Do Politicians Beat the Stock Market?

We tested five strategies on nearly 30,000 congressional stock trades disclosed since 2020. Every one of them failed. The strict filter found five politicians whose historical returns cleared every hurdle we set - risk-adjusted alpha, consistent entry timing, resilience during the 2022 bear market - and by 2025, three of those five had filed zero trades. The two who remained returned 14 percent against the S&P 500's 18. A relaxed version expanded the pool to fourteen candidates and watched the same collapse play out again, just more slowly. The most promising approach, buying whatever stock multiple politicians purchased simultaneously, generated a single standout year that traced entirely to five semiconductor stocks during a volatile stretch in April 2025. Remove those five tickers and the strategy loses money.

Congress, taken as a group, trades about as well as a typical retail investor.

That finding has a caveat worth examining.

Narrow the window from years to weeks, and certain politicians have posted returns that resist easy explanation. The largest stretch belongs to Tommy Tuberville, who outpaced the market by 72 percentage points over three months of trading in stocks adjacent to his Senate Agriculture Committee seat. Nancy Pelosi's spouse posted a 59-point gain in a comparable window, concentrated in technology positions during Congressional deliberation on AI and semiconductor policy. McConnell and Mullin showed smaller edges in shorter bursts: 32 and 2.9 points respectively, each in sectors overlapping their committee assignments.

We found each of those windows by searching for the best stretch each politician produced across the full dataset - the most generous possible framing of the insider-knowledge narrative. Outside those windows, the same politicians show flat or negative returns relative to the index. A 59-point advantage over three months tells you what happened during those three months; it carries no predictive power over the next three. Whether these windows reflect genuine skill or luck compressed into a small sample is the question the rest of this analysis tries to answer.

30,000 trades, and most of them are noise

Since 2012, members of Congress have had to disclose stock transactions within 45 days. The median delay runs about 29 days. By the time a filing becomes public, the trade is nearly a month old.

Our dataset covers 26,020 House trades and 3,665 Senate trades, spanning 2020 through early 2026, from 173 politicians who traded frequently enough to evaluate.

Most of these politicians trade infrequently or buy broad index funds, which reveals nothing about stock-picking ability. The question that matters is whether the small minority who actively trade individual stocks can consistently pick winners - and whether that consistency, if it exists, survives the month-long delay before anyone outside Congress can see what they bought.

An entire industry has grown around the premise that it can. Dozens of accounts on X post filing screenshots within hours. Newsletter services charge monthly subscriptions for curated alerts. At least three fintech startups have built consumer products around real-time congressional trade tracking. The promise is uniform: politicians know something, and you can profit by copying them. None of these services publish evidence that copying has worked over a sustained period, tested against a benchmark, with the disclosure delay built in. That gap prompted this analysis.

Politicians time the market slightly better than chance - and not enough to matter

The first test was simple. When a politician buys a stock, how does their purchase price compare to where that stock trades over the following 60 days?

Imagine the stock's 60-day price range as a ruler from 0 to 1. Buying at the absolute bottom scores 0 - perfect timing. Buying at the top scores 1. Pure chance averages 0.50.

Across all 30,000 trades, politicians collectively scored 0.444. Detectable over thousands of observations. Far too small to trade on. The gap between that score and the 0.500 random baseline amounts to buying a stock roughly two percent closer to its short-term low, which vanishes inside any real portfolio's transaction costs. During the 2022 bear market, the score flipped above 0.50: politicians were buying closer to short-term highs as prices fell around them. A handful of individuals held scores below 0.40 across multiple years, a result genuinely difficult to attribute to luck, but the group average tells the story the table confirms.

The strict filter identified five skilled traders. Three of them vanished.

Five politicians, out of 36 with enough trading history, passed every filter we constructed: risk-adjusted alpha above the S&P 500, consistently good entry timing across multiple years, and performance that held up through the 2022 bear market. We weighted them into a portfolio and tested it forward on 2025 data that had been held out completely.

Three of the five had filed zero trades in 2025. Gone. Blind trusts, retirement from active trading, or something the filings don't explain. The surviving pair returned about 14 percent for the year against the index's 18.

A two-person portfolio is a bet on two specific individuals continuing to trade the way they always have. Nothing in the disclosure data signals when that bet expires.

A wider net caught the same problem

Loosening the requirements produced one genuinely interesting finding. When we delayed portfolio entry by the typical 29-day disclosure lag, about 95 percent of the historical outperformance survived. Whatever edge existed in the training data came from which stocks politicians chose, not the precise timing of their purchases. That distinction matters because it suggests the disclosure delay may not destroy the signal entirely.

The wider funnel admitted fourteen candidates instead of five. It scored them on whether outperformance survived the delay and how far returns diverged from the broader market. More dimensions. More data. Same result.

Five of seven qualifying members filed zero trades in 2025. The blended return roughly matched the index, minus a fraction of a percent. Widening the pool gave the dropout problem more room to manifest without solving it.

Consensus buying: when multiple politicians pick the same stock

The third strategy abandoned individual tracking entirely. Forget which politicians trade well. Ask instead: when two or more of them independently buy the same stock in the same month, does that collective conviction outperform?

The rule was simple. Pool about 30 politicians who showed positive outperformance in training data. When two or more independently disclose purchases of the same stock within a 30-day window, buy it. Hold 60 trading days. Sell.

During the 2023-2024 training period, the consensus portfolio returned 35 percent against the S&P 500's 68. That headline comparison is misleading because the portfolio sits in cash between signals while the index is always fully invested. The risk profile tells the more interesting story: two-thirds of all trades were profitable, the worst drawdown was about 8 percent (less than half the market's 18 percent pullback), and roughly 130 consensus signals fired per year.

One good year out of three, driven by one market event

On held-out 2025 data, the consensus portfolio returned 43 percent against the index's 18. Win rate: 71 percent. Risk-adjusted performance: 1.81, well above the 1.0 threshold professionals consider strong.

Place that result alongside the full training history and the picture changes. The strategy had underperformed the S&P 500 in every prior year, losing to the index by 29 points in 2023 and 8 points in 2024. The 2025 result was the first time it had ever beaten the benchmark.

April 2025 explains why. Semiconductor and large-cap tech stocks fell sharply on tariff news, then recovered sharply, and the consensus portfolio happened to hold those names through the dip. A sudden drop followed by a sharp recovery in your heaviest positions is the ideal scenario for any concentrated buy-and-hold approach. It is also a scenario that cannot be predicted or reproduced on demand.

Five stocks carried the entire strategy

Breaking down the 430 large-cap consensus signals by individual ticker revealed what was actually driving returns. Five companies, all in semiconductors or mega-cap technology, showed consistently positive excess returns. Everything else in the consensus portfolio - healthcare, financials, consumer goods, retail - either matched the S&P 500 or trailed it.

Nvidia and Alphabet led the winners, each averaging 16 to 17 percentage points of excess return across dozens of consensus signals. The losers were just as concentrated. Apple, the most frequently purchased consensus stock at 22 signals, trailed the market on average. Tesla and UnitedHealth fared worse. The table below shows the full breakdown, but the pattern is visible at a glance: the winners cluster in semiconductors, the losers scatter everywhere else.

Strip out the five best tickers and the remaining 400-odd signals underperform a simple index fund. The strategy looks diversified. Under the hood, it is a concentrated semiconductor bet entered with a month-long delay.

Three structural reasons these strategies break

Each failed strategy closed a door on a different version of the congressional-alpha narrative.

Start with the simplest: people disappear. Politicians stop filing disclosures through retirement, blind trust adoption, committee reassignment, or simple loss of interest. Three of five in the strict filter, five of seven in the relaxed version. No historical metric predicts who goes next. Any strategy that depends on specific individuals continuing to trade is one announcement away from collapse.

Then there is the question of where the edge should appear. If politicians possess genuine information advantages, those advantages should concentrate in smaller, less-followed companies where committee-level knowledge about energy policy or pharmaceutical regulation creates the widest gap between what they know and what the market has priced. We applied a $50 billion market cap ceiling to all consensus signals. Ninety-five percent vanished. Politicians overwhelmingly buy the same household names that populate every index fund, and in the thin remaining slice of smaller-company signals, the win rate fell below 50 percent.

The third problem is environmental. In 2024, a year of steady gains without sharp pullbacks, even the semiconductor consensus picks produced only a 44 percent win rate. Split signals by market conditions and the pattern emerges cleanly: purchases during brief pullbacks delivered strong returns, purchases near all-time highs delivered losses. Same politicians in both categories. Whether a trade looked brilliant or foolish depended on whether the market happened to dip around the purchase date.

What's actually happening: dip-buying on a schedule

Holding period analysis confirmed a pattern visible across every approach we tested. Average excess return over the S&P 500 stayed under one percentage point at 30, 60, and 90 trading days. Semiconductor outliers pulled the averages upward. The median excess, a better measure of what the typical trade delivers, was negative at every window.

Politicians buy stocks on a regular schedule, the same way any retail investor with automatic deposits into a brokerage account does. When that steady purchasing coincides with a market dip, the subsequent recovery makes them look prescient. Same people, same stocks, different market conditions: the pattern explains the variance in results better than any skill hypothesis.

Consider where the outperformance should appear if genuine information advantage were driving it. Energy trades from Energy and Commerce Committee members. Defense picks from Armed Services. Healthcare positions from HELP. That sector-specific clustering is the signature of insider knowledge, and it is absent from this dataset. The excess concentrates instead in mega-cap technology stocks that every institutional and retail participant on the planet watches simultaneously, which points away from privileged access and toward something more familiar: momentum-chasing behavior indistinguishable from what any retail brokerage dataset shows.

What the disclosure data is for

Congressional trading data serves the purpose it was designed for: public accountability. Journalists and watchdog organizations use it to surface conflicts of interest between committee assignments and personal holdings, and to flag trades that align suspiciously with upcoming legislation. Any voter can look up what their representatives own. As a transparency mechanism, the system works.

The academic literature on congressional trading tells a story that has shifted over time. Ziobrowski and colleagues found in 2011 that House members generated abnormal returns, but that research covered 1985 through 2001 - a period when disclosures were weaker, filings weren't digitized, and the information took months to reach the public. Since the disclosure law modernized in 2012, several factors have likely narrowed the gap between what politicians know and what the market can see: faster electronic filing, algorithmic scraping of disclosures within hours of publication, and a social media ecosystem that broadcasts every filing to millions of followers.

This analysis has a scope limitation worth stating. Every strategy we tested evaluated buy-side activity - what politicians purchased. We did not analyze sell timing, which means the strongest version of the insider-knowledge argument goes partially unexamined: that politicians' real edge lies in avoiding losses by selling before bad news, the way several senators did ahead of the March 2020 COVID crash. A full sell-side analysis on this dataset is a separate project.

We should also name what would change our conclusion on the buy side. If future data showed consistent outperformance concentrated in the specific sectors each politician regulates - energy trades from Energy Committee members, defense trades from Armed Services, pharmaceutical trades from HELP - that would constitute evidence of information-driven returns we did not find in this dataset. If the dropout problem resolved and we could track the same skilled traders year after year, building a genuine multi-year track record, the individual filter strategy might work. Neither condition held across the six years of data we examined.

Every strategy we tested surfaced patterns that were statistically real and practically useless. The timing advantage is too small to trade on. Individual outperformers vanish without warning. The consensus signal depends on five stocks in one sector during one type of market. Across roughly 30,000 decisions by 173 politicians over six years, the conclusion is not that the data is empty. The conclusion is that elected officials buy stocks the same way their constituents do.