Author Details

Vipul Jain

Vipul Jain
Vice President, Equity Factor Research

Roman Kouzmenko

Roman Kouzmenko
Executive Director, MSCI Core Equity Research

Social Sharing

Extended Viewer

Did insider transactions have secrets to tell?

Did insider transactions have secrets to tell?

  • Among alternative data sources we examined, the insider-transaction dataset stood out with unique characteristics when compared to traditional risk-model style factors.
  • Our backtest of data on insider transactions showed it had positive factor, quintile spread returns and information ratios.
  • When we considered only transactions done by key company executives, excluding those related to employee compensation, 10b5-1 plans and tax purposes, it further improved returns and information ratios.

Given that company insiders1 may be in possession of material, nonpublic information, their trading is usually subject to strict rules, including regulations in most countries requiring trades be disclosed publicly. By observing how insiders traded their company’s stock, we were able to extract more information about expected future company performance than we were from traditional sources.

 

Defining metrics for insider sentiment

In our analysis, we used insider-transaction information available across 50 countries through mandatory regulatory filings. The dataset’s coverage spans a little more than 15 years beginning in 2003 and, as of the most recent period, over 12,000 companies were included globally with approximately 4,150 in the MSCI ACWI IMI and 1,850 in the MSCI USA IMI. Japan is a notable exclusion due to a lack of regulations required to report corporate insider transactions. Other notable exclusions from the MSCI ACWI Index universe are Taiwan, Saudi Arabia, Mexico, Argentina, Colombia, Qatar and the United Arab Emirates. To evaluate the potential explanatory power of insider transactions on stock returns, we explored the data from count, volume and depth of insider trades. Detailed definitions are provided below.

 

Metrics for insider sentiment

Metric Definition
Trade count Difference of total “buy” count and “sell” count over the prior three months divided by total “buy” and “sell” count during the same period.
Trade volume Difference of total “buy” shares and “sell” shares over the prior three months divided by total “buy” and “sell” shares in same period.
Trade depth Difference of total “buy” U.S. dollar amount and “sell” U.S. dollar amount over the prior three months divided by the full security-market capitalization in U.S. dollars.

 

Insider-sentiment factors performed well historically

For this analysis, we defined the insider-sentiment factor as an equal-weighted combination of the count, volume and depth of insider-trade metrics, as defined above. We looked at the cross section of returns in univariate and multivariate settings. For the univariate analysis, we ranked stocks at each month’s end by the insider-sentiment factor, grouped them into quintiles and looked at the return spread between the top and bottom quintiles for the subsequent month. The multivariate results showed the unique potential contribution of a given factor over and above all other risk-model factors.

We tested two different formulations of the insider-sentiment factor:

  • “All transactions,” where metrics were calculated based on all insider transactions
  • “Filtered transactions,” where we focused on transactions by key company executives2 only and excluded transactions linked to employee compensation plans, automated investments from 10b5-1 plans and trades executed for tax purposes, to see if that focus improved the factor’s explanatory power

The two exhibits below show the resulting performance characteristics of these two formulations.

 

Performance characteristics of insider-sentiment factors

  Quintile portfolio statistics Multivariate regression statistics Stability Information
coefficient
Metric Quintile
spread
return, %
Quintile
spread
volatility,
%
Quintile
spread IR
Factor
return, %
Factor
volatility,
%
Factor IR Mean |t| |t| > 2, % Exposure
auto-
correlation
Rank
IC (%)
All
transactions
3.27 7.52 0.43 1.26 0.65 1.93 1.36 25.81 0.84 -0.16
Filtered
transactions
5.93 9.26 0.64 1.97 0.77 2.54 1.62 30.97 0.87 0.54

Due to lower global coverage of dataset between 2003 and 2005, factors are backtested starting from June 2006 to April 2019. Rank ICs are defined as the rank correlation of the month-end signal exposures with security returns in the subsequent month. The multivariate statistics are the results of weekly cross-sectional regressions in the Global Total Market Trading Equity Model. Stability is defined as average month-to-month exposure autocorrelations. The investment universe is the MSCI ACWI IMI. Source: MSCI, 2iQ Research

 

Insider-sentiment factors had positive returns

Insider-sentiment factors had positive returns

Source: MSCI, 2iQ Research

While the t-stats and volatilities were relatively low (but consistent with the levels historically observed for alternative data sets), both formulations had consistent, positive factor returns throughout our analysis period, quintile spread returns and information ratios. Such characteristics have been more typical of alpha measures than traditional risk factors, such as size and beta, and establish the potential of the insider-transaction dataset beyond traditional factors. The stability of this dataset was generally moderate: in the 0.70 to 0.90 range that we label “medium-term.”

 

Insider-sentiment factors had low correlations with traditional factors

The exhibit below shows cross-sectional correlations of filtered transactions’ insider-sentiment factor with the style-factor exposures of MSCI's Barra Global Total Market Equity Trading model. Except for the momentum factor, which had a moderate level of correlation with insider transactions, most of the correlations — including those with other sentiment measures such as analyst sentiment (earnings estimate revisions) and short interest — were near zero. The correlation with momentum could potentially highlight the nature of insider transactions; i.e., insiders would like to sell more when the stock does well and would like to buy more when the stock is cheap.

 

Average monthly exposure correlations of insider-sentiment factors with style factors

Average monthly exposure correlations of insider-sentiment factors with style factors

Due to lower global coverage of the dataset between 2003 and 2005, factors are backtested starting from June 2006 to April 2019. Average of monthly cross-sectional correlation of insider sentiment factors with the style factor exposures of the Barra Global Total Market Equity Trading model. The investment universe is the MSCI ACWI IMI. Source: MSCI, 2iQ Research

The insider-transactions dataset, with its various transaction-level filters, could be a potential avenue to construct factors that seek positive risk-adjusted returns after accounting for traditional factors.

 

 

1Directors, officers or large stakeholders of a company required to disclose their transactions publicly.

2Key company executives include board chair, executive board, supervisory board, executive committee, board of directors and senior nonexecutives.

 

 

Further Reading

More than a feeling: Quantifying consumer sentiment

Peering into peer selection: Quantifying company similarity

Should we be surprised by earnings surprises?

Regulation