13F, macro and flows: from institutional holdings to alpha signals

The 13F universe in numbers

Every institutional investment manager with at least $100 million in US-listed equities and other "13(f) securities" must file Form 13F with the SEC within 45 days of each quarter end. FinzData parses every filing since 2013 into one table of holdings, one row per manager, security and quarter. For the 2026Q2 quarter, filed by mid-August 2026:

Since 2013Q2 the number of filers has grown 2.6-fold, from 3,453 to 8,856, and reported value 4.4-fold, from $18.6 trillion to $82.3 trillion. Most of the value growth is the market itself: the dips in late 2018, early 2020 and 2022 line up with equity drawdowns. The filer count rises in steps at each year end, as managers who crossed $100 million during the year start filing.

Download the full quarterly series: 13f-series.json.

What each dataset tells you

What the research says

None of the findings below is new, and all of them come with the usual caveats about backtests. They show how professional and academic researchers have used these kinds of data.

Ownership breadth

Chen, Hong and Stein (2002) used 13F filings to count how many mutual funds held each stock. Their argument: when short selling is constrained, pessimists can only sit out, so a fall in the number of holders signals that prices reflect only the optimists. They found that stocks whose breadth of ownership fell went on to underperform stocks whose breadth rose.

Managers' best ideas

Cohen, Polk and Silli studied the positions where managers deviate most from what diversification alone would suggest. Those "best ideas" outperformed, while the rest of a typical portfolio added little. With 13F data, the same idea can be applied across every filer, not only mutual funds.

Crowding and comomentum

Lou and Polk measured crowding in momentum stocks from the excess correlation of their returns, which they call comomentum. When comomentum was high, momentum returns were weaker and more prone to reversal, and higher institutional ownership of past winners, measured from 13F data, went with higher comomentum. Crowding is the main way holdings data feeds risk models, not only alpha.

How Man Group measures crowding

Man Group's research on crowding combines several sources: similarity of holdings from 13F filings, short interest and securities-lending utilisation, intraday return correlations, and estimates of trend-follower (CTA) flows and how long a crowded position would take to unwind. Two findings stand out: highly utilised stocks have historically shown much more negative skew, and trend-following profits fall as crowding into a trend becomes more extreme. Man notes that the reporting lag limits measures based on 13F alone, one reason to blend them with faster data. Its Numeric equity team has described replacing crowded stocks with similar, less crowded ones rather than dropping exposure altogether.

Market timing with macro and positioning

In "Market Timing: More than a Mirage", Man researchers survey predictors of equity returns including GDP nowcasts, the yield curve, credit spreads, the VIX term structure, liquidity stress indicators and investor positioning and flows. Their conclusion is that models combining price momentum with macro information do better than any single signal.

AQR's macro momentum

In "A Half Century of Macro Momentum" (Journal of Portfolio Management, 2017), AQR's Jordan Brooks trades equity indices, currencies, 10-year government bonds and short-term interest rates on four themes: the business cycle (one-year changes in GDP and inflation forecasts), international trade (one-year currency moves against an export-weighted basket), monetary policy (one-year change in two-year yields) and risk sentiment (one-year equity excess returns). From 1970 to 2016 the hypothetical, gross-of-fees strategy had a Sharpe ratio of about 1.2, a correlation of about 0.4 with trend following, and a 50/50 mix with trend reached about 1.4.

AQR's fundamental trends and dislocations

AQR's "Fundamental Trends and Dislocated Markets" pairs a Systematic Macro sleeve, built on business-cycle, monetary-policy, trade and sentiment indicators such as the Chicago Fed National Activity Index, industrial production surprises and inflation surprises, with an Opportunistic Macro sleeve that enters when valuations reach roughly two-standard-deviation extremes. The paper reports a hypothetical 1993 to 2018 Sharpe ratio of about 1.2, around 12% a year at 10.5% volatility, and above 1.4 when combined 50/50 with multi-asset style premia.

Putting it together: a factor-model recipe

A sketch of how the four datasets combine in one research process. Treat each step as a hypothesis to test.

1. Make it point-in-time. Key each 13F quarter to the filing date, not the quarter end: holdings for June become usable in mid-August. Use macro vintages as they were first published.
2. Build cross-sectional signals per stock and quarter: change in breadth (holders against the prior quarter), net flows (new plus increased positions, minus closed and decreased), ownership concentration among the largest holders, overlap between large holders as a crowding score, and conviction (a stock's weight in a manager's book relative to its market weight).
3. Neutralise and rank. Remove sector and size effects, convert to z-scores and combine.
4. Condition on the regime. Use growth and inflation momentum, the change in Fed net liquidity and COT positioning to vary exposures over time, for example cutting exposure to crowded names when liquidity is contracting.
5. Test on survivorship-free prices. Include delisted stocks, trade after the filing date and charge realistic costs.

import finzdata as fz
c = fz.Client()

own = c.institutional_ownership(ticker="NVDA", start="2015-01-01")    # holders, shares, flows by quarter
own["breadth_chg"] = own["holders"].diff()
own["net_flow"] = (own["new_positions"] + own["increased"]) - (own["closed_positions"] + own["decreased"])

top = c.holders(ticker="NVDA", period="2026Q2")       # every manager that reported it
book = c.manager_holdings(cik=top["cik"].iloc[0])     # one manager's full book, with weights

liq = c.liquidity("FED_NET_LIQUIDITY")                # funding backdrop
cpi = c.macro("CPIAUCSL", as_of="2024-01-31")         # only what was published by that date
px = fz.Ticker("NVDA").history(start="2015-01-01")    # survivorship-free prices

Caveats

• The lag is real. Holdings are up to 45 days old when published and 135 days old by the next filing. Signals that decay quickly will not survive it.
• Long only. 13F shows no short positions, so a manager's net exposure is unknown. Pair it with short interest and lending data.
• Quarter-end snapshots. Positions held between quarter ends are invisible, and some managers tidy portfolios before quarter end.
• Amendments and confidential treatment. Some positions are filed late or withheld for a time. FinzData keeps the latest amendment for each manager and quarter, and the filing date of each line.
• Hypothetical results. Performance figures quoted above come from the cited papers and are hypothetical, mostly before fees and costs. This page is research, not investment advice.

Try it on the data

The latest quarter of institutional ownership is free with an API key. Every manager's full book back to 2013 comes with a look-back bundle, alongside survivorship-free prices, point-in-time macro, liquidity, TIC flows and COT positioning: $195 for the first 5 years of history, $95 for each further 5, plus $19.95 a month for daily updates.

Get a free API key   Browse 13F managers · 13F dataset details · Pricing

Sources

1. Chen, J., Hong, H. and Stein, J. (2002). Breadth of ownership and stock returns. Journal of Financial Economics.
2. Cohen, R., Polk, C. and Silli, B. Best Ideas. London School of Economics working paper.
3. Lou, D. and Polk, C. Comomentum: Inferring Arbitrage Activity from Return Correlations.
4. Man Group. Crowding. Man Institute.
5. Man Group. Market Timing: More than a Mirage. Man Institute.
6. Brooks, J. (2017). A Half Century of Macro Momentum. AQR Capital Management.
7. Bakrania, Doheny, Fader and Heinrichs. Fundamental Trends and Dislocated Markets. AQR Capital Management.
8. US Securities and Exchange Commission. Form 13F data sets.