Stock Market Cycles
Lieven Baele (Tilburg U.) and Geert Bekaert (Columbia U.)
October 2026
We develop a new methodology to identify bear and bull markets in the US stock market, which only relies on calibrating a bandwidth parameter around each trading day to define peaks and troughs. We find this bandwidth to be 8 months leading to bear (bull) market spells with a median duration of 74 (294) days and median cumulative returns of-16.2% (+37%). Within each stock market phase, we formulate a regime switching model differentiating between normal down, crash and recovery regimes within a bear market and normal up, risk-on and correction regimes within a bull market. The model has two special features: it focuses on mean regimes and uses cumulated returns rather than daily returns. A daily return model with common volatility and mean regimes performs very poorly. We further validate the model using statistical tests, textual analyses using Wall Street Journal articles and an anatomy of important stock market episodes.
Concentration or Regime? The Changing Macro Sensitivities of U.S. Equities
Mohammad Bagher Karimi (independent researcher)
November 2026
We examine whether the sensitivity of U.S. stock returns to real yields, breakeven inflation, oil, gold, and the dollar changed as the S&P 500 became concentrated in a few large technology firms. Using weekly data for 2010-2026 and an index reconstructed from public data, we find significant structural breaks in early 2013, early 2020, and mid-2023. The confidence interval of the last break includes late 2022, but no break is estimated at the release of ChatGPT. An exact decomposition of index betas shows that changes in index weights contribute little. Most changes are common to nearly all firms, including the loss of the market’s positive exposure to breakeven inflation after 2022. The largest firms’ own real-yield betas rose relative to the average stock after 2022, but AI intensity in pre-2022 10-K filings does not explain this beyond firm size. The evidence points to a macroeconomic regime shift, not concentration or AI, as the main source of the change.
Using Economic Regimes to Inform a Rules-Based ETF Allocation
Manuel Bossi (Stevens Institute of Technology)
November 2026
How much of a macro-regime allocation’s return comes from its growth and inflation “season” scores, and how much from the price-based risk controls applied to them? This paper decomposes an economic-season allocation system that the author designed and introduced to a graduate student-managed investment fund at a US university. The evaluated account is a standalone managed sleeve backtest, separate from the live fund. The scores blend four baskets of exchange-traded funds (ETFs), and price rules set the holdings. Over April 2019-September 2026, which is out of sample for the season-basket coefficients, the selected-equity sleeve earned 21.57% a year net of modeled costs with a 16.23% maximum drawdown. At matched forecast risk over April 2017-September 2026 it led a risk-managed global equity account by 8.38 points a year. Most of this edge comes from the price-based layer. PV (price controls plus a volatility cap) holds the four baskets in equal shares under the same price-based controls. At equal forecast risk over April 2019-September 2026, the full system leads PV by 0.63 percentage points a year with selected equities and 0.20 with broad equities (95% intervals [-0.286, 1.739] and [-0.399, 0.984]), a lead concentrated in 2020-2022. For allocators, the price-based layer is the part to build first, and a matched-risk test against a price-based comparator built from the same baskets measures a macro overlay; here it caps the macro increment at about 1.7 points a year. The 2019-2026 record was studied during development, so the evidence is a historical reconstruction.
Stepping Aside Only When Credit Breaks: A High-Yield Spread Regime Filter for Equity Exposure
Randy Harris (Dual Momentum Systems)
October 2026
Price-based momentum timing rules protect equity investors in crashes, but false alarms make them exit and re-enter often. Each whipsaw risks missing a rebound while out, and every exit realizes gains. We test a macro regime rule driven by an external signal instead: the ICE BofA US High Yield option-adjusted spread. The rule steps out of US large cap equities into intermediate Treasuries only when the month-end spread is at or above 6.5 percentage points and has not yet fallen 2 points from its six-month high. From January 1997 to September 2026 it was invested 84% of the time, changed regime 0.8 times a year, and returned 13.0% a year against 10.2% for buy-and-hold, with maximum drawdown cut from -51.0% to -24.6%. Exit levels between 5.5 and 7.5 perform similarly. After tax at 24% short-term and 15% long-term, 83% of realized gains are long-term. The rule’s blind spot is a bear market without credit stress: in 2022 the spread peaked at 5.87 and the rule stayed invested. Pairing the rule with a diversified growth sleeve narrows that gap: the combination returned 13.7% a year after trading costs, with a maximum drawdown of -16.2%.
Simple Dynamic Stock/Bond/Gold Portfolios
Nikhil Devanathan (Ph.D. candidate, Stanford U.), et al.
September 2026
For more than four decades, the 60/40 stock/bond portfolio has served as a benchmark for delivering reasonable returns without excessive risk. More recently, a 50/30/20 stock/bond/alternative portfolio has been suggested. We use gold as the alternative and as an inflation hedge. In this paper we ask: how much improvement over these benchmark fixed-weight portfolios can be obtained using widely available public data and standard methods from quantitative finance?
We restrict ourselves to long-only dynamic portfolios of stocks, bonds, and gold, plus cash, rebalancing monthly, using only publicly available data. We evaluate portfolios on the conventional metrics: return, volatility, Sharpe ratio (computed in excess of the federal funds rate), drawdown, and turnover, in addition to consistency of performance over time, judged by the consistency of the realized annual volatility.
Over the 20–year period 2006–2026, using a conservative estimate of trading costs, we show that all risk-adjusted and drawdown metrics are improved using simple volatility control, where we dynamically mix the fixed-weight portfolios with cash so as to target a fixed volatility. This method relies on a simple estimate of portfolio volatility derived from past returns. We also demonstrate that more sophisticated portfolios based on convex optimization—similar to those used in quantitative hedge funds—yield further substantial improvement in return and risk-adjusted return. We consider two such portfolios, one that uses a simple estimate of future returns based on past returns, and one that forecasts future returns based on past returns and just a handful of widely available public economic data. These portfolios also outperform a suite of standard risk-based allocation methods, such as risk parity and minimum variance, evaluated on the same assets and data.
Risk Premium Neglect in Bonds and Stocks
Peter Vaporakis (Copenhagen Business School)
November 2026
I study how professional forecasters expect financial markets to respond to changes in monetary policy using survey expectations of policy rates, Treasury yields, and stock prices. I find that when forecasters expect tighter future monetary policy, Treasury yields rise less than expected, while stock returns decline and expected returns stay flat. Forecasters therefore systematically overestimate long-maturity Treasury yields and stock returns. These forecast errors are consistent with risk premium neglect: forecasters neglect the risk premium implications of their policy beliefs in their asset price forecasts. A model in which agents neglect the cyclicality of the price of risk reproduces these patterns.
Where Alpha Cannot Be Measured: Regime-Conditional Performance Attribution
Mark S. Tennenbaum (Life UnLocked Partners)
January 2026
Alpha is small and noise is not. A fund’s daily residual runs about 42 basis points on the calmest days and 112 on the most violent, against a long-run best alpha near 4 basis points a day. We separate days by market move size and show alpha is best measured on quiet days rather than loud ones. Across 18 US factor ETFs the method finds no false alpha under the attribution model. Detecting a 10 percent alpha on excluded days with 80 percent power would require a 17- to 33-year record for the median factor ETF, depending on the benchmark. Rising dispersion in a crisis is disproportionately beta separation, not opportunity. Investors need only a return series and a market series.
Learn To Use R For Portfolio Analysis
Quantitative Investment Portfolio Analytics In R:
An Introduction To R For Modeling Portfolio Risk and Return
By James Picerno
