Insider Trading - no data

It appears that Insider purchase and sales data are not available.
The stock factors: Ins#ShrPurch and Ins#ShrSold are always NA.

I would like to use them in rankings.

We should have all except InsOwnerSh%, which still needs to be rebuilt. Please note that data is only available since around 2002. Could that be the problem?

The issue occurs even in current rankings.

Current rankings that include insider trading result in an extra NA for each insider trading factor.

Also, current rankings using insider trading produce nonsense factors.

When I look at the actual insider trading for particular companies, it does not correlate to the insider trading factors.

Which ranking system? Could you give a more specific example?

Thanks

Hi Marco,

An example of the problem is illustrated by a ranking named:
Test insider 10

It includes a composite ranking formula: Ins#ShrPurch / (Ins#ShrPurch + Ins#ShrSold)

Highest is ranked higher.

Recent insider trading was evaluated using www.finviz.com.
Here are the insider composite ranks for two symbols:
CRAY 98.8
PXLW 79.89

CRAY is all sells ( and option exercises)
PXLW is all buys

The insider composite ranks appear to be random.

When I was testing this last weekend, rankings that included insider buys and sells had 2 extra NAs - so I concluded that the data was not actually in the database for any securities.
I haven’t seen that problem again - no extra NAs - but no explanation as to what has changed.

Thanks in advance for your help.
Rick

I think it’s misleading because Ins#ShrSold is reported as -ve. I can either update the documentation or flip the sign. I should probably flip the sign for good since it’s counter-intuitive. In the meantime try thhis

Ins#ShrPurch / (Ins#ShrPurch - Ins#ShrSold)

Thanks for your fast reply, Marco.

I agree with your conclusion: I think changing the sign would be best.

Hi, I encountered this thread while trying to figure out what was going on with Ins#ShrSold field so am bumping because it might be the same issue. I think the Ins#ShrSold is still negative in the database system in some places that feed the backtester, but apparently shows as positive when utilized in screens and screen reports.

When I view the Ins#ShrSold field on screen reports it shows as a positive, and when I run screens I have to screen for a positive Ins#ShrSold if I want results for insider selling. However if I run a backtest on that screen with positive Ins#ShrSold I will get no results, and must change the sign of the rule to search for negative Ins#ShrSold for the backtester to produce results.

Screen shots are attached, but I think it’s the same issue being discussed above. It’s very confusing that the screen and the backtest utilize a different sign for Insider Shares sold. Makes it confusing as to what my formula is actually doing in the ranker also when working with insider fields.






Hi, bumping this. I want to use the field, but would like to get a confirmation from p123 that it’s functioning as intended and isn’t having a problem that will be corrected in future. If something is to be corrected I don’t want to embed in a system.

  • I can run a screen on a condition Ins#ShrSold>0 and get passing companies, but when backtested this produces no transactions and no backtest results (flat backtest as shown above).
  • If I change the condition to Ins#ShrSold<0 I will not get any passing companies on a screen, but when I backtest I will get the transactions .

I guess ultimately I want to use this field as it seems to have predictive power, but wondering if it’s something that should be changed prior putting in any ranking system, screen, or backtest? If it’s static and going to stay the way it is that’s fine (negative in the backtester, but positive in the screener, not sure about the sign used in the ranker), but if possible I’d like to confirm it’s not something in need of change prior to utilizing it. Thanks,

While we’re on this topic, Insider Ownership data is missing or flawed. Can’t run any search using that field, hope it can be corrected soon. Thanks.

mm123, I was just randomly checking today, and I think the issue discussed above has been addressed. The Shares Sold for Insider Activity is now uniformly negative in both the backtester and in the screener.

The issue you discussed in your thread (link below) about Insider Ownership % being NA is still an issue from what I can tell though. Maybe data no longer exists?
https://www.portfolio123.com/mvnforum/viewthread_thread,10589

Spaceman, you are right. The date starting on 6/5/2004 through today has -ve numbers for Ins#ShrSold . This is how Compustat enters it (we never did flip the sign). Compustat insider data starts on this date, with around 1,000 companies. By the end of 2004 the coverage reaches the plateau of around 4000.

The +ve numbers you are seeing are from 5/29/2004 and before. These is the old Reuters data that should have been cleared out. I will clear it out today.

I’ll also make the documentation clearer by grouping the factors (shares totals vs transactions totals) and changing starting date to end of 2004.

Recreating the InsOwnerSh% is one of those projects that has been on the todo list for a too long and is always getting bumped. I will try to get it done this coming week.

Thanks Marco for looking at that. From what I can tell, there’s some pretty useful data in there and some of it even works surprisingly well on large caps. (for example I have a note that for the SP500, in a weekly rolling backtest since 2005 with 12mo hold, the top 10% of insider buying as % of mktcap on average outperforms the equal-weight index RSP by 3.2% annually. That’s huge for a single simple large cap factor that’s likely uncorrelated, and it still works well even in recent years). I think the insider holding % will be helpful to scale the magnitude of insider buying or selling activity, and I think will be better ratio to use than scaling by something like mktcap. Just a guess though.

When I was checking last night per my comment I thought it was fixed, but I started w/ 2005 like I was with institutional data. I didn’t realize the nature of the sign flip at a point in time in insider data, but I guess something was really messing with my studies because my notes have confusion about what I was seeing at different points in time. Because we just see ranks instead of the underlying values sometimes it’s hard to tell why things are behaving strangely. The other part is insider activity is intermittent, my guess likely due to SEC regulations on when they can buy/sell, so there are periods where lots of companies might pass a test, and then few, and then a whole bunch again. It makes the percentiles behave strangely so very loose fit rules seem to work better I think.

It also seems to me insider activity might be a longer term signal, as frequent rebalances based on the signal don’t seem to fare as well as longer term. While all starting months w/ longer term hold seem to have positive alpha with longer holding periods, it does seem there’s variation in performance by month depending on when rebalances are setup (Around April, and late Nov/early Dec seem “better”), and my guess is the differences might be real rather tied to noise and tied to the intermittent nature of activity.

Thanks again for looking at this. I think there’s useful data in here.

Remember that insider trading data on the site is aggregated over the prior six months. It’s useful for trend analysis, but not so much for buy and sell signals based on sizable insider activity. (Of course, there’s a separate question of whether one can use those signals at all anymore, now that sales are more associated with executive compensation and them liquidating stock to buy houses and stuff, but I digress.)

I’m finding there’s still useful signal in there, moreso for smaller companies than large, but even meaningful like the very simple study mentioned above for SP500. It would be nice if it was a variable that could be lagged to create a 12m moving total as well as the default 6mo. That would help smooth some of the discontinuity in the data, but I’ll take what I can get.

Regarding stock compensation and selling: My pet hypothesis: higher levels of insider selling might be a signal of a company executing well and executives making bonuses and selling some of that bonus is a sign of companies executing well on stated goals - resulting in stock likely to perform better going forward. It seems more often the case that insider selling is correlated with better future returns. It’s not a strong signal in most cases, but I think it’s likely a net positive impact for future return.

Would really like to see InsOwnerSh% field be useable. Making any progress?

Ins#ShrPurch data routinely flags exec awards as “purchases”. This is highly misleading, please fix

fedcep- do you have an example ? We are just taking what S&P gives us defined as “The number of shares bought by insiders over the
past 6 month.”

NOTE: We need to add the fact that we are using the 6 month totals from S&P in our documentation. Maybe that’s the source of the confusion?

Hi Marco,

While on this subject can you tell us about the status of InsOwnerSh%? That field has no data.

Thanks,
Ed

Thanks Marco. Example: ticker ZIXI. The following transactions are marked as buys when they were stock awards. Much different signal – the latter is total comp, the former is a clear bullish indicator: ZIXI Stock Quotes and Insider News | Insider Tracking

Edit: FWIW the six-month data sample didn’t generate confusion on my side, it’s explained in the datapoint description at p123