Introduction

In 2012, Congress passed the STOCK act (Stop Trading On Congressional Knowledge), a bill designed to combat insider trading. Using congressional knowledge for private gain is problematic because often times, insider trading impacts a Senator’s ability to serve the public.

While the STOCK act clarified rules by which Members of Congress must abide by in their investments, it is far from a panacea. By nature of their work, Members are briefed on matters of national security, regulation, and the economy, giving them access to nonpublic information that could benefit them in the stock market. Most recently, in February 2020, Richard Burr sold around $1.72 million in personal holdings following a Senate Intelligence Committee briefing regarding COVID-19, all the while downplaying the virus’s economic impact to the public. 1

Thankfully, the STOCK act requires every Member of Congress to file and publicly disclose any financial transaction of stocks, bond, commodities futures and other securities within 45 days. In addition, Members will also file annual disclosures describing their annual incomes and held assets. Over the last month, I have been interning for FinePrint Data, a firm that parses and centralizes (often handwritten/scanned) disclosure forms into machine readable formats for further analysis. The dataset I am working with is one of FinePrint’s core products – a CSV file containing senate disclosures from present through January 2014.

Disclosures from the annual disclosures dataset can answer questions about the wealth and trading habits of senators. Specifically, annual disclosures on held assets help us gauge a particular senator’s net worth. Aggregating this information across party or Senate allows us to glean the wealth of Democrats, Republicans, or Senators in general. We can then compare parties to one another, or Senators to the average population.

Investigating disclosures of specific Senators can also help uncover possible instances of insider trading. Senator David Perdue, for instance, is notorious for being the most active trader in Senate, having conducted 2,596 transactions in one year. His most profitable trades this year have involved Cardlytics (CDLX), a company for whom he served on the board before joining the Senate. In 2020, Perdue sold Cardlytics at $86 a share, following a private conversation with then CEO Scott Grimes. Weeks later he bought Cardlytics back at $30 a share. 2 Since then, his stock has more than quadrupled in value. Are Perdue’s trades savvy investment decision? Or evidence of insider trading?

In this project, I will attempt to answer both general and specific questions. First, I will visualize the wealth of congress using static plots and provide my own analysis using the annual disclosures data set. Then, I will investigate the trading habits of senators, including their trade frequency and volume, using the periodic disclosures data set. Finally, I will demonstrate how trade annotations on ticker time series can be used to gauge profitability of trades and potentially detect insider trading.

Exploratory Data Analysis

1. Investigating The Wealth of Senate

With annual disclosures, we can attempt to pinpoint the net worth of individual Senators by aggregating the values of their assets. We can also make observations about the wealth of Senate as a whole. By grouping the annual disclosures data set by senator name, we can aggregate for the sum of their asset values. Problematically, congressional disclosures do not specify the exact value of each asset held. Instead, senators file the values of their transactions as categorical bins (ex. between $15,000 & $50,000). Thus, it would be wise to consider three metrics when estimating the wealth of congress - aggregation of minimum values, maximum values and mean.

After grouping by senator and engineering the features minimum, maximum and estimated net worth, we can look at the medians of these newly created columns to arrive at a lowball, highball and middleground estimation of the net worth of “the average senator”. In this case, median is a much more descriptive statistic compared to mean, since the wealth inequality in senate is substantial (I will demonstrate this in subsequent paragraphs).

In order to not double count asset disclosures, I’ve sampled only disclosures with 2019 as the filing date (the latest available annual disclosures).

## [1] "Median Senate Net Worth (Estimated):  3822590.5"
## [1] "Median Senate Net Worth (Lowball):  1723110.5"
## [1] "Median Senate Net Worth (Highball):  5493103"
## [1] "% of Millionaires in Senate (Estimated):  0.7"
## [1] "% of Millionaires in Senate (Lowball):  0.555555555555556"
## [1] "% of Millionaires in Senate (Highball):  0.811111111111111"

It’s worth noting that even the lowball median net worth of the average senator is $1,723,110 meaning that at the very least, the average senator is a millionaire. This aligns with our estimate of %millionaries, which indicates that more than half of senate has at least a million dollars in assets.

The average age of senators is ~62. According to the U.S. Census Bureau, the median net worth for homeowners age 65 and older is $201,500. 3 This means that the average senator is, at the very least, ~8 times wealthier than the average American in their age group.

The wealthiest senators are fabulously wealthy. In particular, Rick Scott and Mike Lee have 9 digit net worths (with Romney and Warner following close behind).

4 Wealthiest Senators 2019

In the visual above, each dot represents the net worth of a senator. Three facets of the jitter plot represent middle, high and low estimations of senator net worth. Color represents party. Finally, the purple line annotation is the $200,000 mark and the black line annotation is the $1,000,000 mark.

I had to remove the four richest senators from the graphic because their wealth vastly outstripped that of their colleagues, and caused the visual to become illegible. Nevertheless, we observe stark wealth inequalities in senate. In all three facets, a substantial number of senators had a net worth between the $200,000 and $1,000,000 mark. A majority of senators in all three graphics had a net worth exceeding $1,000,000. Only a handful of senators were worth $200,000 or less.

In the class of senators with net worth exceeding $1,000,000, our highball and middleground estimates indicate wild disparities in wealth. However, this may be because transactions exceeding a certain magnitude begin losing precision. Categories describing small transactions have small margins of error (ex. transactions $0 to $15000 or $15000 have up to 5 digits of imprecision). Categories describing large transactions are comparatively much more imprecise (ex. $2.5M to $5M have up to 7 digits of imprecision).

Our lowball jitter plot indicates that no senators (besides the excluded four) is worth more than $4 million. Our highball and middleground estimates indicate that very few, if any senators are worth more than $8 million.

Our jitter plot indicates that extremely wealthy senators are more likely to be Republican than Democrat. We speculate that Republican senators may also be wealthier than Democrats in general.

## [1] "Democrat Minus Republican Median Net Worth (Lowball) -539003"
## [1] "Democrat Minus Republican Median Net Worth (Highball) -1633594"

By taking the difference between the median net worths of the two parties, we find that the wealth gap between the average Democrat and Republican senator. This value ranges anywhere from $539,003 to $1,633,594; meaning, the average Republican is likely to be half a million to more than a million dollars wealthier compared to the average Democrat.

We can visualize the wealth gap between parties by using a Circle Packing Visualization:

Net Worth Circle Packed By Party

Unsurprisingly, the Republican cluster appears much larger than the Democrat cluster. We observe that in both parties, severe inequalities exist, but excepting Mark Warner, Democrats are relatively similar in net worth. We find that while Republicans come from a variety of financial backgrounds, there are more wealthy Republicans than wealthy Democrats.

2. Are Senators More Active Traders than Average People?

Having investigated the wealth of Senate using annual disclosures, we can now look at their trading habits using periodic disclosures. As with the annual disclosures, I’ve subsetted the senate periodic data to include only transactions conducted in 2020.

Subsetting the data in this manner will allow us to draw conclusions about the annual trading habits of Senators. I’ve opted to select transactions from 2020 rather than 2019 because 2020’s COVID-19 selloff makes it a more interesting year for investigating senate disclosures.

## [1] "Mean # Transactions:  107.2"
## [1] "Median # Transactions:  13"
## [1] "Mean Volume:  14797639.1666667"
## [1] "Median Volume (Estimated):  451516.25"

Exploratory Data Analysis reveals that the median number of trades conducted in senate is ~10 per year, and the median trading volume in a year is 408,006. We notice a substantial difference between the mean and the median, suggesting severe asymmetries in trading patterns.

We arrive at these statistics by grouping 2019 disclosures by name, aggregating row count to arrive at transaction count and summing the mean of the min/max values of each transaction to arrive at an estimated total trading volume. We can use a scatter plot to visualize our findings.

We notice that Kelly Loeffler is an extreme outlier in trading volume and David Perdue is an extreme outlier in transaction count. In particular, we notice that Perdue traded 823 times in 2020, and Loeffler traded 730 times. Moreover, Loeffler traded more than $180 million dollars in total volume in 2020.

Loeffler’s outlier status may come as surprise, given that she did not appear at all in annual disclosures. This is due to the fact that Loeffler was only elected to senate on January 6th of 2020, while the annual disclosures analysis looked at disclosures in 2019. In 2020, Loeffler was the richest Senator, with an estimated net worth of ~$800M.

Does Party Affect Trading Habits?

We can also investigate the trading habits of Republicans and Democrats by subsetting the data across party lines.

## [1] "Median Democrat Transaction Count: 10"
## [1] "Median Republican Transaction Count: 30"
## [1] "Median Democrat Trading Volume ($): 396511"
## [1] "Median Republican Trading Volume ($): 600004"

We find that the average Republican Senator trades ~$200,000 in volume more compared to Democrats, and executes three times the trades.

We find that at lower quartiles, trading behavior does not differ much across party lines. However, active Republican traders in Senate are likely to trade more frequently than active Democrats. This explains the significant difference medians for both frequency and volume between parties.

3. Examining Individual Stocks

One of the more interesting visualizations we can perform with transaction data is annotating a time series with buy/sells of senator(s). These annotations allows us to gauge the profitability of each individual trade, and, coupled with domain specific knowledge, allows us to investigate possible violations of the STOCK act.

I’ve developed an R Shiny app that generates these visualizations for any ticker. The link is here: https://dzhen1.shinyapps.io/finalproject/

We can start by investigating Perdue’s dealings with Cardlytics, a data marketing firm whose board he was once a part of.

We find that Perdue sold large portions of CDLX on January 23rd. Then, in April, Perdue repurchases CDLX at its bottom. Since then, his position has risen over 5 times in value. Perdue’s prudent trades have come under scrutiny recently as it has been revealed that he received a private tip over the phone from Cardlytics’ CEO Scott Grimes about a change in leadership two days prior to his sale. This means that Perdue could have been engaging in insider trading.

Another interesting subject of investigation is the COVID-19 selloff involving Richard Burr, Kelly Loeffler and David Perdue.

In this paper, we’ll examined two of Loeffler’s stocks. The first ticker - DuPont - is a medical supplier; the second ticker - booking.com - is a e-booking travel agency.

By examining our time series, we learn that in the months leading up to the pandemic, Loeffler bought large portions of DD and dumped BKNG. This gives us ground to suspect that, as a Senator, her knowledge of the severity of the impending pandemic may have impacted her investment decisions.

These visualizations do not, by themselves prove insider trading on congressional knowledge. However, visualizing the transactions on each ticker can allow for more compelling and faster investigations into congressional insider trading.

Conclusion

In this paper, I conducted exploratory data analysis on two data sets: senate annual disclosures and periodical disclosures. I found that the average Senator is more than 8 times wealthier than the average American in the same age group. Furthermore, I found that the average Republican Senator is worth anywhere from half a million to a million dollars less than the average Democrat Senator.

In analyzing the periodical transactions data set, I found that the average Republican Senator trades 3 times as frequently as the average Democrat Senator and trades $200,000 more in annual volume.

One conclusion we can draw from this data set is that Republican senators are wealthier and pay more attention to their assets when compared to their Democrat counterparts. This seems like an intuitive hypothesis, since the Republican party generally promotes fiscal conservatism and freedom of market.

However, since only one year of congressional data was examined, it would be unwise to make the extrapolation that Republican Senators are wealthier and trade more than Democrat Senators. Given more time, I would conduct a similar analysis on data dating back to 2014, and perhaps examine house data as well.

Furthermore, while the transaction annotations on ticker time series is an intuitive visualization, it is not in it of itself powerful enough to detect cases of insider trading on congressional knowledge. It is all but impossible to distinguish between a prudent trade and a problematic trade without additional information.

As a visualization tool, my R Shiny app is limited insofar as it is a purely exploratory tool. Thus, some next steps for this app could include automated detections of abnormally profitable trades, which would help political accountability groups streamline detection of insider trading.

  1. Lipton, E., 2021. Senator Richard Burr Sold a Fortune in Stocks as G.O.P. Played Down Coronavirus Threat. [online] Nytimes.com. Available at: https://www.nytimes.com/2020/03/19/us/politics/richard-burr-stocks-sold-coronavirus.html [Accessed 30 January 2021].↩︎

  2. Jacobs, J., 2021. Deep Dive on Sen. Perdue and Cardlytics (CDLX). [online] Fineprintdata.com. Available at: https://www.fineprintdata.com/post/perduedeepdive [Accessed 30 January 2021].↩︎

  3. GoodLife. 2021. Average American Net Worth at Retirement in 2019 | GoodLife. [online] Available at: https://goodlifehomeloans.com/average-american-net-worth-at-retirement [Accessed 30 January 2021].↩︎