The threat of a malthusian crisis in the Habsburg monarchy

Journal of Interdisciplinary History

Volumn 30, Issue 1, 1999, Pages 91-98

  Komlos, John ^a   Heintel, Markus ^b  

^a UNIVERSITY OF MUNICH   (Germany)

^b Munchener Ruck   (Germany)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0010813054     PISSN: 00221953     EISSN: None     Source Type: Journal    
DOI: 10.1162/002219599551921     Document Type: Article

References (28)

1. Komlos, "Stature and Nutrition in the Habsburg Monarchy: The Standard of Living and Economic Development in the Eighteenth Century," American Historical Review, XC (1985), 1149-1161;
2. Komlos, idem, Nutrition and Economic Development in the Eighteenth Century Habsburg Monarchy: An Anthropometric History (Princeton, 1989);
3. Komlos, idem, "The Secular Trend in the Biological Standard of Living in the United Kingdom," Economic History Review, XLVI (1993), 115-144;
4. Komlos, idem, "Anomalies in Economic History: Reflections on the Antebellum Puzzle," Journal of Economic History, LVI (1996), 202-214;
5. Komlos, idem, "Shrinking in a Growing Economy: The Mystery of Physical Stature during the Industrial Revolution," Journal of Economic History, LVIII (1998), 779-802;
6. Lars Sandberg and Richard H. Steckel, "Heights and Economic History: The Swedish Case," Annals of Human Biology, XIV (1987), 101-110;
7. Jörg Baten, Ernährung und wirtschaftliche Entwicklung in Bayern, 1790-1880 (Stuttgart, 1998).
8. Hans-Joachim Voth, "Height, Nutrition, and Labor: Recasting the 'Austrian Model,'" Journal of Interdisciplinary History, XXV (1995), 627-636;
9. Hans-Joachim Voth, idem, "Physical Exertion and Stature in the Habsburg Monarchy, 1730-1800," Journal of Interdisciplinary History, XXVII (1996), 263-275.
10. The "working-day variable," consisting of only four different values, is obviously a poor proxy of labor input. The variable is artificially constructed by adding the number of abolished holy days to 251.
11. Komlos and Ritschl, "Holy Days, Work Days, and the Standard of Living in the Habsburg Monarchy," Journal of Interdisciplinary History, XXVI (1995), 57-66. In his rejoinder, Voth refers to Komlos' data set (of nearly 150,000 observations) as the "larger" data set. This term is misleading, insofar as Komlos had only one data set, not a larger and a smaller one.
12. The "smaller" data set, to which Voth refers is actually the 42 observations published in Komlos, Nutrition and Economic Development, 57, from which Voth attempted to estimate 17 (!) variables. In addition, he used an incorrect statistical method; weighted least squares would have been the right procedure, not ordinary least squares.
13. Voth, "Physical Exertion and Stature," 264.
14. Voth, "Physical Exertion and Stature."
15. William H. Greene, Econometric Analysis (New York, 1993), 179, 180. The most popular example of partial regressions is detrending.
16. Voth, "Physical Exertion and Stature," 266.
17. To be sure, the raw sample is not truncated perfectly, because some men shorter than the minimum height requirement (165.8 cm) were accepted into the Habsburg military. Provided that one discards all observations below 165.8 cm, OLS is still useful in case of a truncated dependent variable; with truncated OLS, the relative size of the regression coefficients remain valid even if the t-statistics do not remain meaningful. On related issues, see Ching-Fan Cheung and Arthur Goldberger, "Proportional Projections in Limited Dependent Variable Models," Econometrics, LII (1984), 533-534;
18. Heintel, "Historical Height Samples with Short-fall. A Computational Approach," History and Computing, VIII (1996), 24-37;
19. Heintel and Baten, "Smallpox and Nutritional Status in England, 1770-1873: On the Difficulties of Estimating Historical Heights," Economic History Review, LI (1998), 360-371;
20. Greene, Econometric Analysis, 245-247, 689;
21. Komlos and Joo H. Kim, "Estimating Trends in Historical Heights," Historical Methods, XXIII (1990), 116-120.
22. Greene, Econometric Analysis, 689-690. We model the time variable as dummy variable for each birth decade beginning with the 1730s. Furthermore, we confine our analysis to adults over the age of twenty-three in order to simplify the regression models by excluding recruits who have not reached their terminal heights. We also exclude recruits born in the nineteenth century, because the W variable remains unchanged after 1770 (Figure 1). With these restrictions, the data set under analysis reduces to N = 36,500.
23. The log likelihood function of the truncated regression model is given in Greene, Econometric Analysis, equation 22.9, 689. Computational details for the mlf are available from the authors.
24. See also Siddharta Chib, "Marginal Likelihood from the Gibbs Output," Journal of the American Statistical Association, XC (1990), 1313-1321. The marginal likelihood criterion is closely related to the Bayesian Information Criterion (BIC) of Schwarz and superior to Akaike's Information Criterion (AIC).
25. See Gerhard Schwarz, "Estimating the Dimension of a Model," Annals of Statistics, VI (1978), 461-464;
26. H. Akaike, "Information Theory and an Extension of the Maximum Likelihood Principle," in B. N. Petrov and F. Csaki (eds.), Second International Symposium on Information Theory (Budapest, 1973);
27. 2 in an ordinary regression model.
28. Greene, Econometric Analysis, equation 22.9, 689. We also computed the log mlfc with both T and W included in the regression. These results are not reported here, because the projection matrix was singular due to multicollinearity. When computed with pseudo (Moore-Penrose) inverse, the log mlfc was negative. We also repeated the partial regression approach using maximum likelihood instead of OLS. In the second step, we used either W or T as the extra independent variable. In both cases, again the log mlfc was negative, that is, smaller than the fifteen or fifty-seven reported in Table 1. Thus, the partial regression models are inferior to the instrumental (dummy) variable, T. As one might expect, Voth's championing of the partial regression model fails, even if done correctly; in cases of high multicollinearity, the usual solution is the rigorous a priori exclusion of one of the multicollinearity, variables.