Mayors' party and commune finances

Evidence

Left and right towns differ in many ways besides their mayor, so comparing them says little. Close elections avoid this. Where a left list beat a right one, or lost to it, within 10 points of the vote, which side won was closer to chance than in a landslide. Across the 2014 and 2020 municipal elections, the communes where the left narrowly won are compared with those where it narrowly lost, on what changed over the following term.

None of the pooled differences between narrowly elected left and right mayors is statistically significant. Between a narrowly elected left mayor and a right one, running costs per head differ by −0.9%, estimated, staff costs by 3.2%, estimated, equipment spending by €−21, estimated per head and end-of-term debt by €−82, estimated per head; private jobs by −1.3%, estimated and income per household by −0.3%, estimated. The placebos, the same changes over the term before the election, are near zero, as they should be. Term by term, two results stand out: over the 2014 term, in accounts to 2017, left mayors ended with less debt per head; over 2020-2025, private jobs grew less in their towns. With this many comparisons one or two are expected to stand out by chance; neither term result should be read on its own. French communes must balance their operating budget and may borrow only to invest, and most of their grants follow formulas: the room a new majority has is narrow.

The design measures what a narrow left win did in races fought between a left and a right list, in communes of 1,000 inhabitants and more (2014) or 3,500 and more (2020, where only those have a political label): communes whose race was not close, or not between left and right, are not in it, and private jobs are Urssaf's salaried employees of private employers, which leave out the self-employed, farm workers and public employees. Each list is classed by the nuance the Ministère de l'Intérieur gave it, grouped as its circulaire of 3 February 2020 groups the nuances, and in 2014, for which it published no grouping, the same way: the left is the far-left and left blocs; the right is the centre and right blocs, so La République en marche, the MoDem and the UDI count as right; the Rassemblement national, the Front national in 2014, and other far-right lists are in neither, so a race whose top two included one is left out. Counting the centre apart gives the same picture. Ranges are 95% intervals from the regression standard errors. McCrary's test of the density of the margin at zero gives a log ratio of −0.22, estimated (standard error 0.13, estimated) at the 10-point bandwidth of the estimates: about a fifth fewer narrow left wins than narrow left losses, within what chance gives at 10 points; at 20 points the same deficit, −0.20, estimated (standard error 0.09, estimated), is outside it. At 5 points the ratio is −0.11, estimated (standard error 0.19, estimated). Close races split only roughly evenly between left and right; the placebos, which find no difference over the term before, are the check that the two sides were otherwise alike. Election results from the Ministère de l'Intérieur; the code is analysis/politics.ml and the results are in results.csv.