GrowthMP {WALS} | R Documentation |
Growth regression data used in Masanjala and Papageorgiou (2008).
GrowthMP
A data frame with 37 observations on 25 variables:
Average growth rate of GDP per capita from 1960 - 1992 at purchasing power parity.
Logarithm of GDP per capita in 1960.
Fraction of years economy open from 1960 - 1990.
Fraction of GDP in mining.
Share of exports of primary products in GDP in 1970.
Ratio of real domestic investment (public and private) to real GDP.
Real exchange rate distortion.
Average years of primary schooling for population over 25 years of age in 1960.
Life expectancy at age 0 in 1960.
Average growth rate of population from 1960 - 1990.
factor. "yes"
if country participates in at least one external
war from 1960 to 1985. "no"
else.
Average number of revolutions and coups per year from 1960 - 1990.
Index of political rights ranging from 1 (most restrictive) to 7 (most freedom)
Index of civil liberties ranging from 1 (most restrictive) to 7 (most freedom)
Index of outward orientation.
Degree of capitalism.
factor. Shows if the country used to be "british"
or
"french"
colony. If neither of them applies, then "none"
.
Fraction of English speakers.
Fraction speaking foreign language.
Probability that two random people are from different ethnolinguistic groups.
Fraction of population Protestant.
Fraction of population Catholic.
Fraction of population Muslim.
Size of country in millions of square kilometers.
Distance from the equator.
The dataset of Masanjala and Papageorgiou (2008) is a subset of sub-Sahara African countries from the data used in Sala-I-Martin (1997). See Table A2. in Masanjala and Papageorgiou (2008) for the original sources of the variables. This dataset is also used for replication purposes in Amini and Parmeter (2012).
To replicate the WALS estimates in Amini and Parmeter (2012),
use all variables except for a constant as auxiliary regressors and divide all
regressors by their in-sample maximum before running
wals(..., prescale = FALSE)
(NOTE: It is not recommended to use
prescale = FALSE
as this runs an old version of the WALS estimator,
prescale = FALSE
should only be used for replication purposes).
The resulting coefficients and standard errors have to be divided by the maximum
of the regressors again to get the values presented in Table I of the paper.
Journal of Applied Econometrics Data Archive. The data was taken from the archive entry of Amini and Parmeter (2012) for replication purposes but they can also be found in the archive entry of Masanjala and Papageorgiou (2008).
Amini SM, Parmeter CF (2012).
“Comparison of model averaging techniques: assessing growth determinants.”
Journal of Applied Econometrics, 27(5), 870-876.
doi:10.1002/jae.2288.
Masanjala WH, Papageorgiou C (2008).
“Rough and lonely road to prosperity: a reexamination of the sources of growth in Africa using Bayesian model averaging.”
Journal of Applied Econometrics, 23(5), 671-682.
doi:10.1002/jae.1020.
Sala-I-Martin X (1997).
“I Just Ran Two Million Regressions.”
The American Economic Review, 87(2), 178–183.
## Replicate second panel of Table I in Amini & Parmeter (2012)
## NOTE: Authors manually scale data, then rescale the resulting coefs and se.
X <- model.matrix(gdpgrowth ~ ., data = GrowthMP)
scaleVector <- apply(X, MARGIN = 2, max)
Xscaled <- apply(X, MARGIN = 2, function(x) x/max(x))
Xscaled <- Xscaled[,-1]
datscaled <- as.data.frame(cbind(gdpgrowth = GrowthMP$gdpgrowth, Xscaled))
fitMP <- wals(gdpgrowth ~ 1 | ., data = datscaled, prescale = FALSE,
prior = laplace(), eigenSVD = FALSE)
tableMP <- cbind("coef" = coef(fitMP)/scaleVector,
"se" = sqrt(diag(vcov(fitMP)))/scaleVector)
printVars <- c("(Intercept)", "lgdp60", "yrsopen", "mining", "primexp70",
"invest")
print(round(tableMP[printVars,], 4))