Medical progress and health inequality: penicillin in Italy

Grant Miller and Jurre Jochemsen. Show Author details

First Penicillin Italy Medical progress and health inequality penicillin in Italy
First Italian Penicilin

Technological progress in public health and medicine has been described as one of the main drivers behind the general increase in life expectancy since World War II (Acemoglu & Johnson, 2007; Cutler et al., 2006; Easterlin, 1999; Preston, 1975; see also posts at our blog here or here). Although these innovations have improved health on average, however, it is less clear what their implications for health inequality are. For example, expensive new treatments may be adopted first by the wealthy, generally increasing health inequality across socio-economic groups. On the other hand, medical advancements can have the opposite effect if they lead to more affordable new treatments than previously existing substitutes, or if they focus on diseases disproportionately prevalent among the poor. 

Penicillin and mortality in Italy

In our research (see here), we address this by examining the introduction of penicillin after the Second World War. Although a single medical innovation, penicillin has been called the “single greatest victory ever achieved over disease.”(Bennett & Chung, 2001). In the century prior to its introduction (ca. 1947), life expectancy was very low (average life expectancy at birth was only 29 years in 1861). Health disparities were particularly large as well: life expectancy in Italy varied by as much as 12 years across regions. These high levels of health inequality persisted until the end of World War II, after which they rapidly declined. We hypothesize that the introduction of penicillin in 1947 played a central role in Italy’s post-war population health convergence.  

To test this, we gathered data on mortality from the Italian National Statistical office between 1924 and 1955. These consist of individual death certificates for every death in Italy that include the cause of death as well as age at the time of death. 

Our empirical strategy draws on the idea that penicillin is mainly effective against infectious diseases to test whether the decline in mortality after the war is associated with the introduction of penicillin. More specifically, we compare the mortality rate of diseases that can be treated with penicillin (penicillin-sensitive diseases) with the mortality rate of diseases that can not be treated with it (penicillin-insensitive diseases). This comparison can be seen in figure 1 below (1947 marks the introduction of penicillin), which shows little sign of decreasing mortality rates for penicillin-sensitive diseases compared to penicillin-insensitive diseases prior to 1947. As soon as penicillin was introduced, however, mortality rates for penicillin-sensitive diseases dropped sharply when compared with mortality rates for penicillin-insensitive diseases. 

However, to confirm whether penicillin is truly responsible for Italy’s post-war health convergence, a more robust analysis is necessary. We test the influence of penicillin in four different ways: (1) analyzing changes in the distribution of the age at death, (2) analyzing the post-1947 drop in penicillin-sensitive (vs. insensitive) mortality rates for every region, (3) testing if regions with higher pre-1947 mortality rates converged towards regions with lower initial mortality rates, and (4) testing for convergence in the standard deviation of the age at death distribution.

Mortality rates due to various diseases in Italy 1924-1955
Mortality rates due to various diseases in Italy 1924-1955

First, we analyze changes in the distribution of age at death by using a Kolmogorov-Smirnov test. In this test, the distribution of age at death in 1955 is compared to the distribution in earlier years. The results are shown in the figure below. We find that throughout the period 1924-1947, the distribution of the age of death was different from the distribution in 1955. By contrast, we find no statistically significant differences between the distributions during the period 1947-1954 and the baseline distribution of 1955. This result implies that the distribution of age at death changed abruptly around 1947. Because this is the same year in which penicillin was introduced, this is highly consistent with the idea that penicillin led to mortality rate convergence.

Equality of age at death mortality in Italy 1924-1955
Kolmogorov-Smirnov test of the equality of age at death mortality distributions in Italy, 1924–1955.

Secondly, we estimate a statistical model that compares the the change in penicillin-sensitive mortality rates from before to after 1947 with the change in penicillin-insensitive mortality rates over the same period.  Doing so, we find that after 1947, mortality rates of penicillin-sensitive diseases dropped in all regions relative to changes in penicillin-insensitive diseases, and that this change was most significant in regions were the initial rates were high. The results thus imply health convergence.

Thirdly, we estimated a separate model that directly estimates the convergence in mortality rates between regions by pre-1947 rates of penicillin-sensitive mortality.  Our results show that after 1947, penicillin-sensitive mortality rates converged about four times as quickly as before. By contrast, penicillin-insensitive mortality rates diverged after 1947, having neither converged nor diverged before then.

Finally, health convergence was measured by looking at variance in the mortality rates. Figure 3 shows that the variance in penicillin-sensitive mortality rates dropped after the introduction of penicillin, and that this decline in  variance is statistically significant. 

Standard deviation of mortality rates in Italy 1924-1955

Ruling out confounding factors

Can we conclude confidently that penicillin was behind the drop in health inequality after 1947? Or are there other factors that can explain this development? We identify three potential concerns with our study. The first relates to competing risks: we are comparing penicillin-sensitive mortality rates with penicillin-insensitive mortality rates, but these two factors can influence each other. In particular, a person that was saved by penicillin can later end up dying of penicillin-insensitive causes. The decline in penicillin-sensitive deaths can therefore increase the penicillin-insensitive deaths, biasing the comparison. Our analyses however showed that in the context of this study, the impact of the competing-risks problem is small.

The second factor we consider is the influence of World War II on health patterns. Mortality rates of infectious diseases often rise during wartime, which means that their decline after 1945 could be attributed to the end of the war instead of the introduction of penicillin. This does not seem to be the case, since our results are very similar when we exclude from our analysis the period 1943-1945, the years of greatest conflict in Italy. Therefore, the potential independent effects of the war on regional mortality do not impact our conclusions.

Lastly, our results could be explained by selective migration linked to people of a certain socioeconomic and/or health status. This is unlikely to have influenced our analyses as postwar migration rates in Italy were the lowest ever recorded and some evidence suggests that it was unrelated to health and/or socioeconomic status.

Medical progress and health inequality

Our research focuses on a single technological advancement in one specific context, but we argue that the epidemiological transition that Italy experienced resembles that of many other countries. Therefore, some of the mechanisms we highlight may be applicable to other regions. More specifically, we show that new technologies, especially when they are relatively more affordable by the majority of the population, can lead to a substantial reduction in both mortality rates and health disparities.

Further information:

  • The associated image of this post was taken from Wikimedia Commons (link).
  • This blog post is based on an article published by the author (and other colleagues) in Demography (link).
  • Jurre Jochemsen (a master student at Wageningen University) has provided assistance with the drafting of the piece.


  • Acemoglu, D., & Johnson, S. (2007). Disease and development: The effect of life expectancy on growth. Journal of Political Economy, 115, 925–985.
  • Alsan, M. Atella, Vincenzo, Bhattacharya, J., Conti, V., Mejía-Guevara, I. and Miller, G. Technological Progress and Health Convergence: The Case of Penicillin in Postwar Italy. Demography 1 August 2021; 58 (4): 1473–1498.
  • Bennett, J. & Chung, K. (2001). Alexander Fleming and the discovery of penicillin. Advances in applied microbiology, 49, 163–184.
  • Cutler, D., Deaton, A., & Lleras-Muney, A. (2006). The determinants of mortality. Journal of Economic Perspectives, 20 (3), 97-120.
  • Easterlin, R. (1999). How beneficent is the market? A look at the modern history of mortality. European Review of Economic History, 3, 257–294.
  • Preston, S. H. (1975). The changing relation between mortality and level of economic development. Population Studies, 29 (2), 231-248

Author details

Henry J. Kaiser, Jr. Professor (Stanford University)

Master Student (Wageningen University)