Investment columnist Joachim Klement: Demographics cannot predict the outcome of US elections
Klement cites a 2026 historical backtesting study by Calvo et al., pointing out that there is no similarity between election predictions based on demographic variables and actual results, arguing that investors who rely on demographics to judge the direction of American politics should re-examine this analytical tool.
The United States is approaching its midterm elections, and investment columnist Joachim Klement uses this occasion to reiterate a judgment he has long held: demographics is not destiny. His core argument is that using demographic trends to predict election outcomes has never succeeded historically.
The trigger for this judgment was an opinion piece by Thomas Edsall in The New York Times this spring. Edsall argued that the 2030 census will shift electoral votes from states with declining populations, such as New York and California, to states with growing populations, such as Texas and Florida; since the former are reliably Democratic states and the latter are reliably Republican states, this will increase the Republican advantage by approximately 12 electoral votes between 2032 and 2040.
Klement points out that this analysis implicitly assumes a key premise: voters in the 2030s will vote along the same demographic fault lines as they do today. That is, it assumes that Black voters and highly educated voters reliably vote Democratic, while older voters and working-class whites reliably vote Republican. Klement considers this a bold assumption—older people die, so no one knows if the elderly in the 2030s will still reliably vote Republican; working-class whites were historically a key constituency for the Democrats, and no one knows if they will still shun the Democrats ten years later; Hispanics currently lean more toward voting Democratic than Republican, but they are highly religious and could well switch to the Republican Party if they feel alienated by Democratic policies.
To support this criticism, Klement recommends a new paper by Richard Calvo and his collaborators. The paper analyzes the historical performance of election predictions based on demographic variables. Klement describes the results: comparing the forecast line based on demographic considerations (dark black) with the actual election result line (light gray) in the charts, there is no resemblance between the two. His conclusion is that predicting election trends using demographics is a complete waste of time.