 15
BUD/GDP
Note: PRIV/GDP and BUD/GDP are the average percentages during the period 1988–97.
Primary data from World Bank Privatisation Database and World Bank Global
Development Network Growth Database.
396jds05.qxd 29/08/03 14:13 Page 129
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 129
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
F I GURE 2
T HE CORRE L AT I ON BE T WE E N PRIVATISATIO N A N D
WORLD BANK ADJUSTMENT LOANS
12
10
8
PRIV/GDP
0
0 2 4 6 8 10 12 14 16
SAL/GDP
Note: PRIV/GDP and SAL/GDP are the average percentage during the period 1988–97. Primary
data from World Bank Privatisation Database, World Bank Global Development Network
Growth Database, and World Bank SAL Database.
396jds05.qxd 29/08/03 14:13 Page 130
The analysis is conducted in two parts. In the first, the control variables are
examined to find appropriate specifications for the I variables. In the second
part, the results of introducing privatisation are reported.
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 131
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 133
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
TABL E 2
T HE E BA RE S ULT S OF T HE I VARIA BLES (AT 0.05 LEV EL)
Specification I
Specification II
Specification III
TABL E 2 (cont.)
Specification IV
Notes: Dependent variable: GYP. The base results are calculated excluding the M and Z
variables. Note that following Levine and Renelt [1992] (see Table 1, notes (a), p.947),
where an EBA indicates a robust result, but the base result is statistically insignificant, the
overall result is interpreted as fragile. ** = statistically significant at 0.05 level, and ***
= statistically significant at 0.01 level.
positive, whereas they are negative for GOVC, confirming previous studies
[e.g., Barro, 1991].
In summary, LGYP exhibits fragile results, and LIFE and POP appear to
be sensitive to different combinations of other variables, resulting in
ambiguous results. On the other hand, the results for GDI and GOVC are
always robust. Nevertheless, it is difficult to choose a particular
specification for the control variables from these results. Since a continued
search for the most suitable combination of the control variables is beyond
the scope of this study and may prove impossible, as some past studies have
demonstrated, then all four specifications are used to conduct an EBA for
the privatisation variable.
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 135
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
TABL E 3
T H E E BA RE S ULT S ( AT 0.05 L E VE L ) F OR TH E PR IVATISATIO N VA R IA BLE
Notes: Dependent variable: GYP. Note that following Levine and Renelt [1992] (see Table 1,
notes (a), p.947), in the case that the EBA indicates a robust result, but the base result is
statistically insignificant, the overall result is interpreted as fragile.
396jds05.qxd 29/08/03 14:13 Page 136
Two cautionary notes are required at this point. First, it is known that the
inclusion or exclusion of certain control variables sometimes eliminates a
bivariate relationship [Easterly and Rebelo, 1993]. It is thus plausible to
argue that our inconclusive results are due to an inadequate choice or
combinations of the control variables. In order to examine this possibility,
all possible combinations of the control variables and the privatisation
variable were scrutinised by applying the EBA technique. This exercise
reconfirmed our conclusion by showing that no combination of control
variables exists that would establish a robust partial correlation between
privatisation and economic growth.
Second, it is known that cross-country regression analysis is sensitive to
the inclusion or exclusion of a particular country. This may arise due to the
existence of influential observations and/or outliers in the regressions. This
possibility is also examined by checking whether or not there are any
influential observations or outliers in our basic regression models. It should,
however, be pointed out that there is no universally accepted indicator that
is able to identify influential observations and outliers. For instance, Donald
and Maddala [1993] argue that the examination of studentised residuals is
the most appropriate method to identify influential observations, and to
detect outliers leverage should also be examined. On the other hand,
Beckman and Cook [1983] and Fiebig [1992] argue that large studentised
residuals may not necessarily point to influential observations, and Fiebig
[1992] specifically points out that influential observations may actually be
related to small residuals.
In addition, even if influential observations and outliers are identified, it
may prove difficult to deal with them. Influential observations and outliers
may arise as a result of data errors and model misspecifications. Suitable
corrections may be hampered because of poor data, especially for
developing countries, and simply deleting the countries in question may still
be appropriate. On the other hand, influential observations may represent
natural variation which provides valuable information, and eliminating
these observations would clearly be inappropriate. Even if influential
observations are kept in the regression, however, the potential effect on the
results of eliminating them ought to be noted [Fiebig, 1992]. One way to
accommodate influential observations in the regression is to apply robust
regression estimators instead of OLS (see Rousseeuw and Leroy [1987] for
a detailed account of robust regression estimators). It should also be noted
that even these detection techniques and robust estimators may fail to detect
outliers [Luke and Schaffer, 2000].
Accepting these limitations, studentised residuals were examined in an
attempt to detect influential observations in our basic regression models. We
also applied some of the influential observation and outlier detection
396jds05.qxd 29/08/03 14:13 Page 137
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 137
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
techniques, DFITS [Welsch and Kuh, 1977], Cook’s Distance [Cook, 1977],
and Welsch Distance [Welsch, 1982]. These techniques, in essence,
encapsulate the information in the leverage versus residual-squared plot into
a single statistic, creating an index that is influenced by the size of the
residuals or outliers and the size of leverage (see Appendix 3 for the
calculation procedures). In addition, we tested for influential observations
and outliers specifically to examine the privatisation variable by applying
DFBETAs [Belsley, Kuh and Welsch, 1980]. The application of DFBETAs
permits us to examine the difference between the coefficients of a specific
variable when a particular observation is included or excluded
Table 4 reports the results of the tests for influential observations.
Although it is difficult to compare the indices across the different
specifications and techniques, these tests identified 15 countries as potential
outliers or a total of 95 observations as influential. Singapore and Malaysia
contributed significantly to the influential observations, 17.9 per cent and
14.7 per cent respectively, followed by Oman 13.7 per cent, and to a lesser
extent Sierra Leone 8.4 per cent. These four countries generated more than
50 percent of the influential observations. In particular, Welsch Distance
indicates that only Malaysia, Oman, and Singapore are potential outliers.
More specifically, of those 20 test results shown in Table 4, Singapore
appears as a potential outlier in 17 of them, Malaysia appears in 14 tests,
and Sierra Leone is identified in eight tests. Oman was detected as
influential by 13 test results only in the first, second, and third
specifications. In addition, the combination of Malaysia and Singapore
appear in 14 test results, including all the test results in the third and fourth
specifications. Furthermore, the results of DFBETAs suggest that these two
countries have substantial positive effects on the privatisation variable, in
particular in the third and fourth specifications.
After identifying the influential observations and thus the potential
outliers in our basic regression models, we attempted to re-estimate the
baselines for privatisation in two steps. First, we re-estimated the baselines
for privatisation by applying two robust regression estimators, the median
least squares (MLS) and a robust regression estimator using iteratively
reweighted least squares, including all of the influential observations (see
Rousseeuw and Leroy [1986] and Luke and Schaffer [2000] for details of
these estimators). We then re-estimated the baselines using OLS by
eliminating all the influential observations. Second, we deleted the
influential observations one by one according to the higher absolute value
of each of the indicators for influential observations. Following these steps,
we could not find any statistically significant results for privatisation in the
first and second specifications but found a number of statistically
significant results for privatisation in the remaining specifications.
396jds05.qxd 29/08/03 14:13 Page 138
TABL E 4
T E S T S F OR I NF L UE NT I AL OBS E RVATIO N S A N D O U TLIERS
Specification I
1 Uganda 2.563 Jamaica –0.543 Oman 1.065 Jordan 0.305 Oman 10.09
2 Jamaica –2.559 Malaysia 0.437 Singapore 0.862 Sierra Leone 0.231 Singapore 7.648
3 Cameroon –2.514 Singapore 0.364 Vietnam 0.789 Jamaica 0.194
4 Jordan –2.341 Peru –0.356 Uganda 0.709 Oman 0.184
5 Sierra Leone –2.041 Oman –0.290 Malaysia 0.672 Singapore 0.120
6 Jordan 0.255 Vietnam 0.099
7 Uganda 0.077
8 Malaysia 0.075
Specification II
1 Cameroon –2.959 Jamaica –0.510 Oman 2.327 Oman 0.687 Oman 23.69
2 Oman 2.837 Oman –0.459 Singapore 0.742 Sierra Leone 0.247
3 Uganda 2.321 Peru –0.456 Vietnam 0.733 Jordan 0.235
4 Sierra Leone –2.287 Malaysia 0.409 Uganda 0.727 Jamaica 0.156
5 Jamaica –2.196 Singapore 0.279 Singapore 0.077
6 Jordan –2.148 Cameroon 0.254 Vietnam 0.074
7 Uganda 0.070
8
Specification III
1 Cameroon –2.810 Malaysia 0.756 Oman 1.327 Sierra Leone 0.323 Oman 13.38
2 Sierra Leone –2.549 Singapore 0.443 Singapore 1.048 Oman 0.285 Singapore 9.074
3 Malaysia 2.324 Peru –0.434 Malaysia 0.971 Singapore 0.170 Malaysia 8.289
4 Singapore 2.296 Korea –0.366 Korea 0.641 Malaysia 0.146
5 Thailand –0.302 Barbados 0.066
6 Argentina –0.294 Korea 0.065
7 Oman –0.260
Specification IV
1 Singapore 2.616 Malaysia 0.794 Singapore 1.081 Sierra Leone 0.326 Singapore 9.213
2 Cameroon –2.411 Singapore 0.537 Malaysia 0.987 Singapore 0.212 Malaysia 8.402
3 Malaysia 2.398 Korea –0.366 Korea 0.646 Malaysia 0.180
4 Sierra Leone –2.325 Peru –0.340 Barbados 0.100
5 Thailand –0.303 Korea 0.080
6 Jamaica –0.295 Jamaica 0.068
7 Jordan 0.065
396jds05.qxd 29/08/03 14:13 Page 139
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 139
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
Accordingly, the first and second specifications were dropped from any
further examination.
Correspondingly, we re-estimated the baselines for privatisation by
applying the two robust regression estimation techniques which enabled us
to incorporate the influence of the influential observations or potential
outliers in the regressions without deleting them. The results are reported in
Table 5 in the third and fourth columns of each specification. In the third
specification, we found statistically significant results for privatisation from
MLS (at the 0.10 level) and Robust Regressions (at the 0.01 level). In the
case of the fourth specification, we could not find statistically significant
results for privatisation from the MLS but found the statistically significant
results from Robust Regression (at 0.01 level). As for the MLS results from
specification four, pseudo R2 is too low, and the signs of the coefficients,
except those for PRIV and POP, changed. In essence, we have found that a
robust partial correlation between privatisation and economic growth can be
established in these specifications, in particular in the third specification, by
using the robust estimators instead of OLS and without deleting the
influential observations.
The results of excluding all the influential observations using OLS (a
total of 10 models were estimated) are shown from the fifth to the last
column in Table 5. In these specifications, we found statistically significant
results for privatisation at the 0.01 level (shown in column (1) in the third
specification and column (6) in the fourth specification) by deleting all the
influential observations identified by studentised residuals. Similarly, we
found statistically significant results for privatisation in the fourth
specification (column 10) at the 0.05 level and in the third specification
(column 5) at the 0.10 level by eliminating all the influential observations
identified by Welsch Distance. As for Cook’s Distance, we also found
statistically significant results for privatisation at the 0.05 level (the third
specification (column 4)) and at the 0.10 level (the fourth specification
(column 9)).
In contrast, the elimination of all the influential observations identified
by DFBETAs resulted in statistically insignificant results for privatisation in
both specifications. In terms of the deleted influential observations
identified by DFITS, the results are somewhat mixed. We found a
statistically significant result for privatisation at the 0.10 level in the fourth
specification (column 8) and statistically insignificant result in the third
specification (column 3). In sum, we tested ten models using OLS in the
third and fourth specifications and found statistically significant results in
seven models.
The results of eliminating the influential observations to examine
whether or not these have led to increases in R2 and/or the statistical
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
TA B L E 5
T H E B A S E L I N E R E S U LT S F O R P R I VAT I S AT I O N
Specification III
396jds05.qxd
OLS MLS Robust Regression (1) Studentised (2) DFBETAs (3) DFITS (4) Cook’s (5) Welsch
Specification IV
OLS MLS Robust Regression (6) Studentised (7) DFBETAs (8) DFITS (9) Cook’s (10) Welsch
Notes: Dependent variable: GYP. The models (1) - (10) were estimated using OLS. * = statistically significant at 0.10 level , ** = statistically significant at
0.05 level, and *** = statistically significant at 0.01 level. t-statistics are in parentheses.
396jds05.qxd 29/08/03 14:13 Page 141
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 141
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
TABL E 6
T H E B A S E L I N E R E S U LT S F O R P R I VAT I S AT I O N E X C L U D I N G
S E L E CT E D I NF L UE NT I AL O BSERVATIO N S
Specification III
Specification IV
Notes: Dependent variable: GYP. These models were estimated using OLS. * = statistically
significant at 0.10 level , ** = statistically significant at 0.05 level, and *** = statistically
significant at 0.01 level. . t-statistics are in parentheses. The likely outliers excluded are
shown in the order of higher absolute values.
396jds05.qxd 29/08/03 14:13 Page 143
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 143
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
TABL E 7
T H E E BA RE S ULT S ( AT 0.05 L E VE L ) F OR TH E PR IVATISATIO N VA R IA BLE
E XCL UDI NG OU TLIER S
Cook’s (13) Min –0.357 –3.23 FDI, SDM3, DEBT Sierra Leone,
Max –0.209 –2.11 GDI, FDI, LA Oman, Robust
Base –0.256** –2.40 – Singapore,
Barbados
Cook’s (15) Min –0.383 –3.34 OPEN, FDI, DEBT Sierra Leone,
Max –0.214 –2.02 FDI, EA, LA Sinagpore, Robust
Base –0.288** –2.61 – Barbados
Notes: Dependent variable: GYP. The EBA results for PRIV are obtained by including the control
variables and three regressors from the pool of the Z variables. ** = statistically significant
at 0.05 level. The likely outliers excluded are in the order of higher absolute values.
negative partial correlation between these two variables using the full 63-
country sample with the robust regression estimators. These results indicate
that the privatisation variable is sensitive to combinations of control
variables and the inclusion of certain countries. In other words, in order to
detect a significant result for the privatisation variable, great care is required
in the selection of a specific conditioning information set and countries.
V. C O N C L U S I O N
The analysis has shown that there is a robust partial correlation between
privatisation and economic growth, suggesting that privatisation has
contributed negatively to economic growth. This conclusion is contrary to
the results obtained by Plane [1997] and Barnett [2000]. The analysis in this
study has used a different approach to that adopted in Plane’s study. By
using an extreme bound analysis and a larger sample of countries we have
been able to undertake a more comprehensive and rigorous analysis of the
396jds05.qxd 29/08/03 14:13 Page 144
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 145
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
REF ERENCES
Alchian, A., 1965, ‘Some Economics of Property Rights’, II Politico, Vol.30, pp.816–29.
Alesina, A. and R. Petrotii, 1993, ‘Income Distribution, Political Instability and Investment’,
NBER Working Paper, No.4486.
Barnett, S., 2000, ‘Evidence on the Fiscal and Macroeconomic Impact of Privatisation’, IMF
Working Paper, July, Washington, DC: IMF.
Barro, R., 1991, ‘Economic Growth in a Cross-Section of Countries’, Quarterly Journal of
Economics, Vol.106, No.2, pp.407–43.
Barro, R. and X. Sala-i-Martin, 1995, Economic Growth, New York: McGraw-Hill.
Beckman, R. and R. Cook, 1983, ‘Outliers’, Technometrics, Vol.25, pp.119–49.
Belsey, D., Kuh, E. and R. Welsch, 1980, Regression Diagnostics, New York: John Wiley.
Bennell, P., 1997, ‘Privatisation in Sub-Saharan Africa: Progress and Prospects during the
1990s’, World Development, Vol.25, No.11, pp.1785–803.
Blomstrom, M., Lipsey, R. and M. Zejan, 1993, ‘Is Fixed Investment the Key to Economic
Growth?’ NBER Working Paper, No.4436.
Boubakri, N. and J. Cosset, 1998, ‘The Financial and Operating Performance of Newly Privatised
Firms: Evidence from Developing Countries’, unpublished draft, Universite Laval.
Boycko, M., Shleifer, A. and R. Vichny, 1996, ‘A Theory of Privatisation’, Economic Journal,
Vol.106, pp.309–19.
Cook, P., 1999, ‘Privatisation and Utility Regulation in Developing Countries: The Lessons So
Far’, Annals of Public and Cooperative Economics, Vol.70, No.4, pp.549–87.
Cook, P., 2001, ‘Competition and its Regulation: Key Issues’, paper presented at an international
workshop on ‘Competition and Regulation’, University of Manchester, 9 Feb. 2001
Cook, P. and C. Kirkpatrick, 1988, Privatisation in Less Developed Countries, Brighton:
Weatsheaf.
Cook, P. and R. Fabella, 2002, ‘The Welfare and Political Economy Dimensions of Private vs.
State Enterprise’, The Manchester School, Vol.70, No.2, pp.246–61.
Cook, R., 1977, ‘Detection of Influential Observations in Linear Regression’, Technometrics,
Vol.19, pp.15–18.
Cooley, T. and S. Leroy, 1981, ‘Identification and Estimation of Money Demand’, American
Economic Review, Vol.71, No.5, pp.825–44.
Davis, J., Ossowski, R., Richardson, T. and S. Barnett, 2000, ‘Fiscal and Macroeconomic Impact
of Privatisation’, Occasional Paper, No.194, Washington, DC: IMF.
DeLong, J. and L. Summers, 1991, ‘Equipment Investment and Economic Growth’, Quarterly
Journal of Economics, Vol.106, No.2, pp.445–502.
Deravi, K., Hegji, C. and D. Moberly, 1990, ‘Government Debt and the Demand for Money: An
Extreme Bound Analysis’, Economic Inquiry, Vol.28, No.2, pp.390–401.
Donald, S. and G. Maddala, 1993, ‘Identifying Outliers and Influential Observations in
Econometric Models’, in G. Maddala, C. Rao and H. Vinod (eds.), Handbook of Statistics,
Vol.2, Amsterdam: Elsevier.
Doppelhofer, G., Miller, R. and X. Sala-i-Martin, 2000, ‘Determinants of Long-Term Growth: A
Bayesian Averaging of Classical Estimates (BACE) Approach’, OECD Working Paper, No.266.
Easterly, W. and S. Rebelo, 1993, ‘Fiscal Policy and Economic Growth: An Empirical
Investigation’, Journal of Monetary Economics, Vol.32, No.2, pp.417–58.
Fiebig, D., 1992, ‘Diagnostic Checking in Practice: The Case of Outliers and Influential Data’,
mimeo, University of Sydney.
Folster, S. and M. Henrekson, 2001, ‘Growth Effects of Government Expenditure and Taxation
in Rich Countries’, European Economic Review, Vol.45, No.8, pp.1501–20.
Galal, A., Jones, L., Tandon, P. and I. Vogelsang, 1994, Welfare Consequences of Selling Public
Enterprises: An Empirical Analysis, New York: Oxford University Press.
Gentleman, J. and M. Wilk, 1975, ‘Detecting Outliers II: Supplementing the Direct Analysis of
Residuals’, Biometrics, Vol.31, No.2, pp.387–410.
Harberger, A., 1987, ‘Comment’, NBER Macroeconomics Annual, pp.255–8.
Jensen, M. and W. Meckling, 1976, ‘Theory of the Firm: Managerial Behavior, Agency Costs,
and Ownership Structure’, Journal of Financial Economics, Vol.3, No.4, pp.305–60.
396jds05.qxd 29/08/03 14:13 Page 146
Kalemili-Ozcan, S., Ryder, H. and D. Weil, 2000, ‘Mortality Decline, Human Capital Investment,
and Economic Growth’, Journal of Development Economics, Vol.62, No.1, pp.1–23.
Kikeri, S., 1998, ‘Privatisation and Labor: What Happens to Countries When Governments
Divest’, World Bank Technical Paper, No.396, Washington, DC: World Bank.
Kikeri, S., Nellis, J. and M. Shirley, 1994, Privatisation: The Lessons of Experience, Washington,
DC: World Bank.
Laffont, J. and J. Tirole, 1991, ‘Privatisation and Incentives’, Journal of Law, Economics and
Organisation, Vol.7, No.1, pp.84–105.
Leamer, E., 1978, Specification Searches: Ad Hoc Inference from Non-Experimental Data, New
York: Wiley.
Leamer, E., 1983, ‘Let’s Take the Con Out of Econometrics’, American Economic Review,
Vol.73, No.1, pp.31–43.
Leamer, E., 1985, ‘Sensitivity Analysis Would Help’, American Economic Review, Vol.75, No.3,
pp.308–13.
Leamer, E. and L. Herman, 1983, ‘Reporting the Fragility of Regression Estimates’, Review of
Economics and Statistics, Vol.65, No.2, pp.306–17.
Levine, R. and D. Renelt, 1992, ‘A Sensitivity Analysis of Cross-Country Growth Regressions’,
American Economic Review, Vol.82, No.4, pp.942–63.
Levine, R. and S. Zervos, 1993, ‘What We Have Learned About Policy and Growth from Cross-
Country Regressions?’ American Economic Review, Vol.83, No.2, pp.426–30.
Londregan, J. and K. Poole, 1990, ‘Poverty, the Coup Trap, and the Seizure of Executive Power’,
World Politics, Vol.423, No.2, pp.151–83.
Lorgelly, P. and P. Owen, 1999, ‘The Effect of Female and Male Schooling on Economic Growth
in the Barro-Lee Model’, Empirical Economics, Vol.24, No.3, pp.537–57.
Luke, P. and M. Schaffer, 2000, ‘Wage Discrimination in Russia: An Econometric Investigation’,
William Davidson Institute Working Paper Series, No.295, University of Michigan Business
School.
Mankiw, N., Romer, D. and D. Weil, 1992, ‘A Contribution to the Empirics of Economic
Growth’, Quarterly Journal of Economics, Vol.107, No.2, pp.407–37.
Martin, S. and D. Parker, 1997, The Impact of Privatisation: Ownership and Corporate
Performance in the UK, London: Routledge.
Megginson, W., Nash, R., and M. Van Randenborgh, 1994, ‘The Financial and Operating
Performance of Newly Privatised Firms: An International Empirical Analysis’, Journal of
Finance, Vol.49, No.2, pp.403–52.
Megginson, W and J. Netter, 2001, ‘From State to Market: A Survey of Empirical Studies on
Privatisation’, Journal of Economic Literature, Vol.39, No.2, pp.321–89.
Parker, D., 1999, ‘Privatisation in the European Union: A Critical Assessment of its
Development, Rationale and Consequences’, Economic and Industrial Democracy, Vol.20,
pp.9–38.
Plane, P., 1997, ‘Privatisation and Economic Growth: An Empirical Investigation from a Sample
of Developing Market Economies’, Applied Economics, Vol.29, No.2, pp.161–78.
Ram, R and T. Schulz, 1979, ‘Life Span, Health, Savings, and Productivity’, Economics
Development and Cultural Change, Vol.27, No.3, pp.399–421.
Rousseeuw, P. and A. Leroy, 1987, Robust Regression and Outlier Detection, New York: John
Wiley.
Sala-i-Martin, X., 1994, ‘Cross-Sectional Regressions and the Empirics of Economic Growth’,
European Economic Review, Vol.38, No.3/4, pp.739–47.
Sala-i-Martin, X., 1997, ‘I Just Run Two Million Regressions’, American Economic Review,
Vol.87, No.2, pp.176–83.
Stokey, N., 1994, ‘Comments on Barro and Lee’, Carnegie-Rochester Conference Series on
Public Policy, Vol.40, pp.47–57.
Temple, J., 1999, ‘The New Growth Evidence’, Journal of Economic Literature, Vol.XXXVII,
pp.112–56.
Tullock, G., 1965, The Politics of Bureaucracy, Washington, DC: Public Affairs Press.
UNCTAD, 1996, Comparative Experiences with Privatisation: Policy Insights and Lessons
Learned, New York: United Nations.
396jds05.qxd 29/08/03 14:13 Page 147
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 147
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
Vickers, J. and G. Yarrow, 1988, Privatisation: An Economic Analysis, London: MIT Press
Yarrow, G., 1999, ‘Theory of Privatisation, or Why Bureaucrats are still in Business’, World
Development, Vol.27, No.1, pp.157–68.
Welsch, R., 1982, ‘Influence Functions and Regression Diagnostics’, in R. Launer and A. Siegel
(eds.), Modern Data Analysis, New York: Academic Press, pp.149–69.
Welsch, R. and E. Kuh, 1977, ‘Technical Report’, in Linear Regression Diagnostics, 923-77,
Cambridge, MA: Sloan School of Management, MIT.
World Bank, 1994, World Bank Assistance to Privatization in Developing Countries, OED
Report, No.3273, Aug.
World Bank, 1995a, Bureaucrats in Business: The Economics and Politics of Reform, Oxford:
Oxford University Press, for World Bank.
World Bank, 1995b, Structural and Sectoral Adjustment Lending: World Bank Experience
(1980–1992), Operations Evaluation Department, Washington, DC: World Bank.
APPENDIX 1
However, the extreme values, max β* (θ) and min β* (θ), may be located in an unlikely
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
parameter space on the locus of constrained estimates in relation to the data. To resolve this
problem, define a statistically feasible likelihood ellipse, say at 0.05 level or 95 per cent, which
is depicted in the dashed lines in the figure, within the likelihood ellipse. This is defined by the
assumption that all doubtful variables are included in the regression. The intersection of the
points in the interior of the ellipse of constrained estimates and this likelihood ellipse is the set of
points (γ1*, γ2*). The area defined by the set is shown by the shaded area in the figure.
Thus, the set of all posterior means for the distribution of (γ1*, γ2*) is here subject to further
constraint in this area, eliminating the unlikely parameter space. The parameter pairs in the area
can be obtained by posterior estimates from prior distribution centred at the origin, and the
measure of specification uncertainty is the difference between the extreme values of β* (θ) over
the area.
FIGURE A1
ELLIPSOIDS OF CONSTRAINED LEAST SQUARES POINTS
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008 396jds05.qxd
APPENDIX 2
PRIV GYP LGYP LIFE POP CON STAB OPEN GDI BUD FDI IMP SDIMP INFL SDINF M3 SDM3 DEBT
Argentina 0.10 0.02 8.58 71.01 0.01 0.04 6.00 0.16 0.17 –0.01 0.01 0.22 0.24 5.65 10.78 0.18 0.05 0.04
Bangladesh 0.00 0.03 7.17 53.30 0.02 0.04 1.00 0.23 0.19 –0.06 0.00 0.80 0.46 0.05 0.03 0.26 0.03 0.02
29/08/03
Barbados 0.03 0.00 8.91 74.34 0.00 0.20 0.00 0.98 0.15 –0.02 0.01 0.08 0.07 0.02 0.02 0.62 0.06 0.07
Belize 0.07 0.03 8.09 72.57 0.03 0.15 0.00 1.16 0.26 –0.03 0.03 0.15 0.10 0.03 0.02 0.43 0.03 0.05
Benin 0.01 0.01 6.90 50.85 0.03 0.11 1.00 0.58 0.16 –0.07 0.00 0.03 0.01 0.07 0.11 0.26 0.03 0.02
Bolivia 0.05 0.02 7.42 56.81 0.02 0.13 3.00 0.48 0.15 –0.03 0.03 0.02 0.02 0.12 0.04 0.35 0.11 0.06
14:13
Brazil 0.04 0.01 8.34 64.76 0.02 0.17 4.00 0.17 0.21 –0.08 0.01 0.29 0.32 10.21 9.53 0.44 0.18 0.03
Burkina Faso 0.00 0.01 6.26 45.87 0.02 0.14 6.00 0.38 0.22 0.00 0.00 0.03 0.01 0.05 0.09 0.23 0.01 0.02
Burundi 0.00 –0.03 6.33 46.71 0.02 0.12 6.00 0.35 0.13 –0.06 0.00 0.27 0.13 0.10 0.07 0.19 0.02 0.04
Cameroon 0.00 –0.04 7.19 53.31 0.03 0.11 2.00 0.40 0.16 –0.03 0.00 0.03 0.01 0.04 0.06 0.19 0.04 0.05
Chile 0.01 0.06 8.29 72.67 0.02 0.10 0.00 0.61 0.25 0.03 0.04 0.13 0.06 0.13 0.07 0.41 0.02 0.08
Colombia 0.03 0.02 8.08 68.23 0.02 0.13 3.00 0.34 0.20 –0.01 0.02 0.09 0.05 0.24 0.03 0.32 0.07 0.08
Page 149
Costa Rica 0.00 0.02 8.12 74.84 0.02 0.16 0.00 0.82 0.26 –0.01 0.03 0.02 0.07 0.18 0.05 0.41 0.03 0.08
Cote d’Ivoire 0.02 –0.01 7.26 50.96 0.03 0.15 1.00 0.68 0.11 –0.14 0.01 0.03 0.01 0.05 0.13 0.28 0.01 0.13
Ecuador 0.00 0.01 7.95 67.04 0.02 0.10 2.00 0.57 0.20 –0.02 0.02 0.13 0.10 0.42 0.16 0.27 0.06 0.09
Egypt, Arab Rep. 0.02 0.02 7.55 61.07 0.02 0.11 2.00 0.53 0.22 –0.04 0.02 0.07 0.19 0.13 0.05 0.87 0.03 0.06
Ghana 0.04 0.02 6.70 55.96 0.03 0.11 0.00 0.50 0.18 –0.04 0.01 0.09 0.10 0.29 0.10 0.17 0.02 0.07
Guatemala 0.00 0.01 7.66 59.69 0.03 0.06 12.00 0.42 0.15 –0.01 0.01 0.10 0.09 0.16 0.11 0.25 0.02 0.03
Guinea 0.00 0.02 6.65 42.55 0.03 0.10 0.00 0.49 0.19 –0.07 0.00 0.07 0.05 0.12 0.11 0.07 0.04 0.04
Guinea-Bissau 0.00 0.02 6.48 41.56 0.02 0.08 1.00 0.51 0.31 –0.26 0.00 0.45 0.20 0.54 0.24 0.16 0.02 0.05
Honduras 0.01 0.01 7.26 65.40 0.03 0.13 3.00 0.75 0.28 –0.06 0.02 0.26 0.30 0.18 0.08 0.33 0.03 0.11
India 0.01 0.04 7.09 57.85 0.02 0.11 11.00 0.21 0.23 –0.06 0.00 0.10 0.04 0.09 0.03 0.48 0.02 0.03
Indonesia 0.01 0.06 7.46 60.21 0.02 0.08 1.00 0.51 0.30 0.00 0.01 0.07 0.08 0.09 0.02 0.43 0.08 0.09
Iran, Islamic Rep. 0.00 0.02 8.08 64.73 0.02 0.12 0.00 0.39 0.28 –0.02 0.00 2.29 1.19 0.27 0.09 0.53 0.09 0.04
Jamaica 0.07 0.01 7.80 72.41 0.01 0.14 2.00 1.19 0.31 –0.03 0.03 0.18 0.09 0.28 0.16 0.53 0.03 0.18
Jordan 0.00 –0.02 8.16 66.71 0.04 0.25 0.00 1.32 0.31 –0.02 0.00 0.05 0.03 0.06 0.06 1.18 0.15 0.13
Kenya 0.01 0.00 6.80 57.55 0.03 0.16 3.00 0.62 0.21 –0.05 0.00 0.14 0.16 0.13 0.07 0.47 0.05 0.10
Korea, South 0.01 0.07 8.63 69.27 0.01 0.10 0.00 0.64 0.36 –0.01 0.00 0.01 0.05 0.06 0.02 0.67 0.14 0.03
Malawi 0.00 0.01 6.20 44.88 0.03 0.16 0.00 0.60 0.19 –0.07 0.00 0.29 0.16 0.29 0.24 0.23 0.04 0.06
Malaysia 0.08 0.06 8.36 69.50 0.03 0.13 1.00 1.65 0.36 –0.03 0.06 0.01 0.01 0.04 0.01 1.05 0.27 0.10
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008 396jds05.qxd
A P P E N D I X 2 (cont.)
PRIV GYP LGYP LIFE POP CON STAB OPEN GDI BUD FDI IMP SDIMP INFL SDINF M3 SDM3 DEBT
Mali 0.00 0.01 6.25 46.51 0.03 0.13 2.00 0.54 0.20 –0.05 0.00 0.03 0.01 0.06 0.10 0.22 0.01 0.03
Mauritania 0.00 0.01 6.67 49.46 0.03 0.13 0.00 1.02 0.20 –0.02 0.01 0.70 0.57 0.06 0.03 0.23 0.05 0.12
29/08/03
Mexico 0.05 0.01 8.58 69.82 0.02 0.10 0.00 0.44 0.23 0.00 0.02 0.05 0.05 0.31 0.29 0.25 0.07 0.08
Morocco 0.02 0.01 7.66 62.01 0.02 0.17 2.00 0.57 0.22 –0.04 0.01 0.05 0.04 0.04 0.02 0.66 0.08 0.10
Mozambique 0.01 0.03 6.63 43.47 0.02 0.13 6.00 0.52 0.17 –0.27 0.01 0.25 0.20 0.43 0.14 0.33 0.05 0.06
Nepal 0.00 0.03 6.91 52.02 0.03 0.09 1.00 0.46 0.22 –0.08 0.00 0.30 0.11 0.10 0.04 0.35 0.03 0.02
14:13
Nicaragua 0.02 –0.01 7.27 62.17 0.03 0.22 7.00 0.80 0.23 –0.15 0.02 0.41 0.63 31.11 45.44 0.36 0.16 0.13
Nigeria 0.01 0.02 6.85 48.31 0.03 0.13 3.00 0.74 0.18 –0.10 0.04 0.64 0.50 0.36 0.24 0.20 0.05 0.09
Oman 0.00 0.01 8.89 67.96 0.04 0.33 0.00 0.86 0.16 –0.09 0.01 0.02 0.01 0.00 0.09 0.31 0.02 0.07
Pakistan 0.01 0.02 7.22 58.24 0.03 0.14 6.00 0.36 0.16 –0.08 0.01 0.06 0.03 0.10 0.02 0.44 0.03 0.05
Panama 0.06 0.01 7.94 71.74 0.02 0.17 11.00 0.73 0.19 0.02 0.02 0.00 0.00 0.02 0.02 0.52 0.13 0.07
Papua New Guinea 0.05 0.01 7.39 53.93 0.02 0.20 9.00 0.97 0.25 –0.03 0.04 0.10 0.06 0.05 0.06 0.34 0.03 0.14
Page 150
Paraguay 0.00 0.01 7.61 67.60 0.03 0.08 0.00 0.82 0.23 0.01 0.02 0.23 0.25 0.20 0.09 0.26 0.05 0.05
Peru 0.09 0.00 7.91 64.39 0.02 0.08 12.00 0.26 0.21 –0.02 0.02 0.36 0.49 10.13 20.11 0.19 0.04 0.03
Philippines 0.02 0.01 7.42 64.00 0.02 0.11 16.00 0.72 0.23 –0.04 0.02 0.05 0.02 0.09 0.03 0.46 0.12 0.07
Senegal 0.02 0.00 7.07 48.25 0.03 0.14 0.00 0.62 0.15 –0.01 0.01 0.03 0.01 0.04 0.24 0.23 0.01 0.06
Sierra Leone 0.00 –0.05 6.79 36.81 0.02 0.10 1.00 0.41 0.07 –0.06 0.00 0.88 0.93 0.48 0.27 0.13 0.03 0.06
Singapore 0.08 0.07 9.24 73.59 0.02 0.09 0.00 3.68 0.35 0.12 0.10 0.02 0.01 0.04 0.02 1.18 0.04 0.11
South Africa 0.02 –0.01 8.12 60.43 0.02 0.20 1.00 0.48 0.17 –0.05 0.01 0.04 0.02 0.12 0.03 0.46 0.02 0.03
Sri Lanka 0.02 0.04 7.61 70.70 0.01 0.10 12.00 0.73 0.24 –0.10 0.01 0.13 0.10 0.11 0.03 0.40 0.02 0.04
Tanzania 0.01 0.00 6.28 51.01 0.03 0.17 0.00 0.51 0.22 –0.09 0.01 0.30 0.23 0.26 0.08 0.24 0.03 0.04
Thailand 0.00 0.07 8.00 67.49 0.01 0.10 5.00 0.80 0.39 0.02 0.02 0.03 0.01 0.05 0.01 0.79 0.07 0.06
Togo 0.01 0.00 6.46 51.41 0.03 0.13 8.00 0.75 0.16 –0.05 0.00 0.03 0.01 0.07 0.11 0.32 0.05 0.04
Trinidad and 0.06 0.00 9.00 70.39 0.01 0.13 2.00 0.83 0.16 –0.02 0.05 0.23 0.18 0.07 0.07 0.53 0.05 0.10
Tobago
Tunisia 0.00 0.02 7.90 65.63 0.02 0.16 0.00 0.89 0.26 –0.04 0.02 0.05 0.05 0.06 0.02 0.50 0.02 0.10
Turkey 0.01 0.02 8.14 64.25 0.02 0.11 5.00 0.39 0.24 –0.05 0.00 0.03 0.04 0.75 0.15 0.29 0.04 0.06
Uganda 0.01 0.04 6.27 48.34 0.03 0.09 2.00 0.30 0.14 –0.07 0.01 0.38 0.40 0.48 0.60 0.10 0.01 0.04
Uruguay 0.00 0.02 8.45 72.11 0.01 0.13 2.00 0.43 0.13 –0.03 0.00 0.12 0.08 0.58 0.28 0.49 0.09 0.06
Venezuela 0.04 0.00 8.82 70.53 0.02 0.08 8.00 0.54 0.18 –0.02 0.02 0.19 0.36 0.50 0.31 0.35 0.08 0.08
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008 396jds05.qxd
A P P E N D I X 2 (cont.)
PRIV GYP LGYP LIFE POP CON STAB OPEN GDI BUD FDI IMP SDIMP INFL SDINF M3 SDM3 DEBT
Vietnam 0.00 0.05 6.79 65.80 0.02 0.08 0.00 0.69 0.21 0.00 0.04 0.46 0.73 0.70 1.23 0.22 0.01 0.03
Yemen 0.00 0.00 7.54 51.49 0.04 0.19 0.00 0.61 0.20 –0.08 0.04 0.17 0.07 0.25 0.16 0.51 0.07 0.51
29/08/03
Zambia 0.05 –0.01 6.63 49.63 0.03 0.17 1.00 0.72 0.13 –0.11 0.02 0.69 0.68 0.77 0.50 0.22 0.07 0.17
Zimbabwe 0.01 0.01 6.99 56.91 0.03 0.19 0.00 0.62 0.21 –0.09 0.00 0.29 0.16 0.21 0.06 0.40 0.05 0.08
Mean 0.02 0.02 7.51 59.83 0.02 0.13 3.06 0.67 0.21 –0.05 0.02 0.22 0.18 1.09 1.48 0.39 0.06 0.08
14:13
Standard Deviation 0.03 0.02 0.80 9.95 0.01 0.05 3.88 0.47 0.06 0.06 0.02 0.33 0.24 4.28 6.40 0.23 0.05 0.07
Notes: PRIV = the magnitude of privatisation as the ratio of cumulative proceeds from privatisation during the sample period 1988-97 to average GDP
during the same period; primary data for the proceeds are taken from World Bank Privatization Database and Privatisation International
(various issues); and the data for GDP are taken from World Bank Global Development Network Growth Database.
GYP = the average real GDP per capita (in constant US dollars, international prices, base year 1985) growth rate during the sample period; and
Page 151
the data are taken from World Bank Global Development Network Growth Database.
LGYP = the log of initial real GDP per capita in 1988 capita (in constant US dollars, international prices, base year 1985); and primary data are
taken from World Bank Global Development Network Growth Database
POP = the average population growth rate during the sample period; and the data are taken from World Bank Global Development Network
Growth Database
LIFE = the initial life expectancy at birth in 1987 (the data in 1988 are not available); and the primary data are taken from World Bank Global
Development Network Growth Database
STAB = political instability calculated as the average number of major political crises, revolutions, and coups during the sample period; and the
primary data are taken from World Bank Global Development Network Growth Database
BUD = the average ratio of central government budget surplus/deficit to GDP (calculated in national currency, constant price) during the sample
period; and the primary data are taken from World Bank Global Development Network Growth Database
OPEN = openness of a country calculated as (export-import)/GDP; and the primary data are taken from World Bank Global Development Network
Growth Database
CON = the average ratio of government consumption to GDP (calculated in national currency, constant price) during the sample period; and the
primary data are taken from World Bank World Development Indicators
GDI = the average ratio of gross domestic investment to GDP (calculated in national currency, constant price) during the sample period; and the
primary data are taken from World Bank World Global Development Network Growth Database
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008 396jds05.qxd
A P P E N D I X 2 (cont.)
152
FDI = the average ratio of foreign direct investment (net inflows) to GDP (calculated in current US dollars) during the sample period; and the
primary data are taken from World Bank World Development Indicators
INFL = the average inflation rates (GDP deflator) during the sample period; and the primary data are taken from World Bank World Development
Indicators
29/08/03
* World Bank Global Development Network Growth Database is available from www.worldbank.org/research/growth/GDNdata.htm
THE J OURNA L O F D E V E L O P ME N T S T U D I E S
396jds05.qxd 29/08/03 14:14 Page 153
P R I VAT I S AT I O N A N D E C O N O M I C G R O W T H 153
Downloaded By: [Hiroshima University] At: 07:16 25 July 2008
APPENDIX 3
WDi = DFITSi n −1 ,
1 − hi
Then, outliers or the observations that ought to be examined further are defined as:
DFITS > 2 , k n , Cook’s Distance > 4/n, and Welsch Distance > 3 k .
396jds05.qxd 29/08/03 14:14 Page 154
∧ ∧
DFBETAs measure the effect on each coefficient by testing β j – β (i)j , which is defined as:
βˆ j − βˆ (i ) j ,
DFBETAsij =
s ( i ) ( X ' X ) −jj1
where is the (X′ X)–1jj is the jth diagonal element of (X′ X)–1. The observations that ought to be
examined further are defined as DFBETAs > 2 / n .