The Overnight Drift

by Nina Boyarchenko, Lars C. Larsen, and Paul Whelan

Review of Financial Studies (2023)

Harsh Vimal Javier Ramos Pérez

Financial Economics Reading Group

Center for Monetary and Financial Studies
June 2024
Part 1/3
Aperitivo: Cliff, Cooper, and Gulen (2008) on S&P 500 Spider ETF

1/24
Main Result

Intraday return averages

Average hourly log returns and average cumulative 5-minute log returns holding the E-mini contract

2/24
Main Result: “Overnight Drift”

Returns earned between 2:00 am – 3:00 am ET = 3.7% p.a. (1.48 bps/day) are

• positive in 20 of the 23 years in the sample

• large relative to the 5.9% p.a. CTC return
• persistent on all days of the week, and in 9 of 12 months of the year
• statistically significant in 17 of the 23 years in the sample (at the 10% level)

Possible Explanations

1 Presence of differentially informed traders in the market

Revelations must be positive, and must happen between 16:15 and 2:00
2 Inventory (risk) management by market makers
Selling pressure during ID ⇒ −ve order imbalance at the end of ID ⇒ MMs become net buyers (bear
inventory risk) ⇒ MMs sell to new market participants in ON at a premium
Grossman and Miller (1988)

3/24
Literature

• Equities make a substantial proportion of their returns in ON compared to ID

Cliff, Cooper, and Gulen (2008), Kelly and Clark (2011)
• Cross-sectional momentum is earned overnight
Lou, Polk and Skouras (2017)
• Investors have predictable demand for immediacy at certain points within the day
Heston, Korajczyk, and Sadka (2010)
• Size and illiquidity portfolios earn their returns around market close
Bogousslavsky (2021)
• CAPM holds overnight
Hendershott, Livdan and Rösch (2018)
• Price reversals are linked to inventory risk
Grossman and Miller (1988), Vayanos (2001), Brunnermeier and Pedersen (2009)
• Price dynamics reflect “slow capital” even in extremely active markets
Duffie (2010)

4/24
Data

E-mini S&P 500

• Traded ∼ 24 hours on CME GLOBEX, allowing for trading outside US market hours
• Leads the price discovery of the S&P 500 index
• Most traded equity index futures contract in the world
Tick history data on E-mini S&P 500 futures from the London Stock Exchange Group (LSEG)

• Sample from January 1998 - December 2020

• Quoted best bids and asks
• Realised traded price, volumes, and the time of transaction
• First 5 levels of the order book
• Focus on the most traded contract

5/24
Part 2/3
Main Result: “Overnight Drift”

Returns earned between 2:00 am – 3:00 am ET = 3.7% p.a. (1.48 bps/day)

• positive: in 20 of the 23 years in the sample

• large: > 50% of the 5.9% p.a. CTC return
• persistent: on all days of the week, and 9 of 12 months of the year
• statistically significant: in 17 of the 23 years in the sample (at the 10% level)

Potential explanations:

1 Presence of differentially informed traders in the market

Revelations must be positive, and must happen between 16:15 and 2:00
2 Inventory (risk) management by market makers
Selling pressure during ID ⇒ −ve order imbalance at the end of ID ⇒ MMs become net buyers (bear
inventory risk) ⇒ MMs sell to new market participants in ON at a liquidity premium
Grossman and Miller (1988)

6/24
Main Result

Intraday return averages

Average hourly log returns and average cumulative 5-minute log returns holding the E-mini contract

7/24
Main Result

Hour 23-24 24-01 01-02 02-03 03-04 04-05 05-06

Mean 0.04 0.46 0.43 1.48 0.35 -0.08 0.15

t-stat 0.29 2.77 2.75 7.13 1.33 -0.32 0.63
p-value 0.77 0.01 0.01 0.00 0.18 0.75 0.53
median 0.00 0.00 0.00 0.65 0.00 0.00 0.00
SD 10.34 12.60 11.95 15.78 21.45 19.63 17.50
Skew -0.67 7.38 -0.32 1.20 0.02 -0.87 -0.45
Kurt 35.33 213.53 34.05 33.78 16.91 21.74 19.83

8/24
OD is positive

9/24
OD is statistically significant

10/24
OD is persistent

11/24
OD is persistent

12/24
Potential Explanation: Inventory Risk

13/24
Volumes

14/24
Cumulative Returns in Clock Time

15/24
Cumulative Returns in Volume Time

16/24
Demand Shock Asymmetry

17/24
Part 3/3
Recap

Fact 1: There exists an “overnight drift” in US futures

Fact 2: These abnormal positive returns are realized in the opening of European markets

Explanations

• OD is caused by EOD order imbalances, which generate overnight price reversals

• These order imbalances cannot be resolved until European trading begins
• Demand shock asymmetry is consistent with constrained intermediation
• Bear rallies generate much stronger reversal than bull rallies
• Returns accrue linearly not in clock time but in volume time

18/24
Discussion

Paper deserves credit for

• Studying futures data to provide a new stylized fact

• Why does return premium accrue unevenly over the night?
◦ Solid argument about volume time
◦ Supporting evidence from closing order imbalance and volatility
• Important contribution to the literature on ID and ON returns
• Directions for future work
◦ Link to stock-level data and test directly for changes in liquidity
Amihud (2002), Hameed, Kang, and Viswanathan (2010)
◦ Interactions with weekend effect
French (1980)
◦ Extend to more asset classes and countries

19/24
Edge Case (Qiao and Dam (2020))

20/24
AlphaTrAI launched NightShares in June 2022

“...although the documented high-frequency return patterns of this paper are not easily profitable, the
persistent presence of the overnight drift suggests that the intraday timing of portfolio adjustments should
be an important consideration for asset managers and institutional investors. Indeed, market developments
suggest that arbitrageurs are trying to capitalize on the patterns identified in this paper, with NightShares
launching two exchange-traded funds (ETFs) in June 2022, targeted specifically at earning the overnight drift,
citing financing costs and end-of-day order imbalances as driving forces behind the pattern.”

21/24
. . . to heavy market optimism.

22/24
. . . to heavy market optimism.

22/24
Night ETFs lagged behind both small-cap and large-cap benchmarks

23/24
. . . and were eventually shut down.

24/24