Gerhard Tintner; Jati K. Sengupta · 1963
Gerhard Tintner and Jati K. Sengupta’s journal article develops a stochastic explanation of economic growth whose expected trajectory is logistic, then estimates it using German per-capita national-income data for 1851–1939. Its argument moves from a critique of deterministic growth theories through a probability model and an estimation procedure to applications using nominal and real income. The central contribution is to connect bounded long-run growth with probabilistic fluctuations, rather than treating a fitted trend as a complete description of economic development.
The opening distinguishes the requirements of short-term forecasting from those of long-term development theory:
Es besteht kein Einwand gegen die Verwendung eines exponentiellen Trends für kurzfristige Voraussagen, aber für die Theorie der wirtschaftlichen Entwicklung scheint ein logistischer Trend besser geeignet zu sein, da es sich immer um langfristige Voraussagen handelt$^{4)}$.
English translation: There is no objection to using an exponential trend for short-term forecasts, but a logistic trend seems better suited to the theory of economic development, since it always concerns long-term forecasts$^{4)}$.
The qualification matters: exponential growth is not rejected universally, and the superiority of the logistic curve is proposed rather than established empirically. Against predominantly deterministic theories yielding linear or exponential trends, the authors cite Haavelmo as an exception and seek a simple stochastic process compatible with a logistic mean. Their generalized birth-and-death framework describes changes in an economic quantity (x), such as national income, through the evolution of its probability density (p(x,t)). A differential–integral equation combines a local derivative, a term involving the current density, and an integral over values above (x). The accompanying interpretation associates these terms with decreases, unchanged values, and increases during a small time interval. Setting one parameter to zero gives the familiar linear birth process.
The decisive economic assumption concerns the distribution of income at a given time. The authors posit a negative exponential density, (p(x,t)=g(t)e^{-xg(t)}), and motivate it through short-run production constraints:
Die Gesamtproduktion oder das Volkseinkommen wird dann durch den Faktor (oder die Faktoren) bestimmt, die am knappsten sind.
English translation: Total production or national income is then determined by the factor (or factors) that are scarcest.
This bottleneck argument connects linear programming with extreme-value statistics. If the most scarce productive input governs output, the relevant statistical problem is the distribution of a sample minimum, rather than an average of independently contributing factors. The authors invoke convergence of minimum-value distributions toward an exponential distribution for large samples. This supplies their economic rationale for the chosen density, although the article does not develop a detailed empirical model of the production factors themselves.
Under this distribution, expected income is (1/g(t)), while its variance is (1/g(t)^2). Consequently, the model specifies dispersion as well as a trend: the standard deviation equals the mean. The further assumption (g(t)=a+be^{-ct}), with positive constants, produces the logistic expectation [ E[x]=\frac{1}{a+be^{-ct}}=\frac{k}{1+Be^{-ct}}, \qquad k=1/a,\quad B=b/a. ] The ceiling (k) is therefore the reciprocal of the limiting value of the density’s time-dependent parameter. The authors verify that this density satisfies their evolution equation and give a generating function from which moments and semi-invariants can be calculated. Their construction demonstrates compatibility between the proposed stochastic law and logistic growth; it does not derive logistic growth without distributional and functional assumptions.
The estimation section makes the model operational by taking the reciprocal of the logistic curve. This converts a nonlinear trajectory into a first-order linear difference equation with constant coefficients, allowing a familiar regression technique to estimate the ceiling and growth parameter:
Wir können die Koeffizienten $(1 - e^{-c})/k$ und $e^{-c}$ mittels der klassischen Methode der kleinsten Quadrate schätzen.
English translation: We can estimate the coefficients $(1 - e^{-c})/k$ and $e^{-c}$ using the classical method of least squares.
Estimates of those coefficients yield (c) and (k); a separate method attributed to E. C. Rhodes supplies (B). This reciprocal transformation is an important conceptual bridge between the probabilistic construction and observed time series. The article estimates the expected trajectory first, then uses its parameters to specify the stochastic equation numerically.
For nominal German per-capita national income, measured in thousands of Reichsmarks, the fitted logistic curve has an asymptote of approximately 1.547269 and an exponential coefficient of 0.0192. The authors translate these estimates back into the parameters of their density and evolution equation. They then repeat the procedure for real income:
Die Daten sind Fünfjahresdurchschnitte zu Preisen von 1913.
English translation: The data are five-year averages at 1913 prices.
The real-income application yields a reported logistic asymptote of approximately 3.196445 and an exponential coefficient of 0.076427. Because these observations are five-year averages, the coefficient should not be read straightforwardly as an annual growth rate comparable to the nominal estimate. In both applications, the final step is to interpret the fitted equation’s terms as probabilities of income falling, remaining unchanged, or rising over a small interval.
The article’s relevance lies in this integration of stochastic dynamics, scarcity-based distributional reasoning, and tractable trend estimation. Its scope is methodological rather than a historical explanation of Germany’s successive economic transformations: wars, institutional changes, and structural breaks are not separately modeled. Nor does it report goodness-of-fit statistics, uncertainty estimates, or tests against competing trends. The printed numerical exposition also contains inconsistencies, so its probability coefficients warrant checking before reuse. The strongest conclusion is therefore a constructive one: a logistic income trend can be embedded in a specified stochastic process and calibrated to historical data. The empirical examples illustrate that procedure without independently establishing either a fixed long-run income ceiling or the proposed distribution of fluctuations.
This work was divided into 2 sections when it entered the library's research corpus—an apparatus for search and citation, not necessarily the author's own table of contents. Each title opens its summary.
Put a question to this work; the Librarian answers from its 2 sections and cites the passage.
Ask the Librarian