Seasonality and presidential election cycle regimes

A four-year regime overlay

The election-year cycle is the four-year political calendar that labels each year as postelection, midterm, preelection, or election. A cycle-year label from 1 through 4 lets a spreadsheet split monthly index returns by that political year.

Editorial: treat those labels as a market-regime overlay. The overlay still has to be crossed with monthly seasonality before any entry, exit, or cash-park rule is accepted as one testable procedure.

The unsplit monthly baseline

Over the 1950-1995 monthly S&P 500 close-to-close sample, the average month gained 0.74%, the index rose in 57% of months, the average annual gain was 9%, and 71% of years closed higher. Dividends were excluded. Those figures are the baseline before any cycle-year split.

Returns and rise-frequency by cycle year

When the same months were split by cycle year, average monthly returns were 0.12% in postelection years, 0.46% in midterm years, 1.50% in preelection years, and 0.84% in election years.

Average annual returns in that sample were 2.00% in postelection years and 4.63% in midterm years, versus 18.72% in preelection years and 10.08% in election years.

Rise-frequency at the year scale was uneven. The index never closed down for a full preelection year from 1950 to 1995, rose in 91% of election years, and closed higher in only 45% of postelection years and 55% of midterm years.

An illustrative system, not a plan

An illustrative system held the index only in preelection and election years and used cash-park in 90-day Treasury cash in the other two years. That regime-aware allocation turned a 1950 stake of 10000 into 733605 by 1995, versus 372388 for buy-and-hold, with a maximum annual loss of -18.7% versus -41.3%.

The reverse allocation, long only in postelection and midterm years, grew 10000 only to 56297 over the same span, with a 3.9% average annual return. That path was worse than an all-cash 90-day Treasury path that grew 10000 to 110905.

Editorial: this is an illustrative system used only to test a relationship, not a complete money-management plan. A risk-adjusted-return comparison would still reduce average annual return by a penalty based on monthly variability before the same procedure is accepted.

A regime label can change drawdown

Several large historical declines, including 1962, 1973-74, and 1982, clustered in postelection or midterm years. That is why a cycle-year filter can change both return and drawdown when it is treated as a regime label rather than a standalone forecast.

Seasonal windows inside the cycle

A seasonal window is a calendar month whose typical return and rise-frequency are measured inside a named year of the cycle, not across all years at once. Crossing month-of-year with cycle year changed the ranking of individual months.

January was weak mainly in midterm years and strongest in preelection years. May was useful mainly in postelection years. June was poor in the first two cycle years and strong in the last two. July averaged more than 1% in every cycle year. September was weak except in election years. October was dangerous in preelection years but otherwise solid.

Editorial: a combined-calendar filter marks a month as tradable only when both its seasonal rank and its cycle-year label support exposure.

The six-month list and the midterm-October start

If only six months in the four-year cycle could be held, the ranked list was preelection January at 5.38%, preelection December at 3.35%, preelection April at 3.27%, midterm October at 2.51%, midterm November at 2.46%, and election-year November at 2.32%. None of those six came from postelection years.

The stronger preelection stretch was described as beginning in midterm October and fading by the end of preelection August, an 11-month window that averaged 26.7% in the sample, or 29.5% annualized. That midterm-October start is why a January 1 year box is a weaker regime cut than the same cycle-year labels applied to the seasonal windows that carried the stretch.