On the Rate of Growth of the Population of the United States since 1790 and Its Mathematical Representation

Raymond Pearl, Lowell J. ReedProceedings of the National Academy of Sciences 1920 · DOI

This 1920 paper by Raymond Pearl and Lowell J. Reed is a landmark in demography and mathematical biology. It moves beyond simple curve fitting to propose a biological "law" of population growth. By applying a logistic curve—similar to those found in autocatalytic chemical reactions—the authors argue that U.S. population growth follows a predictable S-shaped trajectory toward a finite carrying capacity.

TL;DR

Pearl and Reed demonstrate that U.S. population growth is better represented by a logistic S-curve than by traditional linear or geometric models. Their model accurately predicted historical census data and suggested that the U.S. passed its point of maximum growth rate around 1914, heading toward an ultimate limit of approximately 197 million people.

What Problem Does This Paper Address?

In the early 20th century, estimating population between census years (inter-censal years) relied on crude arithmetic or geometric progressions. These methods assumed constant growth rates, which the authors argue are biologically impossible over long periods.

The core challenge was finding a mathematical representation that wasn't just an empirical "best fit" for past data, but one that reflected the underlying organic laws of growth. Specifically, the authors sought a model that accounted for the fact that population growth must eventually slow down as a country reaches the limits of its resources (subsistence).

Methods

The authors compare two mathematical approaches:

  1. Empirical Curve Fitting: They first use a logarithmic parabola ($y = a + bx + cx^2 + d \log x$). While highly accurate for interpolation, they admit it lacks a theoretical biological basis because it can grow infinitely.
  2. The Logistic Law (The "S-Curve"): They propose a theoretical model based on the equation: $$y = \frac{b e^{ax}}{1 + c e^{ax}}$$ This function is asymptotic at both ends (0 and a maximum $k$) and contains a point of inflection where growth shifts from accelerating to decelerating. This mirrors the "autocatalytic" nature of growth, where the rate depends on both the current population and the remaining "potential" for growth.

To fit the logistic curve, they used a "procrustean" method, forcing the curve through three specific census points (1790, 1850, and 1910) to determine the constants.

Datasets

The study utilizes official U.S. Census Bureau data from 1790 to 1910. The authors highlight that these figures include adjustments and corrections made by the Bureau since the original counts.

Table 1: U.S. Census Data 1790-1910

Key Insights and Ideas

  • Biological Analogy: The authors argue that population growth in a limited area is fundamentally a growth phenomenon, much like the growth of an individual organism or a population of yeast.
  • The Inflection Point: The model identifies a critical "turning point" in U.S. history. They calculate that the U.S. passed its period of most rapid population growth around April 1, 1914.
  • Carrying Capacity: By calculating the upper asymptote of their logistic curve, they estimated the maximum population the U.S. could support (given 1920s agricultural technology and food habits) to be roughly 197.3 million.
  • Predictive Power: To test the model's validity, they "predicted" the 1910 population using only data available up to 1860. The result was only 0.6% in error—a remarkable validation of the curve's extrapolative stability.

Key Results

The authors found that their logarithmic parabola provided a significantly better fit than previous models, such as Pritchett’s third-order parabola.

Table 2: Comparison of Errors between Pritchett and the Logarithmic Parabola

As shown in Figure 1, the empirical curve tracks the observed census data with extreme precision:

Figure 1: Observed vs. Calculated Populations (Logarithmic Parabola)

While the logistic curve (Equation xviii) had a slightly higher root-mean-square error than the empirical parabola, the authors favored it for its theoretical soundness. It suggested a future U.S. population density of about 66 persons per square mile, which they defended as a realistic limit based on caloric production requirements.

Limitations and Future Work

The authors were candid about the model's "first approximation" status:

  • Symmetry: The logistic model used is perfectly symmetrical, meaning the deceleration phase is a mirror image of the acceleration phase. The authors noted this is likely too rigid for complex human populations.
  • Static Assumptions: The model assumes that "standard of living" and "agricultural production methods" remain constant parameters. Significant technological shifts (like the Green Revolution) would effectively shift the asymptote.
  • Generalization: The paper concludes by stating the need to generalize the logistic equation to allow for asymmetry and more flexible inflection points, a task the authors were actively pursuing at the time of publication.

Summary updated

BibTeX
@inproceedings{pearl1920rate,
  author = {Raymond Pearl and Lowell J. Reed},
  title = {{On the Rate of Growth of the Population of the United States since 1790 and Its Mathematical Representation}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {1920},
  doi = {10.1073/pnas.6.6.275},
}
Figure