A 20-basis-point monthly return gap clearing only the 10% bar is thin support for trading on a president's religion. The paper's new contribution is its religious classification of Latin American presidents. Mella defends the result with a local-currency gap of 0.41%, significant at 1%. He reports that figure in the text without tabulating it.
How religion enters the return test
Mella starts from the power of the presidency. In Brazil, Chile, Colombia, Mexico and Peru, presidents hold decree, veto and agenda power. Their parties should therefore affect policy, with consequences for cash flows and discount rates. Alongside the familiar left-right classification, Mella asks whether a president's party has Christian Democratic roots. He associates that tradition with fiscal prudence and cross-class coalitions. Corporate-finance research connecting religiosity to lower risk-taking and less hoarding of bad news supplies his link from presidential religious affiliation to firm returns.
He updates Coppedge's 1997 expert survey of Latin American parties to classify each party as Left, Center or Right, and separately as Christian or Secular. Dummies identify the sitting president's bloc each month. Matching them to Compustat Global firm returns from 1989 to 2017 produces 104,497 observations.
Brazil accounts for 53% of the rows.
The headline tables use pooled OLS to regress monthly USD excess returns on those dummies, without an intercept or controls. Business-cycle controls enter in an appendix. The election test instead uses MSCI country index returns on the day after each of 33 elections from 1988 to 2018. Here, left and right describe the winner relative to the runner-up, rather than an absolute political position.
Centrist presidents have average monthly excess returns of 1.90%, compared with 1.59% under the right and 1.42% under the left. Christian-party presidents average 1.70%; secular presidents, 1.50%. After elections, a relatively left winner brings a USD return of -0.77%, against +2.85% for a right winner. The gap is 3.62 points (OLS p 0.042). Christian winners bring +5.45%, compared with +0.495% for secular winners. Mella also finds lower volatility, lower crash risk and higher ROE under Christian presidents. He attributes the premium to cash flows and underreaction, ruling out a risk premium.
The centrist result fits the OECD evidence
Bohl and Gottschalk, cited in the paper, found a US-style left premium in only Denmark, Germany and the US among 15 OECD countries. Latin America's centrist ranking fits that evidence. The religious classification supplies the actual addition.
Reading the pooled estimates as a spread requires care. The bloc dummies change only at the country-month level, while t-statistics of 15 to 37 are calculated from roughly 100,000 firm rows. We found no mention of clustering by month or by country. Equation 1 omits country fixed effects. With Brazil supplying 53% of the rows, a bloc dummy partly captures which Brazilian president held office.
Mella acknowledges the homogeneity assumption as a limitation of pooling: the estimates treat economic conditions, market structure and other institutions as alike across the five countries. He does not discuss standard errors. The averages are also firm-weighted, a different object from a country portfolio an investor could hold.
What can four Christian-party wins establish?
The paper states the problem plainly: "in only four elections, the candidate of the Christian party ended up winning the ballot." Mella also writes that Christian regimes "remain more the exception than the norm in the sample" and identifies low variability in the political variable as a limitation.
With four treated events, the OLS p-values are 0.0718 in USD and 0.131 in local currency. Bootstrapping produces 0.0003 and 0.0008. Tightening inference by two orders of magnitude through resampling four events calls for an explanation. We found no description of the resampling procedure. A further discrepancy remains between the text's 3.37% low end for the gap and the table's smallest figure, 3.74%.
Four events cannot tell you how a fifth will trade.
Capturing the return poses another problem. The reaction occurs in the first session after the vote. Whoever earned the average 5.45% held the index before the count, which requires calling the winner ahead of time.
Lower risk, higher ROE
Most of the risk results run in the same direction. Daily index volatility averages 1.74% under secular presidents, versus 1.41% under Christian presidents. Chile's Christian-minus-secular crash-risk gaps are -0.37 on NCSKEW and -0.14 on DUVOL, both significant at 1%. Mexico's corresponding figures are -0.06 and +0.04, neither significant. The panel EPU gap is 9.73, with 100.00 for secular presidents and 90.27 for Christian presidents. The text calls it significant, although the table gives it no significance star. Christian presidents' ROE coefficient over secular presidents is 0.0363, significant at 1%.
These findings argue against a risk premium. The case for underreaction is harder to accept, partly because Mella argues in both directions. In Section 5.4, he writes that "underreaction is needed to the good news of a president being elected" for the cash-flow account to work. Section 4.2 says the election result should be absorbed immediately: "reflected in one day, as the efficient market hypothesis suggests".
His evidence for underreaction consists of the monthly return premium over a presidential term and the higher ROE (0.0363). The claim therefore rests on returns earned across multi-year presidencies, without a test of post-election drift. Mella says cumulative windows from 2-day to one-week remain significant on the Christian-secular axis, but supplies no figures. Those windows include day one and cannot separate delayed adjustment from the initial reaction.
The subsamples carry the most weight. From 2010 to 2017, the secular-minus-Christian gap is -0.59% per month, significant at 1%. From 1987 to 2009, it is an insignificant -0.03%. Mella characterises the split results overall as mostly consistent in magnitude and sign, with significance lost in two of eight combinations.
Only Chile and Mexico have Christian presidencies in the per-country tables. The post-2010 premium thus depends on at most two countries. An independent classification, the V-Party continuous religiosity score, confirms the direction with a coefficient of -0.0021 and t of -3.61. This addresses part of the concern about researcher discretion in assigning parties to categories. The pooled standard errors remain the same problem.
Two findings would change my view: a Christian gap before 2010 that survives country-clustered errors, or drift starting on day two.
A backtest remains out of reach
The monthly presidential index is available in the paper's supplementary file. We lack Compustat Global coverage for these five markets and the MSCI daily index histories. Substituting US-listed country ETFs would change the firm universe, so a replication using them would test a different claim. We stopped short of a backtest.
The paper leaves us with a hypothesis about Christian Democratic presidents worth pricing at the next Chilean or Mexican election. An allocation rule remains out of reach.