Thu, 23 Apr 1998 10:15:16 +0200 (MET DST) Thu, 23 Apr 1998 10:15:08 +0200 (MET DST) Reply-To: From: "Austrian Embassy" To: Subject: Fw: gernot's interesting comment Date: Thu, 23 Apr 1998 10:20:50 +0200 here is Gernot's reaction to my suggestions on his paper. Even if you do not know German, you will find his interesting literature hint at your disposal. Gernot's problem is real, and it is charged with rather explosive social consequences. I wonder why there has been such a more than 3 day silence silence on his brilliant paper, considering all the issues of exchange rate over-or undervaluations in the wake of the Asian crash in the fall of 1997. I agree with Gernot on the important nature of the problem, but I disagree with him on the research strategy. I did meanwhile therefore a non-linear regression analysis on the interesting relationship between PPP and GDP per capita. This analysis somewhat qualifies Gernot's thesis. The relationship - the best fit is a polynomial regression of the second order - looks like the lower, right hand quarter of an elipse, with PPP being the x-axis and GDP per cap the y axis. When you use a logarithmic sclaing for PPP, then the form of curve transforms itself into the right-hand part of an "U". Simply calculating the division between PPP and GDP or GNP is not a good research strategy, I find. You would have to control for this eliptic trade-off and you would have to calculate the RESIDUALS FROM THE NON LINEAR REGRESSION BETWEEN PPP PER CAP AND GNP PER CAP to arrive at more meaninful results. Seen in such a way, the deviations are not so dramatic, and rather, and we find the structural shortcomings of Kravis' method which in the end determine this eliptic trade-off. If you do not consider this eliptic trade off, you are the victim of Kravis peculiar PPP calculation method - and not necessarily confronted with a real, world economic phenomenon. Sorry, that the wsn machine in Colorado again and again rejected my earlier color graphs in the past, sent around via the wsn network, but let me attempt to draw a brief Graph here to summarize my thoughts: l GNP per cap * l * l * l * l * l * l* * * * --------------------------------------- PPP per cap The empirical values, based on UNDP HDR data, neatly scatter around this curve, the R^2 is above 90%. Only residuals from this function will deliver to us an unbiased estimate of the "Kohler" or "ERDI" effect. Performing a simple division f (Kohler formulation) = PPP per cap/GNP per cap will be a far inferior (second best and flawed) formulation about this effect. Simple correlations with growth, in addition, will be biased as well. We have to take the Matthews effect into account. To repeat the most important development level development performance trade-offs, let me state here: economic growth/adjustment success = a1 + b1* ln (PCItn-1)-b2* (ln(PCItn-1))2 The same function is also applied to income inequality, following a famous essay published by S. Kuznets in 1955. Redistribution gets underway after 1000 $ per capita income is reached; the share of the richest 20% diminishes from approximately 55% to around 40%. Growth and adjustment accelerate with redistribution. DYN LEX = ((log 10 (LEX tn)-log 10 (LEX tn-1)) * 100 Let LEX denote life expectancy or other basic human needs indicators, PCI per capita incomes, ENCONS p.c. energy consumption rates per capita and year in kg oil equivalent, and DYN rates of increases of basic human needs satisfaction. On a world scale and for different groups of countries, levels of human development and increases in terms of human development, reductions in infant mortality et cetera will always significantly correspond to the following function and the first derivative: LEX = a + b1 * (ENCONS p.c.)^(1/(e^2)) - b2 * (ENCONS p.c.)^ln(pi) R2 = 72.4%; F = 157.63; df. = 120; alpha (one-tailed) 5% > 1.289 DYN LEX(tn) = a - b1 * LEX (tn-1) +- b2 * (PCI)(tn-1)^((1/(e^2))-1)-b3 * (PCI)(tn-1)^(((ln(pi))-1))) R2 = 69.8%; F = 91.85; df. = 120; alpha (one-tailed) 5% > 1.289 predictors b2 and b3 only: R2 = 43.3%; F = 45.89; df. = 120; alpha (one-tailed) 5% > 1.289. Formulation also possible with ENCONS p.c., but the PCI data series is more complete Based on UNDP (1993) data for all the countries that report economic growth rates for the periods 1965-80-90, equation (3.4) explains 72.4% of total variance of life expectancy; equation (3.5) - even without life expectancy in 1960 as an additional control variable - explains 45.9% of total variance. human development or gender development or gender empowerment = a1 - b1* ln (PCItn-1) + b2* (ln(PCItn-1))2 or political rights violations or civil rights violations = a1 + b1* ln (PCItn) - b2* (ln(PCItn))2 The following statistical properties of the functions hold: human development index R^2 = 82.4%; F = 281.0 gender development index R^2 = 80.1%; F = 240.8 gender empowerment index R^2 = 60.0%; F = 90.0 political rights violations R^2 = 38.0%; F = 36.8 civil rights violation R^2 = 40.0%; F = 39.9 Human Development Index = a + b1 * (real purchasing power p.c.)^(1/(e^2)) - b2 * (real purchasing power p.c.)^ln(pi) Kind regards, and thanks to Gernot for his important paper. Arno Tausch ---------- > From: Gernot Kohler > To: austria@it.com.pl > Subject: > Date: Mittwoch, 22. April 1998 16:36 > > Danke vielmals fuer Ihren Kommentar. Ich habe das Wiener Institut (WIIW) > ausfindig gemacht (web) und fand einen abstract of a research report by > Havlik. Er argumentiert allerdings *fuer* currency undervaluation. Nachdem > ich meinen Essay fertiggestellt hatte, habe ich noch ein Buch gefunden, das > Sie interessieren wird, da es die Beziehung zwischen "undervaluation" und > Wirtschaftswachstum untersucht: > Pan A. Yotopoulos > Exchange rate parity for trade and development: > Theory, tests, and case studies > Cambridge University Press, 1996 > ISBN 0 521 48216 > Unter anderem: "causality goes from undervaluation of NER to low rates of > growth of GDP" (p. 143) [NER=nominal exchange rate] und: "intervention in > the foreign exchange market" seen as attractive (p. 286). Zusaetzlich zu > multivariater Analyse hat er vier Fallstudien, unter anderem um > festzustellen/nachzuweisen, dass "government" (gute Regierung) wichtig ist > in diesem Zusammenhang. Passt wahrscheinlich gut zu Ihrem "lean and socially > just government". > > Freundliche Gruesse, > Gernot >