Reverse Stock Splits and Liquidity in ETFs

Abstract

This study aims to address a gap in the existing literature pertaining to the liquidity of exchange-traded funds (ETFs). Specifically, we examine the effect of reverse share splits on ETF liquidity. In contrast to equities, the utilization of ETFs enables the separation of signaling and liquidity considerations. Findings suggest that liquidity improves after reverse splits in both univariate and multivariate results. In the absence of delisting concerns, results support the liquidity hypothesis of stock splits.

1. Introduction

In a conventional stock split, the total number of outstanding shares is increased, corresponding to a proportional decrease in the stock price. The literature posits that firms split their shares due to signaling and/or liquidity considerations (He and Wang 2012). Signaling is the practice of communicating management’s knowledge about a company’s potential future performance through stock splits (Brennan and Copeland 1988). The liquidity hypothesis states managers may opt to split their company’s stock as a strategic maneuver to enhance liquidity (Yu and Webb 2009). One pathway to improve liquidity is to enhance stock marketability. In the context of forward splits, increased marketability can attract retail investors who prefer lower equity price ranges.

In contrast to a stock split, a reverse stock split involves a reduction in the number of shares, accompanied by a proportional increase in the price per share. The literature suggests that reverse stock splits can be used to avoid delisting, broaden the investor base, reduce shareholders for a private takeover, and cut costs (Martell and Webb 2008). Previous research implies that reverse splits can also enhance the marketability of stocks by adjusting prices to a range perceived as more favorable or secure by investors (Peterson and Peterson 1992). Bacon et al. (1993) find that managers employ reverse stock splits to adjust stock prices to a more advantageous range and expand their ownership bases. In the context of reverse splits, increased marketability can attract active investors who prefer lower relative transaction costs via a higher price range. According to Han (1995), the underlying justification for implementing reverse splits lies in its impact on liquidity. Liquidity plays a crucial role in markets as it ensures that investors can easily buy or sell shares at fair prices without significantly impacting the market. High liquidity allows for efficient trading and provides investors with better execution and low transaction costs. Additionally, a liquid market is more attractive to institutional investors who require the ability to enter or exit positions quickly and in large quantities.

This study examines the effect of reverse stock splits on exchange-traded fund (ETF) liquidity. ETFs are unique instruments to examine since signaling effects can be disentangled from liquidity effects. In contrast to the managers of individual firms, the managers of ETFs lack access to proprietary information pertaining to the specific firms encompassed within the fund. Therefore, signaling cannot be considered a primary motive for implementing a reverse split in ETFs (Dennis 2003). In addition, delisting avoidance is not a primary concern in our sample, as all ETFs have a pre-split price above USD 1. Cox et al. (2022) suggest that the primary factor is improved marketability for stocks trading above USD 1 and conducting a reverse split. In this study, the bid–ask spread estimator of Abdi and Ranaldo (2017) is used to proxy for liquidity. Both univariate and multivariate results indicate a smaller mean spread proxy after reverse splits compared to before. Our results show that liquidity improves after reverse splits, as smaller post-split spreads indicate. These results are robust to the inclusion of spread determinants, including price, volume, and volatility, and to three different time windows (15, 30, and 60 days) around the reverse split event. To the best of our knowledge, this is the first study to examine the effect of reverse stock splits on the liquidity of exchange-traded funds.

2. Literature Review

2.1. Stock Splits and Liquidity

This study is related to the literature on splits and liquidity effects in equity markets. A common measure of liquidity used in literature is the relative spread. The relative spread is calculated by dividing the spread by the stock price to determine the spread on a percentage basis.1 Conroy et al. (1990) examine the effects of stock splits on bid–ask spreads for 133 NYSE-listed firms and find that percentage spreads significantly increase after splits. Similarly, evidence for increased relative spreads as a result of stock splits for a large sample of NYSE, AMEX, and NASDAQ firms is presented by Gray et al. (2003). Furthermore, Huang et al. (2015) find that relative spreads increase after splits in a sample of all NYSE, AMEX, and NASDAQ firms that split their stock. The previous findings suggest that stock splits lead to a decrease in liquidity, evidenced by the observation of larger relative spreads following stock splits. In a study conducted by Huang and Weingartner (2000), their findings support previous research indicating an increase in percentage spreads following stock splits. However, upon further analysis that takes into account the determinants of spreads, they observe that post-split spreads are not statistically different from pre-split spreads.

The effect of stock splits on ETF liquidity is also examined in prior literature. Dennis (2003) examines liquidity for the QQQ Nasdaq-100 tracking ETF after shares are split and finds that both relative quoted and effective bid–ask spreads are higher after the split. Bandyopadhyay et al. (2010) study the effect of stock splits on 12 iShare ETFs. They document increased relative spreads after the split. Yu and Webb (2009) examine the effect of ETF splits on liquidity for 20 ETFs with splits between 2000 and 2006. In contrast to prior studies, they find no difference in relative spreads after stock splits.

2.2. Reverse Stock Splits and Liquidity

This paper also relates to the literature on reverse stock splits and liquidity in equity markets. Historically, reverse splits are much less prevalent in markets than (forward) splits. For example, Gray et al. (2003) identify 1109 stock split events during 1993–1996, yet indicate too few instances of reverse splits to include in their analysis. Han (1995) investigates the effect of reverse splits on the liquidity of firms listed on the NYSE, AMEX, and NASDAQ from 1963 to 1990. Han documents a significantly lower standardized percentage spread for equities after reverse splits.2 A smaller spread after a reverse split is indicative of enhanced liquidity.

According to Cox et al. (2022), more recently, the number of forward splits has decreased over time, while the number of reverse splits has increased in equity markets. They examine forward and reverse splits over the years 2007–2019 and identify 341 forward splits (58% of splits occur during the first five years) and 324 reverse splits (77% of reverse splits occur during the last five years). Although more data on reverse splits are now available, there is a lack of research investigating the impact of reverse splits on the liquidity of exchange-traded funds.

3. Data and Methodology

3.1. Data

This paper aims to analyze the impact of reverse splits on the liquidity of ETFs. We identify 23 instances of reverse stock splits of exchange-traded funds in the years 2011 to 2022. The ETF sample we employ excludes leveraged ETFs since these are more susceptible to volatility and large price swings. March-Dallas et al. (2018) show that liquidity characteristics differ between leveraged and unleveraged ETFs.

Table 1. ETF samples. The table includes information about each ETF reverse split. Ticker represents the ticker symbol associated with the ETF. Reverse Split Date is the date on which the reverse split occurred. Reverse Split ratio defines the exchange ratio of old shares for new shares.

ETF Name	Ticker	Reverse Split Date	Reverse Split Ratio
Glbl X SuperDiv REIT	SRET	December 20, 2022	1 for 3
Glbl X Blockchain ETF	BKCH	December 20, 2022	1 for 3
SPDR Gold MiniShares	GLDM	February 23, 2022	1 for 2
IShares:Gold Trust	IAU	May 24, 2021	1 for 2
United States Oil Fund	USO	April 29, 2020	1 for 8
Glbl X MLP and Engy Infra	MLPX	April 28, 2020	1 for 3
Glbl X MLP ETF	MLPA	April 28, 2020	1 for 6
ProShares:K1 Fr Crd Oil	OILK	April 21, 2020	1 for 5
VanEck:Oil Services	OIH	April 15, 2020	1 for 20
VanEck:Energy Income	EINC	April 15, 2020	1 for 3
SPDR S&P Oil&Gas Exp	XOP	March 30, 2020	1 for 4
SPDR S&P Oil&Gas E&S	XES	March 30, 2020	1 for 10
United States Nat Gas	UNG	January 5, 2018	1 for 4
iShares:Mtge RE	REM	November 7, 2016	1 for 4
iShares:MSCI Gl M&MP	PICK	November 7, 2016	1 for 2
iShares:MSCI Gl GMiners	RING	November 7, 2016	1 for 2
FT:Natural Gas	FCG	May 2, 2016	1 for 5
Glbl X Silver Miners ETF	SIL	November 18, 2015	1 for 3
Vanguard 500 Idx;ETF	VOO	October 24, 2013	1 for 2
VanEck:Jr Gold Miners	GDXJ	July 1, 2013	1 for 4
United States Nat Gas	UNG	February 22, 2012	1 for 4
Invesco Solar	TAN	February 15, 2012	1 for 10
United States Nat Gas	UNG	March 9, 2011	1 for 2

provides the list of the exchange-traded funds used in this study.

The data encompass a range of reverse split dates spanning from March 9, 2011 to December 20, 2022. The reverse splits observed in our sample are not preceded or followed by another split (for the same ETF) within the preceding or subsequent 350 days. Following Shue and Townsend (2021), we aggregate reverse splits into one sample regardless of the specific conversion ratio between shares due primarily to their infrequent occurrence. Moreover, the reverse split ratios exhibit variations across different ETFs. The exchange-traded funds originate from diverse market sectors, encompassing energy, commodity, and equity indices. For each ETF in our sample, we obtain data from Yahoo Finance that includes the split-adjusted daily open price, high price, low price, close price, and volume.3

3.2. Variable Definitions

The accurate estimation of liquidity is essential for understanding market dynamics. The bid–ask spread is often used as a proxy for liquidity (Chordia et al. 2001). Corwin and Schultz (2012) develop a bid–ask spread (High–Low) estimator based on daily high and low prices. Although the High–Low estimator has a high correlation with effective spreads derived from Trade and Quote (TAQ) data, it requires an adjustment for nontrading days and assumes that stock prices do not change while the market is closed (Le and Gregoriou 2020). In practical applications, it is not uncommon for these assumptions to be violated, resulting in a decline in the accuracy of the estimates.

Abdi and Ranaldo (2017) propose a modified (AR) estimator to proxy for liquidity that utilizes a distinct information set comprising daily close, high, and low prices. The AR estimation is characterized by its independence from trade direction and lack of necessity for adjustments during nontrading periods. Consequently, the AR spread proxy offers a more accurate measure of liquidity when compared to the High–Low estimator (Le and Gregoriou 2020). The AR spread estimator is used as a proxy for liquidity in this study and is calculated for ETF i and day t as follows:

$${Spread}_{i,t}=\sqrt{\mathit{max}\left\{4\left(c_{i,t}-{\eta }_{i,t}\right)\left(c_{i,t}-{\eta }_{i,t+1}\right),0\right\}}$$

(1)

where η_i,t is the average of the high and low log-price, and c_i,t is the close log-price.4 Similarly to Bandyopadhyay et al. (2010), we focus on the percent spread, which is formulated as follows:

$${PercentSpread}_{i,t}=\frac{{Spread}_{i,t}}{{ClosePrice}_{i,t}}\times 100$$

(2)

where ${Spread}_{i,t}$ is the AR estimated spread and ${ClosePrice}_{i,t}$ is the close price for ETF i on day t.

Prior studies, such as Chordia et al. (2000), describe volatility, trade activity, and price as the primary spread components that should be controlled for in the context of liquidity studies. Volatility is estimated by the Garman–Klass (GK) volatility measure (Garman and Klass 1980). The GK volatility is estimated as:5

$$Volatilit{y}_{i,t}=\frac{1}{2}{\left[\mathit{Ln}\left(H_{i,t}\right)-\mathit{Ln}\left(L_{i,t}\right)\right]}^2-\left[2\mathit{Ln}\left(2\right)-1\right]{\left[\mathit{Ln}\left(O_{i,t}\right)-\mathit{Ln}\left(C_{i,t}\right)\right]}^2$$

(3)

where H_i,t, L_i,t, O_i,t, and C_i,t, represent the high, low, open, and close prices, respectively, for ETF i on day t. An essential characteristic of the GK volatility is its superior efficiency compared to the conventional close-to-close volatility estimator. Specifically, the GK volatility demonstrates an approximately eight-times greater efficiency in estimating volatility. The GK volatility assesses the level of volatility present within a specific time interval and does not necessarily exhibit an increase in value as the number of trades executed within the interval increases.6

Trade activity is proxied by dollar volume and is calculated as follows for ETF i on day t:

$${Volume}_{i,t}=Ln\left[{TradeVolume}_{i,t}\times {ClosePrice}_{i,t}\right]$$

(4)

where ${TradeVolume}_{i,t}$ is the daily number of shares traded and ${ClosePrice}_{i,t}$ is the close price for ETF i on day t. Consistent with Clifford et al. (2014), the natural logarithm of volume is considered to remedy skewness. The Price_i,t variable is defined as the close price for ETF i on day t.

3.3. Methodology and Hypothesis

The main objective of this study is to investigate the impact of reverse splits on ETF liquidity. In a reverse share split, the number of shares is reduced, resulting in an increased stock price. Prior research shows that transaction costs, as proxied by the spread, are positively related to the inverse price level (Chan et al. 2013). Therefore, a price increase should lead to decreased transaction costs and increased liquidity. The price level also impacts margin trading, as low-priced ETFs cannot be purchased on margin. The threshold for non-marginable ETFs varies depending on the brokerage (e.g., below USD 3 for Schwab and Fidelity). However, higher-priced ETFs will be eligible for margin trading, which may attract institutional investors and improve liquidity. Tradability and marketability are also affected by the price level and liquidity. Low-priced ETFs may be tick-constrained, resulting in an artificially high relative bid–ask spread. If prices increase post-reverse split, the relative spread decreases, resulting in increased liquidity and lower transaction costs for active traders. Lower transaction costs can also make ETFs more marketable to active traders. For example, Vanguard reverse split its S&P 500 ETF (VOO) to lower spreads and reduce investor transaction costs.7 Furthermore, ETFs may maintain price levels that exceed exchange-mandated minimum listing standards by employing reverse splits, a byproduct of which can potentially improve liquidity. In sum, based on potential reverse split effects and the findings of Han (1995) in equity markets, this study proposes the following research hypothesis:

Hypothesis 1. 

There is an increase in liquidity after a reverse stock split in ETFs.

To examine the difference in liquidity before and after reverse splits, t-tests, and non-parametric Wilcoxon matched-pair signed-rank tests are utilized. In addition, the following regression models are estimated to conduct a formal examination of the hypothesis:

$$PercentS{pread}_{i,t}={\alpha }_0+{\alpha }_1{Post}_{i,t}+{\alpha }_2{Volatility}_{i,t}+{\alpha }_3{Volume}_{i,t}+{\epsilon }_{i,t}$$

(5)

$${Spread}_{i,t}={\alpha }_0+{\alpha }_1{Post}_{i,t}+{\alpha }_2{Price}_{i,t}+{\alpha }_3{Volatility}_{i,t}+{\alpha }_4{Volume}_{i,t}+{\epsilon }_{i,t}$$

(6)

where Percent Spread_i,t is the AR estimated spread normalized by the close price, Spread_i,t is the AR estimated spread, Post_i,t is a dummy variable assigned a value of one after a reverse split and zero before a reverse split, Price_i,t is the close price, Volatility_i,t is the Garman–Klass estimator, and Volume_i,t is the natural logarithm of dollar volume for ETF i on day t. Huang and Weingartner (2000) emphasize the significance of considering the relationship between spreads and microstructure variables, such as price, volatility, and volume, in the context of stock splits. Standard errors proposed by Rogers (1993) are employed to ensure the robustness of statistical analyses in the presence of heteroscedasticity.

A statistically significant negative coefficient on ${Post}_{i,t}$ would indicate smaller post-split spreads, which provide evidence of increased liquidity in ETFs after controlling for spread determinants. In accordance with Martell and Webb (2008), the present study employs a time frame encompassing 30 trading days before and 30 trading days after a reverse split. The reverse split date is excluded from the sample. In addition, both a 15-day and 60-day period are incorporated to ensure robustness of the time window.

4. Results and Discussion

4.1. Descriptive Statistics and Univariate Results

The summary statistics for the liquidity proxies and spread determinants are provided in

Table 2. Summary statistics. This table presents the summary statistics of spread measures and determinants for 30 days before and 30 days after a reverse split. Percent Spread is the AR estimated spread normalized by the close price, Spread is the AR estimated spread, Price is the close price, Volatility is the Garman–Klass volatility, and Volume is the natural logarithm of dollar volume. Univariate results for the differences in means, and differences in medians are presented in the last two columns. T-tests to test the difference between means and non-parametric Wilcoxon matched-pair signed-rank tests to test the differences between medians are performed. *, **, and *** denote significance at the one percent level, five percent level, and ten percent level, respectively.

	Mean	Median	Stan Dev	Mean	Median	Stan Dev	Mean Diff	Median Diff
	After Reverse Split			Before Reverse Split
Percent Spread	0.036	0.025	0.039	0.056	0.027	0.080	−0.020 *	−0.002 **
Spread	0.009	0.007	0.008	0.014	0.008	0.016	−0.005 *	−0.001 *
Price	41.548	23.340	40.878	42.497	23.921	40.592	−0.950	−0.581
Volatility	7.864	3.551	12.101	20.003	4.857	50.606	−12.139 *	−1.306 *
Volume	7.217	7.198	1.022	7.142	7.077	1.068	0.169	0.121

. The univariate results for the comparison of means and medians of spreads and associated variables are presented in the last two columns.

The mean Percent Spread is smaller after the reverse split (0.036) than before the split (0.056). The mean difference is negative (−0.020) and statistically significant, implying that Percent Spread decreases after a reverse split. The mean and median of Spread are also significantly smaller in the period after a reverse split relative to the period before for exchange-traded funds. Narrow bid–ask spreads are consistent with reduced trading costs and increased liquidity. In terms of economic significance, the mean percent spread decreases by 35.7% after the reverse split. This decrease in estimated spread has a substantial impact on trading costs for active traders, especially for large orders. Enhanced liquidity after a reverse split is consistent with the results of Han (1995), who finds similar results for equities listed on the NYSE, AMEX, and NASDAQ that experienced a reverse stock split. As for the microstructural determinants of spread, Price declines by 2.2 percent, Volatility declines by 60.7 percent, and Volume increases by 1.1 percent in the period after a reverse split. The mean and median difference in Volatility is negative and statistically significant, implying that volatility decreases after a reverse split. A decline in volatility after a reverse split is consistent with Koski (2007), who finds that volatility decreases for equities after reverse splits. Shue and Townsend (2021) also document a decline in volatility following reverse splits. The reverse split does not affect price and volume. The finding for volume contrasts with Han (1995), who documents an increase in trading volume after reverse splits for equities. Overall, the univariate results imply a decrease in liquidity after a reverse stock split in ETFs.

The behavior of the percent spread for the 30 days before and 30 days after a reverse split is pictured in Figure 1.

The figure shows a large decline in the percent spread over the first ten days after a reverse split and an overall lower percent spread after a reverse split relative to the previous period.

4.2. Multivariate Results

This section presents the findings for the impact of reverse splits on liquidity while accounting for the influence of liquidity determinants.

Table 3. Multivariate results for percent spread. This table presents the coefficient estimates for the following model: ${PercentSpread}_{i,t}={\alpha }_0+{\alpha }_1{Post}_{i,t}+{\alpha }_2{Volatility}_{i,t}+{\alpha }_3{Volume}_{i,t}+{\epsilon }_{i,t}$. Percent Spread is the AR estimated spread normalized by the close price. Post is a dummy variable assigned a value of one after a reverse split and zero before a reverse split. Volatility is the Garman–Klass volatility. Volume is the natural logarithm of dollar volume. Panel A presents results for 30 days before and after a reverse split, Panel B presents results for 15 days before and after a reverse split, and Panel C presents results for 60 days before and after a reverse split. *, **, and *** denote significance at the one percent, five percent, and ten percent levels, respectively. The t-statistics are based on the Rogers (1993) robust standard errors.

Panel A: 30 Days	Coefficient	t-Stat.
Intercept	0.134 *	10.91
Post	−0.009 *	−3.06
Volatility	0.001 *	3.62
Volume	−0.013 *	−8.85
Adjusted R2	0.267
N	1380
Panel B: 15 Days	Coefficient	t-Stat.
Intercept	0.119 *	7.36
Post	−0.007 ***	−1.74
Volatility	0.001 ***	1.81
Volume	−0.011 *	−4.73
Adjusted R2	0.151
N	690
Panel C: 60 Days	Coefficient	t-Stat.
Intercept	0.111 *	15.35
Post	−0.006 *	−3.49
Volatility	0.001 *	4.07
Volume	−0.011*	−13.20
Adjusted R2	0.272
N	2760

presents results for the percent spread.

The coefficient on the Post dummy variable is the main variable of interest in the table. The coefficient (−0.009) associated with Post in Panel A is negative and statistically significant. A similar result for Post is presented in Panel B and Panel C. The main implication is that following a reverse split in ETF markets, a decline in the percent spread is reported after controlling for volatility and volume. The results are consistent across the three different time windows.

Table 4. Multivariate results for spread. This table presents the coefficient estimates for the following model: ${Spread}_{i,t}={\alpha }_0+{\alpha }_1{Post}_{i,t}+{\alpha }_2{Price}_{i,t}+{\alpha }_3{Volatility}_{i,t}+{\alpha }_4{Volume}_{i,t}+{\epsilon }_{i,t}$. Spread is the AR estimated spread. Post is a dummy variable assigned a value of one after a reverse split and zero before a reverse split. Price is the close price. Volatility is the Garman–Klass volatility. Volume is the natural logarithm of dollar volume. Adjustments are made to the t-statistics based on Rogers (1993). *, **, and *** denote significance at the one percent, five percent, and ten percent levels, respectively.

Panel A: 30 Days	Coefficient	t-Stat.
Intercept	0.014 *	4.66
Post	−0.003 *	−4.00
Price	−0.000	−1.28
Volatility	0.000 *	3.35
Volume	−0.000	−0.93
Adjusted R2	0.262
N	1380
Panel B: 15 Days	Coefficient	t-Stat.
Intercept	0.007 **	1.98
Post	−0.002 **	−2.55
Price	−0.000 **	−2.14
Volatility	0.000	1.64
Volume	0.001	1.05
Adjusted R2	0.186
N	690
Panel C: 60 Days	Coefficient	t-Stat.
Intercept	0.010 *	6.29
Post	−0.001 *	−3.70
Price	−0.000	−0.86
Volatility	0.000 *	3.80
Volume	−0.000	−0.91
Adjusted R2	0.246
N	2760

presents results for spread while controlling for price, volatility, and volume.

The primary variable of interest is the coefficient on Post. A negative and statistically significant coefficient is reported across all panels in the table. Results show evidence to support an increase in ETF liquidity due to a reverse split, as indicated by a decline in the spread. This finding contrasts with the results of Huang and Weingartner (2000) for forward stock splits in equity markets. They find that post-split spreads are insignificantly different from pre-split spreads after controlling for price, volatility, and volume.

4.3. Discussion

The presented results support the notion that ETF liquidity is enhanced as a result of reverse stock splits. In effect, we find evidence to support the liquidity hypothesis. However, prior research has argued that one potential reason to reverse split shares is to avoid delisting due to a low trading price. To be listed, most exchanges have a minimum price threshold, often equal to USD 1. Delisting concerns may serve as a primary driver for reverse splits, particularly when the pre-split price is low, as delisting would result in a deterioration of market quality and drastically reduced liquidity. Crutchley and Swidler (2015) examine firms that engaged in multiple reverse stock splits and find evidence that firms reverse split shares to avoid delisting or widen institutional ownership. In their sample of firms, 92 percent of equities have a pre-split price of less than USD 5. Furthermore, Martell and Webb (2008) suggest that reverse splits have been historically perceived as desperate measures to increase share prices to satisfy minimum price requirements. In their sample of NYSE, AMEX, and NASDAQ stocks, ninety-five percent of equities have a pre-split stock price below USD 5.

Table 5. ETF price distribution. This table presents the ETFs sorted into five price bins based on prices one day before and after the reverse split event. The total sample size of ETFs is 23. P represents the close price. Prices before a reverse split are not adjusted by the split ratio. Prices after the reverse split are adjusted by the split ratio.

	Number of ETFs
	p ≤ USD 1	p ≤ USD 5	USD 5 < p ≤ USD 10	USD 10 < p ≤ USD 20	USD 20 < p ≤ USD 30	p > USD 30
Before Reverse Split	0	8	10	2	1	2
After Reserve Split	0	0	0	7	8	8

presents a summary of the price distribution for the pre-reverse split and post-reverse split ETF sample used in this study.

Table 5 shows that only eight ETFs have a price less than USD 5 before a reverse split. Furthermore, none of the ETFs trade below USD 1 in our sample, the threshold for delisting. Cox et al. (2022) suggest that enhanced marketability is the principal factor in reverse splits involving stocks trading above USD 1. In contrast to the equity data used in Crutchley and Swidler (2015) and Martell and Webb (2008), 65% of the ETFs employed in this study have a pre-split price larger than USD 5. ETFs may sustain artificially elevated prices to evade delisting through prior reverse splits. In our ETF sample, only one ETF (UNG) has had more than one reverse split. Therefore, delisting concerns are not a primary objective for the ETFs that conduct reverse splits in our sample. The table also shows a general shift in prices from less than USD 10 before a reverse split to greater than USD 10 after a reverse split, supporting the marketability explanation for reverse splits.

Mackintosh (2022) indicates that ETFs are extremely liquid, transacting in excess of USD 170 billion daily. However, the daily contribution of retail traders to ETF trading is less than 5%. Mackintosh suggests institutional investors, such as market makers, banks, and hedge funds, engage in substantial ETF trading. In effect, institutional traders are the active traders in ETF markets. ETFs may be incentivized to create trading opportunities with low transaction costs and high liquidity to attract trading from the active market participants. O’Hara et al. (2019) suggests that a larger relative tick size attracts active traders. Note that a larger relative tick size results as a consequence of a reverse split. In addition, Clifford et al. (2014) examine the drivers of ETF flows and suggest that ETF flows increase following high volume and small spreads.

Therefore, we suggest ETFs reverse split shares to improve liquidity, enhance marketability, and drive more cash flow into the split ETF.

Relevant to the discussion is the effect of ETF reverse splits on open orders and options. Open orders are canceled as a result of a reverse split. This can impact market quality and liquidity in the short term until positions are rebalanced. Options on ETFs are adjusted subsequent to a reverse split. In particular, the strike price and the number of deliverable shares in each option contract are adjusted based on the reverse split multiplier, whereas the number of option contracts is not affected. In aggregate, the option position and total payoff remain consistent with the value before the reverse split.

As an additional robustness test, we utilize the difference-in-differences estimation technique to investigate the impact of reverse splits on ETF liquidity. To control for general market changes, we create a control sample of non-splitting ETFs. ETFs are matched based on industry (or fund category) and returns, similar to the approach utilized in Yu and Webb (2009).8 

Table 6. Summary statistics matched sample. This table presents the summary statistics of spread measures and determinants for 30 days before and 30 days after a reverse split for the matched ETF sample. Percent Spread is the AR estimated spread normalized by the close price, Spread is the AR estimated spread, Price is the close price, Volatility is the Garman–Klass volatility, and Volume is the natural logarithm of dollar volume. Univariate results for the differences in means, and differences in medians are presented in the last two columns. T-tests to test the difference between means and non-parametric Wilcoxon matched-pair signed-rank tests to test the differences between medians are performed. *, **, and *** denote significance at the one percent level, five percent level, and ten percent level, respectively.

	Mean	Median	Stan Dev	Mean	Median	Stan Dev	Mean Diff	Median Diff
	After Reverse Split			Before Reverse Split
Percent Spread	0.088	0.048	0.117	0.123	0.060	0.174	−0.035 *	−0.012 *
Spread	0.009	0.007	0.007	0.012	0.008	0.012	−0.003 *	−0.001 *
Price	23.942	14.130	34.101	23.362	12.279	33.535	0.580	1.850
Volatility	5.971	2.552	9.549	13.192	3.180	29.098	−7.222 *	−0.628 *
Volume	6.440	6.191	1.147	6.371	6.047	1.140	0.069	0.144 ***

presents the summary statistics for the matched sample of non-splitting ETFs.

In the post-split period, the mean percent spread (0.088) is lower than the pre-split period value (0.060). The mean spread decreases by 0.003 from the pre-split to the post-split period for the sample of matched ETFs. The mean price and volume do not change between the periods, while volatility decreases from 13.192 to 5.971.

Table 7. Difference-in-differences results. The table presents results for the difference-in-differences estimation for the percent spread. Post-Split and Pre-Split represent the time period after and before a reverse split, respectively. Values are calculated based on the time frame 30 days before and after a reverse split. Split and No Split represent ETFs that reverse split and do not reverse split their shares. *, **, and *** denote significance at the one percent level, five percent level, and ten percent level, respectively.

	Post-Split	Pre-Split	Difference
Split	0.036	0.056	−0.020 *
No Split	0.088	0.123	−0.035 *
Difference	−0.052 *	−0.067 *	0.015 **

presents the results for the difference-in-differences approach.

The percent spread decreases by 0.020 (statistically significant at one percent level) from the pre-split to post-split period for ETFs that reverse split their shares. This change represents a 35.7 percent decrease in the percent spread following a reverse split. The matched ETFs that do not undergo a reverse split show a decrease of 0.035 in percent spread, which corresponds to a 28.5 percent decrease. The mean of the difference-in-differences value is positive (0.015) and statistically significant at the five percent level, which implies that the impact of the reverse split on the percent spread is larger among ETFs that reverse split their shares than among non-splitting matched ETFs. The results are consistent with the notion that liquidity improves in ETFs following a reverse split. We also consider the effects of other covariates (volume and volatility) and include them as controls in

Table 8. Difference-in-differences multivariate results. This table presents the coefficient estimates for the following model: ${PercentSpread}_{i,t}={\beta }_0+{\beta }_1{Post}_{i,t}+{\beta }_2{Split}_{i,t}+{\beta }_3\left({Post}_{i,t}\right)\left({Split}_{i,t}\right)+{\beta }_4{Volatility}_{i,t}+{\beta }_5{Volume}_{i,t}+{\epsilon }_{i,t}$. Percent Spread is the AR estimated spread normalized by the close price. Post is a dummy variable assigned a value of one up to 30 days after a reverse split and zero up to 30 days before a reverse split. Split is a dummy variable assigned a value of one for an ETF with a reverse split and zero for an ETF without a reverse split. Volatility is the Garman–Klass volatility. Volume is the natural logarithm of dollar volume. *, **, and *** denote significance at the one percent, five percent, and ten percent levels, respectively. The t-statistics are based on the Newey and West (1987) correction.

	Coefficient	t-Stat.
Intercept	0.193 *	15.66
Post	−0.024 *	−4.23
Split	−0.065 *	−11.18
(Post)(Split)	0.022 *	2.69
Volatility	0.001 *	20.56
Volume	−0.014 *	−7.56
Adjusted R2	0.208
N	2760

.

The difference-in-differences regression in Table 8 is estimated using the generalized method of moments (GMM). The Newey–West correction is applied to assure robustness to serial correlation and heteroskedasticity. The main variable of interest is the coefficient on the interaction term, (Post)(Split). The significant positive coefficient (0.022) implies that percent spread changes by a greater amount for reverse split ETFs relative to non-split ETFs following a reverse split, in the presence of control variables.

5. Conclusions

This study examines the effect of reverse splits on ETF liquidity. Employing ETF data allows us to disentangle possible signaling and liquidity considerations. To the best of our knowledge, no prior studies have explored the effect of reverse splits on ETF liquidity. By examining this relation, we are able to fill a gap in the current literature on ETF liquidity. Our results suggest that ETF liquidity improves after reverse splits. We find support for the liquidity hypothesis of stock splits and note that delisting is not a concern in our ETF sample. Avenues for further research include examining the effect of reverse splits on intraday ETF liquidity, market depth, and institutional ownership.