Of Presidents and Babies

In 2014, Ben Blatt used Social Security Administration (SSA) data to suggest anecdotally that Americans have a habit of naming their newborns after sitting presidents. Blatt found that the number of babies (per 100,000 births) given the same name as a president increase when this politician runs for the presidency and during his tenure in the Oval Office.

For example, the relative popularity of the name “Grover” increased during the presidential election of 1984, which Grover Cleveland won, and remained higher than expected during the Cleveland administration. This pattern repeated itself when Grover Cleveland was elected president for a second, non-consecutive term in 1892, but to a lesser degree.

Quasi-Experiment

I sought to replicate the results from Blatt’s analysis with data from the same source using a quasi-experiment (the code for which can be found here). Specifically, I tested the hypothesis that the relative popularity of a president’s name is larger:

1. During the president’s tenure than during the period preceding his presidency, and
2. Than the popularity of the names of other presidents.

This quasi-experimental reformulation of Blatt’s hypothesis is a more stringent test of the idea that sitting presidents influence the naming decisions of the parents of newborns because it could suggest that the influence of presidents on baby names is specific to (1) the tenure of a sitting president and (2) their particular name (as opposed to similar presidential first names).

Data

I used SSA data of names from Social Security card applications for births that occurred in the United States between 1910 and 2014. (Note that these data were only available by year, whereas Blatt’s data count baby names by day).

The analyzed data were limited to presidents who took office after the earliest year in the newborn data; that is, all presidents since, and including, Woodrow Wilson. For all presidents, first names were used instead of nicknames (e.g., James rather than Jimmy).

Method

For each president, I counted two values:

1. The number of babies given his name during his administration [1] (e.g., the number of babies given the name William during Bill Clinton’s administration; prez_during), and
2. The number of babies given the name of all non-sitting presidents during the same period of time (others_during).

The ratio of the first two numbers (prez_during / others_during) estimates the popularity of the sitting president’s name against the popularity of the names of all of the other presidents during the sitting president’s administration.

For each president, these two counts were repeated for a time period of the same length before his administration (e.g., the number of babies given the name William during the 8 years that preceded Bill Clinton’s 2-term administration; prez_pre and others_pre).

The ratio of these two numbers captures the same popularity difference as before, but during the period before this name becomes the sitting president’s name (prez_pre / others_pre).

If Americans have a bias to name their babies after sitting presidents, then the difference between these two ratios (prez_during / others_during – prez_pre / others_pre) should, on average, be positive at a rate higher than chance. In other words, whatever the popularity advantage of a president’s name to that of other presidential names before that president takes office, that popularity advantage should increase during the president’s tenure.

Results

Across the 17 presidents in the data, the mean difference of these ratios was 0.008 (SD = 0.02), an estimate not statistically reliable with a two-tailed t-test (t = 1.59, p = 0.13). Nonetheless, the ratio difference was in the expected direction for 11 of the 17 presidents (65% of the time). The figure below shows the mean ratio difference with the standard error of the mean as its error bar.