Copycat Portfolios – Do political trades pay off?

Introduction

In 1999, a six-year-old trader became the 22nd most successful money manager in the US, beating more than 6,000 professional Wall Street brokers. This trader threw darts at a list of stocks and picked those where the darts landed. Oh and by the way, that six-year-old trader was a chimpanzee [1]. That's how bad most people, even professionals, are at stock picking. Beating the market consistently is hard. Information is noisy, emotions get in the way, and even sophisticated strategies can crumble and underperform a simple index fund. But wait a minute, don't go hiding your money under your mattress just yet.

What if there was a group of traders that both performed well, but was also legally required to share with you everything they were buying and selling ? Sounds too good to be true, right ?

Enter the United States Congress.

Members of Congress, both Representatives and Senators, sit in a uniquely powerful position. They write laws, regulate industries, oversee federal agencies, and make decisions that can reshape entire sectors of the economy, and many of them, while doing so, invest in the stock market. If you're thinking that doesn't sound right, you are not alone. For a long time, this raised an obvious concern, should lawmakers be allowed to trade stocks while detaining information the rest of us don't have access to ?

In 2012, Congress addressed this question by introducing the STOCK Act (Stop Trading on Congressional Knowledge) [2]. This law requires members of Congress, their partners, as well as their children, to publicly disclose their stock trades within 45 days of execution. Every single purchase, sale, detailed, time-stamped, and publicly available. Are you thinking what we are thinking ?

What if, instead of competing against hedge funds, algorithms, and dart-throwing primates, we simply copied the trades of US politicians ?

This project dives into that question. Using disclosed congressional trading data, we will analyze the performance of these lawmakers, compare them to the broader market, and explore whether a systematic copycat strategy could turn political transparency into investment edge.

The Basics

A Quick Stock Market Refresher

Before jumping into the data, let's take a quick step back and go over some stock market basics. At its core, the stock market is just a marketplace. Companies can sell ownership, called stock, sliced into shares, to investors, who buy and sell those shares, in the hope that they can profit from companies growing. If the company you are invested in grows, your shares become more valuable. If it doesn't... well, let's just say not every trade ages gracefully.

A trade is simply the act of buying or selling a certain amount of shares, for a certain price, of a certain company. For every trade you make, there is a person buying or selling you shares. In the market, prices change depending on this supply and demand relationship, because new information arrives, expectations change, or simply because suddenly a lot of people decide they like or hate a particular stock.

The Exchange: NASDAQ

Stocks are traded on particular marketplaces called exchanges. There are many exchanges, but for this project, we will focus on the NASDAQ. The NASDAQ is one of the biggest exchanges in the world and is especially known for technology and growth-oriented companies like Apple, Microsoft, Nvidia, and more than 4,000 others. All of the individual stock price data used in this project comes from the NASDAQ stock prices dataset, which gives us a deep and rich source to work with [5].

Beating what exactly ? Our benchmark

Any investment can look impressive on its own. The real question is: impressive compared to what? To answer this, we need a benchmark, a reference that represents how the overall market performs.

We use the S&P 500 as that benchmark. The S&P 500 is an index, meaning it is not something you can buy directly, but a measure that tracks the performance of the 500 largest publicly traded companies in the United States. It is commonly used as a proxy for “the market”. Since an index itself cannot be invested in, we use SPY, an index fund that closely tracks the S&P 500 by holding the same companies in similar proportions. Investing in SPY is therefore roughly equivalent to investing in the U.S. stock market as a whole.

Throughout this project, when we talk about “beating the market”, we mean achieving better performance than SPY over the same period.

With the basics out of way, we can finally dive into the fun part: the data.

The Data

At the heart of this project are the congressional stock transaction disclosures, made public under the STOCK Act. Our dataset contains all disclosed trades, from 2012 to early 2024, covering both Representatives and Senators. Each record tells us who traded, what was traded, when they traded and disclosed it, and whether it was a buy or a sell.

A few limitations of these disclosures that are worth keeping in mind:

• Stocks only: We will not look at options, futures, and other derivatives, due to a lack of precise pricing details in the disclosures.
• Trade values are in ranges: Total values of trades are usually disclosed in ranges, not exact amounts, so we will use the midpoint as an approximation.
• Only currently listed stocks: Delisted securities will be excluded in the analysis, since NASDAQ doesn't provide historical price data on them.
• We don't know politicians' prior investments: We assume that they start their term with an empty portfolio and take the convention that they can only hold a positive number of shares. In particular, if they sell shares from prior investments, resulting in them selling more shares of a stock than they theoretically hold, we simply consider that they have sold all their shares and remove that stock from their portfolio, without accounting for that difference.

The cleaned dataset we will be working on is fully exporable below. If you are curious about what your favourite congressman is buying, or who is buying your favourite stock, feel free to dig into the data yourself.

So this is great, but what can we actually do with these transactions ?

Portfolios

Using the disclosed transactions, we rebuild each politician's portfolio, day by day, adding shares when they buy, removing them when they sell, and letting market prices do the rest. We obtain the market prices data from our NASDAQ dataset and from there, calculating returns is simply a matter of tracking how the portfolio evolves.

Curious to try it yourself? The mini-game below lets you recreate a $500 copycat portfolio, buy or sell the highlighted stocks, and fast-forward the disclosure timeline to see how your returns evolve alongside the market.

Once the portfolios are built, we can finally start comparing performances. And that's where things get interesting.

How well do politicians perform ?

Before trying to copy anyone, we need to answer a simpler question: are politicians actually good investors? Academic work has long asked the same question, notably highlighting abnormal returns among members of the U.S. Senate [4].

Now that we can reconstruct what each politician holds over time, we can step back and look at the big picture. Instead of focusing on individual success stories, we start by asking how politicians perform on average, as a group, compared to the market.

But what could be considered as a good indicator of performance?

Actually, there is no single right answer, it really depends on what you want to favor... High returns? Consistent stable returns? Daily returns higher than the S&P 500?

Throughout our analysis, we will use the Information ratio [3]. In simple terms, it gives each portfolio a grade based on its daily returns compared to the S&P 500. A unique property of this indicator is that it rewards portfolios that consistently beat the S&P 500 without deviating too much from its trend and penalizes those whose returns fluctuate a lot compared to the S&P500, favoring low-risk diversified portfolios. When the information ratio approaches zero, our performance is essentially in line with the market. A positive information ratio indicates outperformance, while a negative one signals underperformance relative to the market.

The yearly distributions below show these Information Ratios across all politicians.

The result is fairly sobering.

Most politicians cluster tightly around zero, suggesting that as a group, they do not meaningfully outperform the market. A few do quite well, a few do very poorly, but on average there’s no obvious edge to exploit here. In other words: simply copying any politician at random is unlikely to get you very far.

But averages can hide important details. So next, we break the data down further.

How well do subgroups of politicians perform?

A natural next question is whether certain groups of politicians trade better than others. We start with one of the most obvious distinctions: institutional role.

Senators VS House members

Do Senators trade better than House members — or the other way around?

To get a first intuition, we compare their performances using the distribution of Information Ratios shown in the boxplot below.

At first glance, the two groups look… almost identical. The spreads, medians, and overall shapes of the distributions are remarkably similar, suggesting no clear advantage for either chamber.

To confirm this impression, we will run a statistical test comparing average performance between Senators and House members. But before doing so, as we will run several statistical tests throughout our journey with you, a small reminder of what it actually means cannot hurt, even though you are free to skip it if you are just here for the results!

Basics of statistical tests

Basically a statistical test works as follows. You state a null hypothesis (H0) that you would like to reject, for example in our case, as we aim to find a difference between groups, our null hypothesis could be : "(H0) Both groups essentially have the same average Information Ratio, the difference obtained is just meaningless variance". Then, your goal is to find one single value, called a statistic or in most cases a t-stat, summarizing all the observations you have and aiming to capture most meaningful properties of their distribution. Under the null hypothesis, the probability distribution of this value must be known as the test aims to answer the following question: "If your null hypothesis is true, what was the probability of obtaining a sample that gives you a t-stat at least as extreme as this one?". That probability returned by the test is called the p-value and if it is smaller than 0.05, we usually say that we reject our null hypothesis at the 5% significance level (if it is true, there was less than a 5% chance of getting a statistic at least as extreme as we obtained). It doesn't mean that the null hypothesis is incorrect, it simply means that we have statistical evidence against it.

Now I know what you're thinking... Building such a test seems like complicated and annoying math stuff, right? Well don't worry, some kind people already created all the tools we need to perform such tests for many different hypotheses and many different sample distributions. We simply have to choose the right test for each specific situation and most importantly, not misinterpret our results : even if it is really tempting to see it like this, the probability returned by the test is not the probability that our null hypothesis is true! If you were among the 95% of people that thought it was, we encourage you to read the question the test aims to answer again before diving into our statistical tests...

Back to our analysis

In our situation, we can use a statistical test called a Welch t-test with the same hypothesis as in the example above : "(H0) Both groups essentially have the same average Information Ratio, the difference obtained is just meaningless variance". This test gives us a t-stat of 0.228 and a corresponding p-value of 0.8208. Hence, the conclusion is that we cannot reject our hypothesis as we don't have statistical evidence that the groups perform differently on average.

If role doesn’t matter, maybe ideology does.

Republicans VS Democrats (VS Independents)

To get a first intuition, we compare the performance of Democrats, Republicans, and Independents using the boxplot below.

Something immediately stands out. The Independent group behaves very differently — and for a simple reason: there are only two Independents with enough trading activity to analyze. With such a small sample, any comparison would be unreliable, so we set them aside and focus on the two main parties.

Looking at Democrats versus Republicans, a subtle pattern begins to emerge. On average, Democrats appear to perform slightly better, particularly when looking at the median Information Ratio. The difference isn’t dramatic, but it’s consistent enough to catch the eye.

To understand whether this gap reflects a real difference or just visual noise, we take a closer look at the full distributions.

The broader picture tells a similar story. Democrats show slightly higher Information Ratios on average, while Republicans exhibit a wider spread of outcomes. The gap is modest, but persistent across the distribution.

Still, markets are noisy, and small differences can easily arise by chance. To check whether this pattern holds up, we apply the same Welch t-test as before, giving us a t-stat of 1.398 and a p-value of 0.1637. The result is cautious: while the numbers lean in favor of Democrats, the difference is not statistically significant with the data we have.

In other words, party affiliation alone doesn’t provide a reliable trading edge.

That brings us back to individuals rather than groups. If neither chamber nor party explains performance, perhaps the real story lies with how specific politicians trade, rather than who they represent.

How well do they perform individually ?

So far, politicians as a group haven’t exactly screamed market-beating strategy. But maybe we’ve been asking the wrong question. Instead of looking at averages, let’s zoom in on individuals.

To help with that, we provide two interactive tools. The performance explorer lets you compare a politician’s portfolio returns to the market over time, while the portfolio explorer shows what that politician actually held on a given date, especially useful to quickly see whether a politician's portfolio is diversified or whether they are "all-in on one stock" gamblers.

To identify strong performers, we rank politicians using their Information Ratio, a metric that rewards not just high returns, but consistent outperformance relative to the market. This allows us to identify lawmakers who do well without taking excessive risk. The leaderboard below updates based on the selected time frame, chamber, and party.

At the top of the rankings, two names stand out: Morgan McGarvey for the Democrats and Markwayne Mullin for the Republicans. Interestingly, even though Mullin ends up with a higher Information Ratio, McGarvey's raw returns look more impressive at first glance as you will see below.

Why? The plots will tell the story. Mullin's portfolio closely follows the market, but consistently stays only slightly above the S&P 500. McGarvey's portfolio, on the other hand, swings much more aggressively. Higher highs, but also much higher volatility.

Since the Information Ratio balances performance and stability, this explains why Mullin comes out on top overall. It's a classic trade-off: flashy gains versus consistent outperformance.

To better understand where this difference in volatility comes from, let's take a closer look at what these portfolios actually contain.

The pie charts make the difference crystal clear. Morgan McGarvey's portfolio is almost entirely driven by just three stocks, while Markwayne Mullin's is spread across many different companies. McGarvey's strong performance largely comes from being in the right place at the right time: riding the 2023-2024 surge of Amazon and Nvidia.

Performance Isn't the Whole Story

Picking a couple of winning stocks can pay off — but it’s also fragile. One bad move, one missed rally, and the story can change quickly. However, this doesn't mean that it is a bad strategy, especially if it leads to higher returns.

In conclusion, high returns are nice, but they are even nicer when they come with consistent outperformance of the market. Hence the use of the Information ratio.

With this lens and the tools above, we can now identify a small group of politicians who appear to trade well over time.

Which leads naturally to the next question:

What happens if we copy their trades?

The Copycat

Disclosure Delays: Why Timing Matters

Copying a portfolio is one thing. Copying it on time is another. The main limitation of any copycat strategy is the disclosure delay: the longer the delay, the higher the risk that the opportunity has already passed. That's why we take a close look at how long politicians actually wait before reporting their trades.

The figure below shows the distribution of these delays where we define the delay $\Delta t$ as the number of days between the date of the trade and the date the trade was disclosed on.

Although the law sets a 45 days limit, real-world behavior varies widely. We can see that although most politicians do play by the rules, there is still a non-negligeable portion of trades that are disclosed months, if not years later. Such long delays make many disclosures stale, and this is exactly why taking these disclosure delays in the choice of candidates to copy is important.

Who do we copy ?

Not all politicians are worth copying. So we start by grading them. Each politician gets a score based on two simple criteria: how well they trade and how fast they disclose.

First, we look at past performance using the Information Ratio. In short: are they actually good investors?

Second, we check how late they report their trades. Long disclosure delays mean stale information, and stale signals make bad strategies. Politicians who repeatedly disclose too late are penalized.

We combine both aspects into a single grade and rank politicians accordingly, favoring those who trade well and play by the clock.

The Strategy

Now that we know who we can copy, let's build our investment strategy and see how it performs.

Our copycat strategy is intentionally simple. No leverage, no options, no clever tricks. Just transparency, patience, and repetition.

At the beginning of each year, we look back at the previous four years of disclosed trading activity. Using that history, we rank politicians based on the grading system described above: rewarding strong, consistent performance and penalizing slow disclosures. From this ranking, we select the top N politicians. We then merge their portfolios into a single composite portfolio, giving each politician equal weight. In other words, no one gets special treatment: the best performer doesn’t dominate the portfolio, and diversification comes naturally from copying multiple people at once. Once the portfolio is built, we hold it for one year. During that year, the portfolio isn’t frozen. If the politicians we’re copying make new trades and disclose them in time, we update the portfolio accordingly, just as a real copycat investor would. If politicians copied in the portfolio leave office during the year we copy them, the portfolio is updated using the other politicians in the ranking. At the end of the year, we reset, re-rank everyone using the most recent four-year window, and build a new portfolio for the next year. Then we repeat.

Evaluating the Strategy

To understand how concentration affects performance, let's test this strategy under three different settings:

• Top 7 politicians: more diversified, more stable
• Top 3 politicians: a balance between focus and diversification
• Top 1 politician: highly concentrated, high conviction

All three versions follow the same rules. They start in 2018, are rebalanced once per year, rely on the past four years of data, and use the same grading system to rank politicians. The only thing that changes is how many politicians we copy. This allows us to isolate a simple question:

Is it better to spread our bets across many good traders, or to concentrate on just a few? Let’s see what the data says.

To answer this, we evaluate each strategy along two dimensions: performance and statistical reliability.

First, we look at the Information Ratio, which measures how much excess return a strategy generates relative to the market, adjusted for risk. Higher values indicate more consistent outperformance.

Second, we test whether that outperformance is likely to be real or just noise, by checking whether average excess returns are statistically greater than zero. Since financial returns can be correlated over time, we report results under two assumptions: one assuming independence, and a more conservative one that accounts for autocorrelation using Newey–West adjustments.

The table below summarizes the results. As a rule of thumb, higher Information Ratios are better, and lower p-values indicate stronger evidence that the performance is not due to chance.

A clear pattern emerges.

The Top 3 strategy consistently dominates the trade-off between performance and reliability. It achieves the highest Information Ratio and provides the strongest statistical evidence of outperformance. Under the simpler assumption of independent returns, it crosses the conventional 5% significance threshold. When accounting for autocorrelation, the evidence weakens but remains suggestive.

The takeaway is not that copying politicians is a guaranteed money machine, but that moderate concentration matters. Copying too many politicians dilutes the signal, while copying just one increases volatility. Focusing on a small group of strong, timely traders appears to strike the best balance.

From here on, we focus on the Top 3 copycat strategy.

The Strategy in Action

The chart below shows the cumulative returns of the copycat portfolio (N = 3) compared to SPY, our benchmark for “the market”.

Taken together, the picture is clear. Over this period, copying the trades of a small group of top-performing politicians would have meaningfully outpaced the market. The advantage isn’t fleeting or concentrated in a single moment: it compounds over time.

So, where does that leave us ?

Final Thoughts

So, what did we learn?

First, despite their unique position and access to information, politicians as a group do not magically beat the market. Most perform about as well as you’d expect from any random collection of active investors. No free lunch there.

But zooming in tells a different story.

A small subset of politicians consistently outperforms, and when we focus on those who both trade well and disclose reasonably quickly, a simple copycat strategy starts to emerge. By copying a small group of top performers, not too many, not just one, we find a portfolio that, over time, meaningfully outpaces the market.

Can you actually do it yourself ?

Yes.

Congressional trade disclosures are public by law, and anyone can access them. If you’re curious to explore further, here are a few places where disclosures are published and aggregated:

• U.S. House Financial Disclosures: released by the Clerk of the House [6]
• U.S. Senate Financial Disclosures: published by the Secretary of the Senate [7]

If you were to try implementing a strategy like this in practice, a few lessons from our analysis are worth keeping in mind:

• Disclosure delays matter, late reports are far less useful.
• Diversification still matters, even when copying “smart money”.
• This is slow investing, not day trading.
• And most importantly: past performance is not a guarantee of future results.

This project isn’t an endorsement to blindly follow politicians’ trades. It’s an exploration of what happens when forced transparency meets financial markets, and how publicly available data can be turned into insight.

In a world where beating the market is famously hard, sometimes even for professionals, it turns out that one of the most interesting signals was hiding in plain sight, buried in PDFs no one wanted to read.

Until now.

Whether you use this as inspiration, curiosity, or cautionary tale, one thing is certain: data has a way of changing how we see power, markets, and the stories we tell about both.

And sometimes, it even beats a monkey with darts.

References

Team

Meet the copycat crew

Evan Ayoub

Loan Bianchi

Adrien Bousquié

Jérôme Courdacy

Majandra Garcia