Required in Empirical Social Science: Numbers

By Aaron S. Edlin and Michael Love:

Knowing the magnitude and standard error of a symbolic measure is more important than simply knowing the sign of the measure and whether it is statistically significant. However, we find that even in the top journals, where experienced social scientists choose their headline results – the results they put in the papers – most ignore this theory and do not report the magnitude or accuracy of their findings. They do not provide numerical title results in 63% ±3% of empirical economics papers and 92% ± 1% of historical political science or social science papers between 1999 and 2019. In addition, they never report accuracy (0.1% ± 0.1% ) in the subject results. Many social scientists seem wedded to a tradition of null hypothesis testing rather than a tradition of measurement. There is another way: medical researchers often report numerical superiority (98% ±1%) and accuracy (83% ± 2%) in the subject’s results. Trends suggest that economists, but not political scientists or social scientists, are warming to numerical reporting: the share of empirical economics articles with numerical results has doubled since 1999, and economic articles with numerical results are receiving more citations (+19% ± 11%). .

About someone on Twitter?



Source link