Calendar Effects in US GICS Sector ETFs: An Empirical Seasonality Framework

This study examines calendar-based return patterns across eleven US GICS sector ETFs (the SPDR Select Sector suite) relative to the SPY broad-market benchmark over January 2019 through the latest available sample. Using monthly total returns constructed from Yahoo Finance adjusted closes, we estimate unconditional calendar-month and quarterly averages, sector–month heatmaps, cyclical versus defensive dispersion metrics, and exploratory one-sample t-tests for month-level drift. Cyclical industries exhibit higher calendar dispersion (average seasonality standard deviation approximately 2.1% versus 1.6% for defensives); November emerges as the dominant positive calendar month for SPY and most sector ETFs; September ranks among the weakest months. Technology and consumer discretionary sectors show the strongest calendar-profile alignment with SPY (correlations above 0.89), while energy decorrelates most strongly. Statistical tests identify several month effects at the 10% level, though limited annual observations per calendar month constrain inferential power. The framework provides transparent, reproducible sector seasonality diagnostics for research and risk monitoring—not a trading recommendation.

Calendar anomalies in equity markets have been documented since at least Rozeff and Kinney (1976), with subsequent evidence that seasonal patterns can persist at the industry level across countries (Heston and Sadka, 2008). While aggregate market seasonality has received extensive attention, sector-level heterogeneity remains economically important for tactical asset allocation and risk budgeting.

This research addresses three core questions:

• Do US sector ETFs exhibit stable, month-specific return patterns that differ materially across industries?
• Do macro-sensitive cyclical sectors display greater calendar dispersion than defensive sectors?
• To what extent do sector calendar profiles co-move with SPY, and which months show statistically distinguishable average returns given small-sample constraints?

We adopt a descriptive inference framework—unconditional calendar means, hit rates, and relative performance versus SPY—rather than claiming risk-adjusted alpha or out-of-sample predictability. The sample spans pre- and post-pandemic regimes, Federal Reserve tightening and easing cycles, and the 2022 inflation shock.

Several economic channels plausibly generate sector-specific calendar patterns:

• Earnings seasonality — corporate disclosures cluster in January, April, July, and October, creating recurring information shocks that affect growth sectors differently from defensives.
• Institutional flows — pension rebalancing, window dressing, and tax-loss harvesting can produce turn-of-year and November strength in liquid sector ETFs.
• Macro sensitivity — cyclical industries (energy, industrials, financials, materials) may amplify business-cycle-linked calendar effects; defensives (staples, utilities, health care) dampen them.
• Risk appetite cycles — year-end volatility compression and risk-on positioning often favour growth and technology into Q4.

Bouman and Jacobsen (2002) document Halloween-type seasonal regularities internationally; although aggregate US effects have weakened in recent decades, sector-level structure may remain. Formally, if $r_{i,t}$ denotes the total return of sector ETF $i$ in month $t$, calendar seasonality implies $\mathbb{E}[r_{i,t}\mid \mathrm{month}(t)=m]$ varies with $m\in \{1,\dots ,12\}$ in a pattern measurable over multi-year windows but not necessarily stable enough to forecast without regime risk.

We evaluate four testable propositions:

• H₁ (sector heterogeneity): Calendar-month average returns differ across GICS ETFs; seasonality dispersion ${\sigma }_i^{\mathrm{season}}=\mathrm{std}({\bar{r}}_{i,1},\dots ,{\bar{r}}_{i,12})$ varies materially across $i$.
• H₂ (cyclical amplification): $\mathbb{E}[{\sigma }^{\mathrm{season}}\mid \mathrm{cyclical}]>\mathbb{E}[{\sigma }^{\mathrm{season}}\mid \mathrm{defensive}]$.
• H₃ (benchmark co-movement): Growth-oriented sectors (Technology, Consumer Discretionary) show higher correlation between their twelve-month calendar profile and SPY's than idiosyncratic sectors such as Energy.
• H₄ (Q4 strength): Average returns in October–December exceed April–September for growth-oriented cyclicals, consistent with year-end risk appetite.

For month-level inference, the null is $H_0:\mathbb{E}[r_{i,m}]=0$ for each sector–month pair $(i,m)$. Rejection at $p<0.10$ is exploratory evidence of calendar drift, not risk-adjusted alpha.

The universe comprises eleven SPDR Select Sector ETFs—XLK, XLF, XLE, XLV, XLY, XLP, XLB, XLI, XLU, XLC, XLRE—with SPY as the benchmark. Daily Yahoo Finance adjusted closes aggregate to calendar-month total returns $r_{i,t}=P_{i,t}/P_{i,t-1}-1$ from 2019-01-01 onward (~eight annual cycles per calendar month).

The empirical pipeline implements five analytical layers:

• Layer 1 — unconditional calendar means: ${\bar{r}}_{i,m}=N_m^{-1}{\sum }_{t:\mathrm{month}(t)=m}{r}_{i,t}$ and hit rate $H_{i,m}=N_m^{-1}\sum 1\{r_{i,t}>0\}$.
• Layer 2 — relative performance: ${\Delta }_{i,m}={\bar{r}}_{i,m}-{\bar{r}}_{\mathrm{SPY},m}$ isolates rotation versus the benchmark.
• Layer 3 — seasonality dispersion: ${\sigma }_i^{\mathrm{season}}=\mathrm{std}({\bar{r}}_{i,1},\dots ,{\bar{r}}_{i,12})$.
• Layer 4 — cyclical/defensive grouping: cyclicals (Technology, Financials, Energy, Consumer Discretionary, Materials, Industrials, Communication, Real Estate) vs defensives (Health Care, Consumer Staples, Utilities).
• Layer 5 — inferential diagnostics: one-sample t-tests $t={\bar{r}}_{i,m}/(s/\sqrt{n})$ at $p<0.10$; calendar-profile correlation with SPY; Q4 (Oct–Dec) vs Q1/Q2 (Apr–Sep) spreads.

Analysis over the 2019–2026 sample reveals the following patterns:

• Cyclical amplification (H₂): average ${\sigma }^{\mathrm{season}}$ ~2.1% for cyclicals vs ~1.6% for defensives.
• November dominance: SPY November average ~4.9%; Financials (6.6%), Technology (5.8%), Industrials (5.8%), Materials (5.8%) among sector leaders.
• September weakness: SPY September average ~−2.1%; multiple sectors share September as worst month.
• Benchmark co-movement (H₃): Technology–SPY calendar correlation 0.93; Consumer Discretionary 0.90; Energy 0.48 (most idiosyncratic).
• Q4 strength (H₄): Technology Oct–Dec average ~3.3%; Financials Q1/Q2 minus Q4 spread ~−2.5%.
• Exploratory inference: Utilities July ~4.5% ($p<0.01$); November cluster across sectors at $p<0.10$—treat as hypothesis-generating given ~eight obs per month.
• Hit rates: SPY 100% positive Julys, 86% positive Novembers, 43% positive Septembers in sample.

The framework quantifies sector seasonality from liquid ETF proxies, but limitations warrant careful interpretation:

• Unconditional means conflate regimes (pre-pandemic, 2020 crash/recovery, 2022 rate shock); structural breaks can reverse patterns.
• ~eight annual observations per month limits t-test power; 132 sector–month pairs inflate false-discovery risk at $p<0.10$.
• ETF returns reflect basket composition and GICS methodology changes, not pure stock-level seasonality.
• No adjustment for market beta or Fama–French factors—positive calendar months may reflect risk-premium concentration.
• Yahoo Finance adjusted closes may differ from CRSP/Bloomberg total-return indices.

Practical advantages remain: full transparency, reproducibility from public data, interpretable metrics (${\bar{r}}_{i,m}$, ${\Delta }_{i,m}$, ${\sigma }_i^{\mathrm{season}}$), and comparability with RRG and factor tools on this site. Intended use: research prioritization, risk monitoring, and investment-committee communication—not mechanical calendar trading.

US GICS sector ETFs exhibit measurable, heterogeneous calendar structure over 2019–2026:

• Cyclical sectors show greater calendar dispersion than defensives.
• November and September function as recurrent positive and negative poles.
• Growth-oriented sectors track SPY's seasonal rhythm; energy diverges.
• July and November effects appear in exploratory inference, though small samples limit confirmatory strength.

Natural extensions: SARIMA seasonal modelling (as in the NIFTY 50 study), Fama–French residual seasonality, bootstrap CIs for ${\bar{r}}_{i,m}$, and Bonferroni/FDR correction across sector–month pairs. Data refreshes via npm run data:sector-seasonality-analysis when Yahoo Finance prices update.

The Results section reproduces the full seasonality dataset. Each panel serves a distinct diagnostic role:

• Summary cards — cyclical vs defensive seasonality σ, SPY best calendar month, universe metadata.
• Sector vs SPY table — calendar-profile correlation, Q4 vs Q1/Q2 spreads, best/worst months per sector.
• Heatmap — sector rows × calendar-month columns; colour encodes average monthly return (green positive, red negative).
• Sector drill-down — monthly returns, relative vs SPY, hit rates, quarterly bars, and SPY benchmark pattern for a selected ETF.
• Cyclical vs defensive table — per-sector seasonality σ and return volatility by macro group.
• Statistical tests — exploratory month effects with $p<0.10$.