The Quarterly Review of Economics and Finance 80 (2021) 627–634

Contents lists available at ScienceDirect

The Quarterly Review of Economics and Finance

journal homepage: www.elsevier.com/locate/qref

The impact of order flow on event study returns:

New evidence from zero-leverage firms
Sijia Zhang a , Andros Gregoriou b,∗
a
School of Economics, Jilin University, No.2699 Qianjin Road, Changchun, 130000, China
b
Brighton Business School, University of Brighton, Mithras House, Lewes Road, Brighton, BA2 4NT, UK

a r t i c l e i n f o a b s t r a c t

Article history: We empirically examine the initial loan announcement period of 96 zero-leverage firms listed on the
Received 24 September 2019 FTSE 350 index. Our research demonstrates that there is a clear tendency that trades are executed at
Received in revised form the ask price during the initial loan announcement period, which are regarded as favorable firm-level
30 December 2020
news. Similar results are observed for subsamples formed on the basis of trade size. Order flow disruption
Accepted 15 April 2021
causes a bias in the calculation of returns around the company event announcement.
Available online 19 April 2021
Crown Copyright © 2021 Published by Elsevier Inc. on behalf of Board of Trustees of the University of
Illinois. All rights reserved.
JEL classification:
G14
G15

Keywords:
Order flow ratio
Liquidity
Bid-ask spread
Bid-ask bounce

1. Introduction ing pressure being significantly greater during the positive news
announcement period.
Over the last 20 years, there has been evidence that the existing Our primary objective is to obtain the empirical evidence of
bid-ask bias and order flow imbalance interact with stock return. buy-sell order imbalances and return anomalies during the ini-
Bid-ask bounce means that the price of a stock bounces rapidly tial loan announcement period of zero-leverage firms. According
forth and back within the very limited range between the bid price to Strebulaev and Yang (2013), a zero-leverage firm is defined as
and ask price. The bid-ask bias is a possible explanation for return a company that does not possess any short or long-term debt.
anomalies as return reversals might simply be a shift from trans- Zero-leverage firms occupy an important position in the market.
actions at bid prices to transactions at ask prices. Most of previous Strebulaev and Yang (2013) find in the U.S., an average 10.2 % of
studies (e.g.: Alzahrani et al., 2013; Gregoriou, 2008) on order flow large public firms follow zero-leverage policy during 1962–2009,
imbalance focus on block trading, which are minimum trade size of and there are almost 22 % firms that keep less than 5% book leverage
10,000 shares. They discover that a temporal imbalance between ratio. Dang (2013) discovers in the UK financial market, more than
buy and sell orders relate with the informed traders who are trad- 12.8 % of firms operate a zero-leverage policy between 1980 and
ing based on the private information. Buyer-initiated trades have 2007. Ghoul et al. (2018) report that the percentage of zero-leverage
a stronger price impact than seller-initiated trades, and this asym- firms in 2010 (17 %) has approximately doubled from 1990 (8 %).
metry has been attributed to higher informed trading in purchases Bessler et al. (2013) conclude that most zero-leverage firms are less
than in sells. As a result of these, stocks prices and further orders profitable, have a higher level of return volatility, are smaller and
react accordingly. Lo and Coggins (2006) indicate that order flow younger than other levered firms. They suffer from credit restric-
conveys information beyond trading volume. Gosnell et al. (1996) tions and therefore financial constraints. Hadlock and Pierce (2010)
state that order flow imbalances are instantly transparent with buy- report that the low or zero leverage firm which has short history
and small market chaptalization suffers more from adverse selec-
tion and information asymmetry. Therefore, they prefer to build
up their reputation in credit markets by borrowing and repaying
∗ Corresponding author.
from banks. The announcement of an initial bank loan is a unique
E-mail address: a.gregoriou@brighton.ac.uk (A. Gregoriou).

https://doi.org/10.1016/j.qref.2021.04.014
1062-9769/Crown Copyright © 2021 Published by Elsevier Inc. on behalf of Board of Trustees of the University of Illinois. All rights reserved.
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

event in the financial market because it contains information about zero-leverage firms traded on the London Stock Exchange (LSE)
the first time that a zero-leverage firm encounters a debt. Zhang over a 16-year time period. Moreover, our sample firms are signifi-
and Gregoriou (2019a) observe an improvement of liquidity and cantly larger than previous studies, 96 zero-leverage firms listed on
positive market reaction when zero-leverage firms are issued their the FTSE 350 across many different industries. The prior research,
initial bank loan. The stock price goes up whereas bid-ask spread for example, Brown et al. (1997) and Hasbrouck and Seppi (2001)
goes down, while the level of liquidity is improved overall. If order study order imbalances for thirty and twenty stocks, over one and
flow disruption occurs on the initial loan announcement date, this two years, respectively. Furthermore, the LSE allow us to identify
could provide an explanation of abnormal stock trading activities the exact trading direction without referring to any trade identi-
around the initial loan announcement. fication algorithms, such as the Lee and Ready (1991) rule, which
To our knowledge, this is the first empirical study that exam- would inevitably introduce some estimation errors.
ines the order flow ratio of zero-leverage firms during the initial The remainder of the paper is organized in the following way.
loan announcement. Most of previous studies (e.g.: Alzahrani et al., In the next section we provide a review of the previous litera-
2013; Gregoriou, 2008) on order flow imbalance focus on block ture. Section 3 discusses the data collection, Section 4 describes
trades, which are minimum trade size of 10,000 shares. It is not clear the methodology used. Section 5 reports the empirical analysis of
whether similar results hold for zero-leverage firms or any other the stock market reaction of bank loan announcements of zero-
small or infrequently traded stocks. We begin with the empirical leverage firm. Finally, Section 6 concludes our findings.
analysis by computing both the transaction price return and mid-
point price return of zero-leverage firms around the initial loan
announcement date. We then explore the placement of a trade 2. Literature review
within the bid-ask spread by measuring the order flow ratio, for
all trades and small, large trade groups. Finally, we examine the When the zero-leverage firms decide to lever up, bank loans
relationship between order flow ratio and transaction price-based, will be their first choice since they need to establish a good credit
mid-point-based returns. We attempt to answer the question does history and reduce their costs of public debt. Berger and Udell
the order imbalance provide the explanation of abnormal return of (1995) believe that the process of the bank issuing an initial loan
zero-leverage firms during the initial loan announcement period, to a zero-leverage borrower can reduce information asymmetry
and how the different trade size responds to the news which is problems. The transactions give banks the opportunity to gather a
reflected by the order flow imbalance and return bias. vast amount of information on their borrowers at minimum costs.
Our paper also makes a contribution on the analysis of the Financial intermediation hypothesis believes that banks are able
position of trades within the spread around the corporate event to examine potential borrowers more effectively than the public
announcement. Blume et al. (1989) and Lease et al. (1991) exam- as they have access to private information about the companies
ine the relationship between order flow imbalance and stock (Lummer and McConnel, 1989). Zhang and Gregoriou (2019a) find
return during the bad news announcement period. In contrast, in that the initial loan announcement results in overall improvement
our research the initial loan announcement can be regarded as a of liquidity and sends a positive signal to the market. The signaling
favourable event. It could indicate that the order flow imbalance effect of initial loan announcement to the market causes a posi-
causes buying or selling pressure around the initial loan announce- tive market reaction. It produces superior information which can
ment of zero-leverage firms. According to Zhang and Gregoriou resolve information asymmetry in the market and this is reflected
(2019a), adopting debt for the first time causes a positive reaction by reduction in the bid-ask spread.
in market value of the firms, decreasing the bid-ask spread and the A company announcement event can result in a movement of
stock market volatility. This is because bank lending activities cer- closing prices toward either the bid (a predominance of sell orders)
tify firm quality and signal credit worthiness to outside investors. or the ask quote (a majority of buy orders), and it can be measured
As bid-ask spread size is one of the sources of stock return errors, by order flow ratio. Gosnell et al. (1996) reveal that order flow is
any changes in the bid-ask spread will cause the changes of bid-ask significantly imbalanced around the news announcements. For bad
error automatically. If the spread is minimal, the bid-ask bounce news in their research, the transaction price location shifts towards
becomes unnoticeable. Therefore, the volatility of spread during a dominance of sell orders immediately after the announcement,
the announcement period creates a favourable environment for and the imbalance in the order flow ratio last in the long term.
detecting the bid-ask bias. Therefore, as the favourable corporate event, the announcement of
In addition, we provide evidence on the difference between clos- initial loan will attract more buyer-initialled orders and the transac-
ing price return and mid-point based returns, and the relationship tion price location should shift towards a dominance of buy orders.
between them and the order flow imbalance ratio around the initial Moreover, Lee (1992) finds that during the post-announcement
loan announcement. The bid-ask spread widens for zero-leverage period, there is no difference between the number of large buy and
firms’ stock (Zhang and Gregoriou, 2019a). Therefore, the bid-ask sell orders, but there is a significant increase in the volume of small
bounce which is caused by trades taking place at the bid or ask buy and sell orders. Zhang and Gregoriou (2019b) examine the price
quotes as opposed to the midpoint should be more influential. As behavior of zero-leverage firms after the initial loan announce-
the stock returns are computed based on transaction prices, the ment. They observe that the buy order has a greater impact on
statistically significant profits may not be economically realized. price changes, once price effects are estimated using quote returns
In order to mitigate the influence of shifts between the bid and to eliminate the bid-ask bias, the asymmetry effects between buy
the ask price, we measure stock returns using the midpoints rather and sell orders are dramatically decreased. Bissoondoyal-Bheenick
than the closing transaction prices. et al. (2019) show that the effect of seller/buyer-initiated trades
One of the other contributions of our study is the size of our on bad/good volatility is asymmetric, the effect of seller-initiated
sample. Most previous research analyzes order flow imbalances trades on bad volatility is consistently larger than that of buyer-
concentrating on a short time period, for example, Alzahrani et al. initiated trades on good volatility. On the other hand, Kim and
(2013) examine the imbalance between purchases and sales in the Stoll (2014) examine the relation between informed trading and
Saudi stock market for almost 4 years. Levi and Zhang (2015) ana- order imbalance around the announcement event. Order imbal-
lyze the changes of buyer and seller initiated orders around the ances do have price effects, but there is little evidence that the
event announcement during 26 months. Our data span is signif- order imbalance is significantly correlated with the forthcoming
icantly longer than previous studies. We investigate the event of announcements.

628
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

Chordia et al. (2019) discover that the volatility of order imbal- issue date and loan amount. This enables us to calculate the loan
ance can be a measurement of information asymmetry cost. An amount to total assets ratio. In total, there are 96 zero-leverage
increase in volatility of order flow imbalance is associated with firms within the FTSE 350 over our sample period that have suf-
greater adverse selection cost. Huang et al. (2020) report that the ficient data for our study during the period. The frequency of the
volatility in order imbalance works well in representing the costs zero-leverage adoptions of loan debt by year can be seen in the
of information asymmetry when trading against informed institu- Appendix A.
tional investors in the Taiwan stock market. Zhang et al. (2019) Panel A of Table 1 presents the firm-level information of zero-
demonstrates that in the Chinese stock market, the daily order leverage firms. The zero-leverage firms are smaller than leveraged
imbalances are highly persistent, but the price pressure caused by firms, as we predicted, the spreads of zero-leverage firms are higher
imbalances cannot last beyond a trading day. This suggests that than the firms with debt. Lease et al. (1991), believe that the order
China’s stock market is efficient enough to absorb the imbalances. flow bias is larger for stocks of smaller firms or with high price
Kaul and Nimalendran (1990) find that the bid-ask error com- variability, since the bid-ask spread, as a percentage of price, is
ponent of transaction returns can explain over 50 % of daily return larger for such stocks. Especially when returns are computed with
variances, resulting in excess volatility and price reversals in the transaction prices.
short term. Therefore, return series constructed using bid, ask, or
the average of bid-ask prices could result in more reliable infer- 3.2. Trading data
ences. Lease et al. (1991) use both transaction-price-based returns
to examine possible trading profits and quote-midpoint-based Panel C presents the trades analyzed in our research. The LSE
returns to evaluate equilibrium prices movements. They demon- identifies a trade as a purchase or a sale on their database. The over-
strate that the bid-ask bounce explains a significant proportion all sample consists of 28,342,080 buy orders and 26,101,600 sell
of the excess negative returns earned on seasoned equity offer- orders, which is substantially larger when compared with previous
ings. If closing prices just prior to the announcement accept at the research studies. We then examine the buy and sell orders by the
bid price while closing prices after the announcement accept at trade size. We use a share-based proxy following Gregoriou (2008)
the ask price, then an additional bid-ask bounce bias would be and most of the previous studies, with small trades defined as the
introduced. Gosnell et al. (1996) believe that due to randomness trade size less than 10,000 shares and the large trades as greater
bouncing between the bid and ask prices, the closing-price return than 10,000 research (block trades). In our study, the buy and sell
has a higher volatility than that of quote price or the mid-point orders for the small trade subgroup is 16,880,654 and 11,030,300
return. A spurious gain or loss caused by trades moving away from respectively, whereas for large trades, the buyer-initialled orders
the bid-ask midpoint. This gain or loss is later reversed as trans- are 11,461,426 and 15,071,300 for seller-initialled orders. Follow-
actions resume at the bid-ask midpoint. Han and Lesmond (2011) ing Gregoriou (2008) among others, small (large) trades are defined
argue that the bid-ask bounce inherent a liquidity bias, the price as transactions of less (more) than 10,000 shares.
movement is not a “true” price movement, but that it neverthe-
less increases the volatility of stock returns and consequently price
4. Methodology
reversal. Lo and Coggins (2006) maintain that the degree of return
reversal is positively related to the level of order flow imbalance.
4.1. Order flow ratio
Chordia and Subrahmanyam (2004) find that there is a positive rela-
tion between order imbalances and returns. Blume et al. (1989)
Lo and Coggins (2006) define the net order flow as the dif-
find that the relevance between order flow imbalance and return is
ference between buyer-and seller-initiated trading volumes. We
much stronger for large firms such as the firms listed on the S&P 500
follow Gregoriou (2008) and Lease et al. (1991), by computing the
index. They experience a much greater decrease in price because
within-spread location of the closing price for each zero-leverage
large firms suffer more from excess selling pressure.
firms’ stock i for each event day t as:

3. Data (ask − price)

Order flow ratio = (1)
(ask − bid)
3.1. Firm data
We assume that buyers purchase at the ask price where sellers
trade at the bid price. As Lease et al. (1991) mentioned, the nearer
Our dataset records the zero-leverage firms listed on the FTSE
order flow ratio is to 1, the more likely the transaction price is at
350, that have their first bank loan announcement between 1
the bid quote, indicating the selling pressure. While the closer this
January 2000 and 31 December 20151 . The data are collected
ratio is to 0 the greater the likelihood the trade is at the ask, an
from Thomson Reuters Eikon, a worldscope fundamental financial
indication of buying pressure.
database. The FTSE 350 firm is defined as the aggregation of the
FTSE 100 and FTSE 250 index listed firms, which consists of the 350
4.2. Event Return around the initial loan announcement
largest firms listed on the LSE with respect to market capitalization.
The FTSE 350 represents around 89 % of the entire trading volume
In general, we use the standard market model method for
on the LSE, enabling us to obtain a very accurate description of the
both transaction price based return and midpoint quote based
UK stock market.
return. We estimate the abnormal returns by using an economically
For each company we obtain the average closing price, bid price
market-adjusted model:
and ask price, trading volume (number of stocks traded), and mar-
ket valuation (total share outstanding multiple by share price) at AR i,t = R i,t -R m,t (2)
the month prior to the announcement. We also collect information
concerning the initial loan undertaken by the firm including the Where ARi,t is the abnormal return earned by stock i at time t. Ri,t is
the return on stock i at time t, and Rm,t is the value-weighted market
index return at time t. As a proxy for the market’s return, we use the
1
We require that firms are not financial companies, utility companies or closed-
FTSE index returns for the sample period. The announcement date
end mutual fund. The stock has enough data from 2000 to 2015, including trade (day 0) is the day of the initial bank loan announcement. In terms of
volume, average daily price, daily closing price, and bid-ask spread. the short-term period, we examine abnormal returns (AR) for the

629
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

Table 1
Descriptive Statistics.

Panel A: Borrower Descriptive Statistics

Zero-leverage firm All firm

Market Capitalization (£000,000) 3995 4231

Stock Price(£) 507.633 523.435
Standard deviation of return (%) 1.250 1.231
Trading Volume 6,325,000 6,520,000
Absolute Spread 2.537 2.031
Relative Spread (%) 0.015 0.011
Effective Spread (%) 0.412 0.369

Panel B: Loan Descriptive Statistics

Mean (£000,000) Median (£000,000) Std.Dev Min (£000,000) Max (£000,000) Skewness

Initial Loan 1537.29 248.92 7916.28 44.88 69049 7.94

Panel C:Order-initial by trading volume

Total Small trades Large Trades

Buys 28,342,080 16,880,654 11,461,426

Sells 26.101.600 11030300 15071300

Table 1 presents the summary statistics of 96 zero-leverage firms listed on the FTSE 350 that undertook an initial loan during the time period of 2000−2015. Panel A provides
the descriptive statistics on borrower firms. Market capitalization is calculated as the natural logarithm of the firm’s market capitalization measured by pounds. Absolute
spread is defined as ask price minus bid price. Relative spread is defined as ask price minus bid price, then divided by quote midpoint. Effective spread is defined as two times
trade price minus quote midpoint. The midpoint is ask price minus bid price divided by two. Panel B provides the disclosed amount to be borrowed by the zero-leverage
firm measured in pounds. Panel C presents the purchases and sales of sample firms for the period, including the total, small and large trades. We define the small trades as
the trade size smaller than 10,000 stocks and the large trades as the trades size of more than 10,000 shares.

announcement day [day 0], three-day period around the announce- Specifically, the regression takes the following forms:
ment day [-1, +1], five-day period around the announcement day
Ri,t = 0 +1 Initiali,t +2 Oi,t +3 Voli,t +4 Sizei,t +5 StdDevi,t
[−2, +2] and eleven-day period [-5, +5] around the announcement
day. We calculate the long event period as long as 181 days [-90, +6 Pricei,t +7 LAi,t +εi,t (6)
+90] around the announcement day.
As we have discussed in Section 2, in an attempt to eliminate
the return biases which are induced by order flow imbalances, we Ri,t ’ = 0 +1 Initiali,t +2 Oi,t +3 Voli,t +4 Sizei,t +5 StdDevi,t
calculate the return based on midpoint price as well.
+6 Pricei,t +7 LAi,t +εi,t (7)
AR’i,t = R’i,t -R m,t (3) Where t = 1 corresponds to the pre-announcement period and t = 2
corresponds to the post-announcement period. The subscript i
We have defined R as returns calculated using closing prices, now
means 96 initial loan announcement of zero-leverage sample firms
we introduce R’ as returns calculated using the midpoint price. R
in our research. The dependent variables, R and R’ is the closing
contains pricing errors due to the bid-ask spreads. R’, however, is
price-based return and midpoint-based return, for stock i at time t,
constructed using only the midpoint price suggesting that it will
respectively. Independent variables include Oi,t which is the order
not be under the influence of the bid-ask spread errors. We can
flow ratio for stock i at time t. Initiali,t , defined as the dummy vari-
thus construct tests to evaluate both the R and R’ and obtain a direct
ables which is equal to one if the trading occurs after the initial
estimate of the relative importance of asymmetric price impact and
loan announcements otherwise it is equal to zero. Voli,t is the nat-
bid-ask errors.
ural logarithm of the average daily volume of trading in stock i at
Cumulative abnormal return (CAR) of the event window can be
time t, Sizei,t is the natural logarithm of the market capitalization
calculated using the following formulas:
of zero-leverage firm i at time t, StdDevi,t is the natural logarithm of

s the stock prices’ standard deviation, which captures the volatility
CAR i,t(-q,+q) = ARi,t (4) in the true price for stock i at time t. Pricei,t is the natural loga-
t=q rithm of the daily average share price of stock i at time t, and L/Ai,t
is the natural logarithm of the loan amounts to total assets ratio.

s
All the firm-level variables are computed from one month before
CAR’i,t(-q,+q) = AR’i,t (5) the announcement. We are primarily interested in the coefficient
t=q of Oi,t , which we predict should have a negative relationship with
stock returns.
Where q is equals to 1, 2, 5, 10, 20,. . .. . ..90 respectively, represents
the different time period during the event window. The standard
5. Empirical results
t-test is used to test whether the ARi,t , CARi,t , AR’i,t , and CAR’i, differ
significantly from zero.
5.1. The price effects associated with initial bank loan

4.3. Financial Returns and order flow ratios First, in order to examine the effects of initial loan announce-
ment on the zero-leverage firms stock price, we measure the
We use the ordinary least squares method to obtain the cross- abnormal return based on both the closing price and midpoint price.
sectional variables of both closing price and midpoint returns for We observe the estimation results in Table 2, the effect of the ini-
zero-leverage firms around the initial loan announcement period. tial loan announcement on the zero-leverage firms is significantly

630
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

Table 2 zero-leverage firm’s stock. Overall, whether we use the quote mid-
The Price Effect Associated with Initial Bank Loan of Zero-leverage Firm.
point or closing price-based estimate of returns we observe the
Event Day Panel A: Closing price Panel B: Midpoint positive abnormal return after the initial loan announcement. The
returns(N = 96) returns (n = 96) results indicate that using an alternative return method to isolate
AR (%) T-stat AR (%) T-stat the microstructure bias, causes the same direction results with the
different magnitude. After we purged the bias due to the bid-ask
(-90,0) 0.167 1.72* 0.103 1.79*
(-80,0) 0.212 2.60** 0.000 1.92* bounce, the remaining loan announcement returns are still posi-
(-70,0) −0.327 −3.57** −0.024 −2.83** tive, which could be caused by the positive news and the liquidity
(-60,0) 0.519 5.46*** 0.445 2.34** effect.
(-50,0) 0.392 4.08*** 0.288 2.77**
(-40,0) 0.273 2.86** 0.269 1.69*
(-30,0) 0.256 2.67** 0.244 2.43** 5.2. Trades tendency and order flow ratio
(-20,0) 0.135 1.43* 0.120 1.96*
(-10,0) 0.167 1.56* 0.031 0.95
Table 3 presents the percentage of closing price occurring at
−5 0.268 2.78** 0.028 0.79
−4 0.196 2.08* 0.177 1.91* the ask, bid and at the midpoint for the 181-day event window.
−3 0.103 1.14 0.214 2.26** After the initial loan has been announced, the tendency of closing
−2 0.449 4.27*** 0.417 3.93*** prices to be at the ask price are highly significant. The percentage
−1 0.517 4.99*** 0.633 5.98*** of closing transaction prices occurring at the ask quote are larger
0 2.394 12.20*** 2.365 12.84***
1 1.645 10.14*** 1.870 9.13***
than at the bid quote from 30 days before the announcement date.
2 1.395 9.73*** 1.908 9.25*** In the event day, the percentage of closing prices occurring at the
3 0.816 7.49*** 1.000 8.97*** ask quote and bid quote is 42.58 and 30.70 respectively. This indi-
4 0.657 5.92*** 0.734 6.88*** cates that there is a decrease in sell orders and an increase in buy
5 0.565 5.69*** 0.518 5.28***
orders on the day of the initial loan announcement. Buy orders are
(-1, +1) 4.556 16.06*** 4.868 14.99***
(-2, +2) 6.400 17.55*** 7.193 16.87*** related with some good news rather than bad news. Large one-day
(-5,+5) 9.005 19.37*** 9.564 18.73*** price increases are occurring on the announcement day. These are
(0,10) 0.258 2.25** 0.297 2.41** likely to be associated with substantial buying pressure, enhanc-
(0,20) 0.428 3.91*** 0.490 3.94*** ing the probability that a closing transaction is at the ask price and
(0,30) 0.152 1.58* 0.233 2.59**
(0.40) 0.161 1.74* 0.258 8.27***
leading to a continuation during the next day. We obtain similar
(0,50) 0.170 1.76* 0.186 1.98* findings for small trades (Panel B) and large trades (Panel C). On
(0,60) 0.151 1.52* 0.174 1.77* the announcement day for small trades 43.60 % of trades occur at
(0,70) 0.319 2.99** 0.481 4.18*** the ask price and 33.89 % of trades are at the bid price. For large
(0,80) 0.167 1.43* 0.198 1.77 *
trades, 49.25 % of trades are at the ask price and 37.17 % of trades
(0,90) 0.155 1.32 0.243 2.66**
are at the bid price. In the one day after the announcement, 44.06 %
Table 2 provides the daily average returns (AR) from the pre-announcement to the
and 48.96 % of trades occur at the ask quote for the small and large
post-announcement period calculated by closing-price (Panel A) and midpoint price
(Panel B), which provides the bid-ask bias purged return. The sample consists of 96
trades group respectively, while 33.06 % and 36.99 % are at the bid
zero-leverage firms listed on the FTSE 350 that undertook an initial bank loan during quote. Our results are consistent with Chan and Lakonishok (1993)
the time period of 2000−2015. T-statistics are presented to show if sample loans’ who mention that buying a stock is associated with favorable firm-
AR are significantly different from zero. Event day (day zero) presents the day of the
bank loan announcement. Two tailed tests of significance are reported as follows,
***less than 1 %, **less than 5 % and * less than 10 %. Table 3
Closing Prices relative to Bid and Ask Quotes surrounding the Initial Loan Announce-
ment Date of Zero-leverage Firms.

positive from the pre-announcement period. For stock return cal- Panel A: ALL TRADES
culated by closing price, the abnormal returns are positive from 60
Distribution of Closing Prices
days prior to the announcement date, indicating the possibility of
information leakage before the loan announcement. The abnormal Day At ask At bid Order flow ratio
returns achieve the peak at the event date, 2.394, and are statisti- −90 29.80 37.64 53.06
cally significant. In the event day +1, the abnormal return is 1.645 −60 30.66 34.65 51.06
with a t-test equal to 10.14. During the 5 day (-2, +2) and 11 day (-5, −30 39.22 34.89 48.91
−10 39.67 37.59 49.28
+5) period around the announcement date, the cumulative return −5 34.34 31.81 48.69
is 6.400 and 9.005 respectively. −4 38.22 32.96 48.33
Panel B of Table 2 reproduces the event study in Panel A −3 41.01 35.27 47.01
using daily abnormal returns calculated from the midpoint quote −2 39.67 34.90 47.29
−1 36.66 29.74 45.13
between the bid and ask price. There is clear evidence that the bid-
0 42.58 30.70 41.66
ask effects and order flow ratio have some implications on the stock 1 42.34 31.41 41.97
markets. On the initial loan announcement date, we observe an 2 41.66 31.62 42.00
abnormal return of 2.365 percentage (t-statistic of 12.84), and in 3 40.08 29.02 41.78
the first trading day after announcement, the abnormal return is 4 39.87 30.61 43.26
5 40.69 31.72 43.93
1.870 (t-statistic of 9.13). During the post-announcement period
10 37.34 31.01 46.09
for midpoint-based abnormal return, we observe relatively larger 30 37.66 31.12 45.24
positive returns than transaction-based returns. The average cumu- 60 37.07 31.87 47.24
lative abnormal return of 4.868 is observed over the 3-event day 90 36.01 30.90 47.25
Panel A: Test of mean order flow ratio on Day
interval [-1, +1], with a highly significant t-value. Event windows [-
0 to the Comparison period
2, +2] and [-5, +5] have CARs of 7.193 and 9.564 respectively which Pre-announcement Post-announcement
are both highly significant. According to Lease et al. (1991), it can be period period
concluded that after the firm-level favorable news announcements, T-test 4.88*** 9.74***
the market participants start to increase their inventory level in the

631
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

Table 3 (Continued) small trades in Panel B, the mean ratio in our event period is gen-
Panel B: SMALL TRADES erally smaller than 0.5, indicating that the traders tend to buy
shares with small volume. However, in Panel C (large trades) 10
Distribution of Closing Prices
days prior to the announcement the order flow ratio is 51.16, indi-
Day At ask At bid Order flow ratio cating more sell orders than buy orders. According to Lee (1992),
−90 34.24 30.23 49.17 the order flow imbalance before the news announcement can be
−60 36.48 30.65 48.32 regarded as a measurement of the information leakage, as it mea-
−30 36.93 31.69 48.96 sures how quickly market participants adapt to the news. In our
−10 37.26 31.61 48.45
study, the large traders respond to the news from 1 day before the
−5 35.77 29.74 48.53
−4 39.64 32.85 47.66 announcement whereas the small traders’ react on the announce-
−3 40.66 33.62 47.28 ment day.
−2 40.01 32.02 46.59 Comparing the order flow ratio between small and large trades,
−1 37.49 30.23 45.33 large traders are associated with stronger trading reactions. This
0 43.60 33.89 43.26
1 44.06 33.06 42.04
is because the high buying activities concentrate more around
2 45.78 36.52 43.58 the announcement day. For example, the 3-day period during
3 44.63 34.74 43.16 the announcement for large trades is 42.03, 40.18 and 40.97,
4 43.53 32.90 42.97 whereas for small trades in Panel B it is 45.33, 43.26 and 42.04
5 44.71 33.68 42.04
respectively. The response of large trades is more dramatic than
10 42.78 33.74 43.63
30 40.69 31.92 44.29 the small trades, and the reactions of small traders are slower
60 38.86 30.78 46.47 than the large traders. The statistical test shows these results
90 37.43 29.14 46.04 (the order flow imbalance) are significantly different from both
Panel B: Test of mean order flow ratio on Day the pre-and post-announcement period based on the t-test at
0 to the Comparison period
Pre-announcement Post-announcement
the 1% significance level. It indicates that after the initial loan
period period announcement, buy orders are much larger than the sell orders,
T-test 3.14** 6.88*** and it does not subside during the short term. Following the ini-
tial loan of zero-leverage firm announcement, stock market buying
Panel C: LARGE TRADES increases, creating an upward trend in closing prices and daily
Distribution of Closing Prices returns.
Day At ask At bid Order flow ratio
5.3. Financial returns and firm characteristics
−90 30.06 38.51 54.65
−60 31.24 37.32 53.24
We apply the regression method in order to address the order
−30 37.85 37.73 50.00
−10 37.64 34.54 51.16 flow imbalance that may be evident in the relation between stock
−5 39.03 36.97 49.37 returns and initial loan announcement. Table 4 reports that the
−4 39.99 33.70 48.94 closing price-based returns have a strong negative correlation with
−3 41.80 34.71 46.01
the order flow ratio. This is reflected by the coefficient of -0.057
−2 42.93 35.29 45.68
−1 43.11 33.12 42.03 with a t-statistic of -4.69. As we have previously mentioned, the
0 49.25 37.17 40.18 nearer this ratio is to 0, the more likely the closing price is at the
1 48.96 36.99 40.97 ask price. The negative coefficients between order flow ratio and
2 48.27 37.42 41.04 stock returns indicate that the more likely the trades will be exe-
3 47.95 35.00 40.13
cuted at the ask price, the larger return of stocks. For mid-point
4 46.44 36.80 44.21
5 40.08 31.11 44.56 returns which are based on the midpoint price, the coefficient is -
10 39.04 32.19 47.48 0.042 with a t-statistic equal to -2.14, which is slightly smaller than
30 39.85 32.85 46.96 the closing-based return, a 1% decrease of order flow ratio is associ-
60 38.37 31.59 48.24
ated with an increase of 4.2 % in the average mid-point price-based
90 38.24 31.74 48.38
Panel C: Test of mean order flow ratio on Day
returns. It implies that when we purged the bid-ask bias by measur-
0 to the Comparison period ing the return by midpoint, the negative association between order
Pre-announcement Post-announcement flow and stock returns still exist. But the magnitude of coefficient
period period implies the correlation between order flow ratio and mid-point-
T-test 4.96*** 11.03***
based returns is smaller than the correlation between order flow
Table 3 reports the event day relative to the announcement day (day 0), the percent- ratio and closing price returns.
age distribution of closing prices in relation to the bid and ask quotes, and order flow
The findings for small and large trade subsamples are in the
ratios for the 96 FTSE 350 listed zero-leverage firms that undertook an initial bank
loan during the time period of 2000−2015. The results for all trades are reported in Panel B and C respectively. For small trades, the coefficient between
Panel A, the small and large trades results are shown in Panel B and Panel C respec- order flow and closing price returns is -0.011 with a t-statistics of
tively. The order flow ratio is calculated as (ask price-trading price)/(ask price – bid -1.69, for mid-point-based returns, the coefficient is -0.010 with
price). Two tailed tests of significance are reported as follows, ***less than 1 %, **less a t-statistic of -1.58. There are no significant differences between
than 5 % and * less than 10 %.
closing price-based return and the mid-point price return. In Panel
C, the coefficients of order flow variables are significant for large
specific news, since it reflects the choice of one specific security out trades, -0.106 (t-value =-5.43) for closing price return and -0.094
of many. (t-value=-3.68) for mid-point returns. Also the large traders show
The last column of Table 3 presents the order flow ratio. We cal- a more pronounced difference between price-based returns and
culate the average ratio of (ask-close)/(ask-bid) for the individual quote-based returns.
stocks on each event day surrounding the initial loan announce- The results also suggest that the dummy variable representing
ment. In Panel A, on the announcement date, the mean ratio initial loan announcements appears to have a positive relationship
decreases from 45.13 on day -1 to 41.66 and continues to be smaller with both the closing price-based and mid-point-based return. It
than 0.5 (nearer to 0) during the post-announcement period. For indicates that the initial loan announcement has a positive influ-

632
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

Table 4 Table 5
Regression between Transaction Returns and Firm Variables by Quote Price Effects. Robustness Test: The Price Effect Associated with Initial Bank Loan of Zero-leverage
Firm.
Closing price return Mid-point return
Event Day Panel A: ask-to-ask Panel B: bid-to-bid
Coef. T-stat Coef. T-stat returns (n = 96) returns (N = 96)
Panel A: All trades
AR (%) T-Stat AR (%) T-Stat
Constant −3.066 −11.02*** −2.533 −9.58***
O −0.057 −4.69*** −0.042 −2.14* (-90,0) 0.348 3.40*** 0.171 1.86*
Initial 0.023 2.05* 0.018 1.44* (-80,0) 0.259 2.66** 0.397 3.45***
Vol 0.054 4.18*** 0.004 1.21 (-70,0) 0.230 2.07* −0.268 −2.43**
Size 0.000 1.22 0.007 1.31 (-60,0) 0.615 5.11*** 0.326 3.21***
StdDev −0.012 −3.03** −0.001 −1.09 (-50,0) 0.427 4.18*** 0.267 2.68**
Price 0.035 2.37** 0.001 1.10 (-40,0) 0.218 2.69** 0.032 1.22
L/A 0.000 0.98 0.000 1.04 (-30,0) 0.305 3.24*** 0.167 1.78*
(-20,0) 0.233 2.17* 0.278 2.82**
Panel B: Small trades
(-10,0) 0.029 1.12 0.188 1.74*
Constant −2.713 −7.64*** −2.085 −6.88***
−5 0.297 2.90** 0.196 1.98*
O −0.011 −1.69* −0.010 −1.58*
−4 0.243 2.42** 0.102 1.19
Initial 0.036 2.51** 0.032 2.46**
−3 0.155 1.23 0.072 0.93
Vol 0.047 3.06** 0.009 1.42
−2 0.481 4.21*** 0.363 3.83***
Size 0.002 1.04 0.011 1.61*
−1 0.609 5.09*** 0.445 4.17***
StdDev −0.016 −1.78* −0.000 −0.76
0 2.883 13.25*** 2.315 11.19***
Price 0.024 2.74** 0.007 1.37
1 2.129 11.37*** 1.233 8.01***
L/A 0.000 0.73 0.000 0.52
2 1.049 9.14*** 0.790 6.80***
Panel C: Large trades 3 0.868 7.78*** 0.712 6.61***
Constant −1.069 −8.69*** −2.140 −5.28*** 4 0.633 5.38*** 0.610 5.34***
O −0.106 −5.43*** −0.094 −3.67*** 5 0.615 5.12*** 0.558 4.93***
Initial 0.020 1.86* 0.021 2.13* (-1, +1) 5.621 7.23*** 3.993 13.27***
Vol 0.061 3.15** 0.007 1.25 (-2, +2) 7.151 16.00*** 5.546 15.77***
Size 0.000 1.04 0.002 0.87 (-5,+5) 9.965 19.18*** 7.796 18.81***
StdDev −0.003 −1.20 −0.000 −0.46 (0,10) 0.280 2.90** 0.117 1.42*
Price 0.033 2.22** 0.001 0.73 (0,20) 0.346 3.78*** 0.225 2.69**
L/A 0.000 0.42 0.000 0.61 (0,30) 0.288 2.75** 0.284 2.95**
(0.40) 0.390 4.06*** 0.297 2.97**
Table 4 provides the daily average returns (AR) from the pre-announcement to (0,50) 0.061 1.82* 0.351 3.74***
the post-announcement period calculated by both ask quote (Panel A) and bid (0,60) 0.367 3.86*** 0.448 4.90***
quote (Panel B) which provides the bid-ask bias purged return. The sample consists (0,70) 0.678 5.83*** 0.102 1.33
of 96 FTSE 350 listed firms that undertook an initial bank loan during the time (0,80) 0.188 1.78* 0.161 1.79*
period of 2000−2015. T-statistics are presented to show if sample loans’ AR are (0,90) 0.196 2.01* 0.238 2.43**
significantly different from zero. Event day (day zero) presents the day of the bank
loan announcement. Two tailed tests of significance are reported as follows, ***less Table 5 provides the daily average returns (AR) from the pre-announcement to the
than 1 %, **less than 5 % and * less than 10 %. reports the relationship between stock post-announcement period calculated by both ask quote (Panel A) and bid quote
return and firm-level variables, for 96 zero-leverage firms in FTSE 350 index, over (Panel B) which provides the bid-ask bias purged return. The sample consists of 96
the period from 1 January 2000 to 31 December 2015, for both closing price-based FTSE 350 listed firms that undertook an initial bank loan during the time period
return R and mid-quote-based return R’. Panel A reports the results for all trades, of 2000−2015. T-statistics are presented to show if sample loans’ AR are signifi-
the findings of small and large trades are shown in panel B and C respectively. The cantly different from zero. Event day (day zero) presents the day of the bank loan
regression models estimated are the following: announcement. Two tailed tests of significance are reported as follows, ***less than
Ri,t =␮0 +␮1 Initiali,t +␮2 Oi,t +␮3 Voli,t +␮4 Sizei,t +␮5 StdDevi,t +␮6 Pricei,t +␮7 LAi,t +εi,t 1 %, **less than 5 % and * less than 10 %.
Ri,t ’=␮0 +␮1 Initiali,t +␮2 Oi,t +␮3 Voli,t +␮4 Sizei,t +␮5 StdDevi,t +␮6 Pricei,t +␮7 LAi,t +εi,t.
The subscript i means 96 initial loan announcement of zero-leverage sample
firms. The dependent variables, R and R’ is the closing price-based return and 5.4. Robustness test
midpoint-based return, for stock i at time t, respectively. Independent variables
include Initiali,t , defined as the dummy variables which is equal to one if the trading After the measurement of stock return by both transaction price
occurs after the initial loan announcements otherwise it is equals to zero. Oi,t refers return and midpoint return, we employ the ask-to-ask return, and
to the order flow ratio for stock i at time t. Voli,t is the natural logarithm of the
bid-to-bid return for robustness in Table 5. Overall, the asymmetry
average daily volume of trading in stock i at time t. Sizei,t is the natural logarithm
of the market capitalization of zero-leverage firm i at time t, StdDevi,t is the natural within the quote price return always exists regardless of the event
logarithm of standard deviation of the stock prices as to capture the volatility in the announcements. For example, during the event day the abnormal
true price for stock i at time t. Pricei,t is the natural logarithm daily average share returns based on ask price is 2.883 whereas the bid price return is
price of stock i at time t, and L/Ai,t is the natural logarithm of the loan amounts to
2.315. On the first trading day after the announcement, the abnor-
total assets ratio. Two tailed tests of significance are reported as follows, ***less
than 1 %, **less than 5 % and * less than 10 %.
mal return is 2.129 for ask price return whereas it is 1.233 for
bid price return. According to Zhang and Gregoriou (2019a), the
buy order is motivated by both the information and liquidity fac-
tors before the initial loan announcement. Market makers trade
ence on the stock return, even after controlling for the order flow according to their requirements based on both private informa-
ratio. However, the results for mid-quote returns are generally less tion and liquidity. During the announcement period, the favorable
significant than the transaction-based return, suggesting that when firm-level news attracts more buy orders leading to the liquidity
we purged the bid-ask bias and controlling other firm-level vari- increasing, resulting in low bid-ask spread. The ask-to-ask abnor-
ables, the stock has little reaction to the initial loan announcement. mal returns in Table 5 are larger than bid-to-bid returns. These
Moreover, the results show that the coefficients of the standard effects are permanent because the increasing liquidity persists in
deviation which measures the volatility of midpoint return are the long-term.
smaller than the closing-price based return (-0.012 for closing price
return whereas -0.001 for midpoint price return). Confirming the 6. Conclusion
results of Han and Lesmond (2011) and many other researchers,
closing price-based idiosyncratic volatility is usually larger than In this paper we empirically examine the initial loan announce-
that of the quote midpoint-based volatility. ment period of 96 zero-leverage firms listed on the FTSE 350 index.

633
S. Zhang, A. Gregoriou The Quarterly Review of Economics and Finance 80 (2021) 627–634

Our research demonstrates that there is a clear tendency that trades Berger, A. N., & Udell, G. F. (1995). Relationship lending and lines of credit in small
are executed at the ask price during the initial loan announce- firm finance. Journal of Business, 68(3), 351–381.
Bessler, W., Drobetz, W., Haller, R., & Meier, I. (2013). The international
ment period, which are regarded as favorable firm-level news. zero-leverage phenomenon. Journal of Corporate Finance, 23, 196–221.
Order flow disruption causes a bias in the calculation of returns Bissoondoyal-Bheenick, E., Brooks, R., & Do, H. X. (2019). Asymmetric relationship
around the company event announcement. Buying pressure affects between order imbalance and realized volatility: Evidence from the Australian
market. International Review of Economics & Finance, 62, 309–320.
the stock performance during the event period, reflected by the Blume, E. M., MacKinlay, C., & Terker, B. (1989). Order imbalances and stock price
price increases and spreads decrease. Small traders respond dif- movements on October 19 and 20, 1987. The Journal of Finance, 44(4), 827–848.
ferently from large traders to the same information signals. Large Brown, P., Walsh, D., & Yuen, A. (1997). The interaction between order imbalance
and stock price. Pacific-Basin Finance Journal, 5(5), 539–557.
trades respond to the news quickly whereas the small trades show
Chan, L. K. C., & Lakonishok, J. (1993). Institutional trades and intraday stock price
a puzzling reaction during the announcement. The empirical evi- behavior. Journal of Financial Economics, 33(2), 173–199.
dence shows that the order flow imbalance does not subside during Chordia, T., & Subrahmanyam, A. (2004). Order imbalance and individual stock
returns: Theory and evidence. Financial Economics, 72(3), 485–518.
the short term in the post-announcement period. Furthermore,
Chordia, T., Hu, J. F., Subrahmanyam, A., & Tong, Q. (2019). Order flow volatility and
the order flow ratio is negatively related to the stock return. As equity costs of capital. Management Science, 65(4), 1520–1551.
the order flow ratio goes towards zero, this is related with more Dang, V. A. (2013). An empirical analysis of zero-leverage firms: New evidence
buy orders, causing the price and return to increase. Our esti- from the UK. International Review of Financial Analysis, 30, 189–202.
Ghoul, S. E., Guedhami, O., Kwok, C., & Zheng, X. (2018). Zero-leverage puzzle, An
mation process is based on a large sample with approximate 28 international comparison. Review of Finance, 22(3), 1063–1120.
million share purchases and 26 million share sells, over the time Gosnell, T. F., Keown, A. J., & Pinkerton, J. M. (1996). The intraday speed of stock
of period 2000–2015. These phenomena may explain several sig- price adjustment to major dividend changes: Bid-ask bounce and order flow
imbalances. Journal of Banking & Finance, 20(2), 247–266.
nificant event study results that have been documented in recent Gregoriou, A. (2008). The asymmetry of the price impact of block trades and the
years. It shows that the stock market liquidity reacts strongly when bid-ask spread: Evidence from the London Stock Exchange. Journal of Economic
a bank reports favorable news via the announcement of the initial Studies, 35(2), 191–199.
Hadlock, C. J., & Pierce, J. R. (2010). new evidence on measuring financial
loan agreement. We conclude that the order flow imbalance can constraints: Moving beyond the KZ index. The Review of Financial Studies, 2(5),
explain part of bid-ask spread bias and asymmetry, in the trading 1909–1940.
direction of zero-leverage firms when they encounter debt for the Han, Y., & Lesmond, D. (2011). Liquidity biases and the pricing of cross-sectional
idiosyncratic volatility. The Review of Financial Studies, 24(5), 1590–1629.
first time.
Hasbrouck, J., & Seppi, D. J. (2001). Common factors in prices, order flows and
liquidity. Journal of Financial Economics, 59(3), 383–411.
Appendix A. The frequency of the zero-leverage adoptions Huang, H., Tsai, W., Weng, P., & Wu, M. (2020). Volatility of order imbalance of
institutional traders and expected asset returns: Evidence from Taiwan. Journal
of loan debt by year
of Financial Markets, in press.
Kaul, G., & Nimalendran, M. (1990). Price reversals: Bid-ask errors or market
Year Frequency overreaction? Journal of Financial Economics, 28(1-2), 67–93.
Kim, S., & Stoll, H. R. (2014). Are trading imbalances indicative of private
2000 28 information? Journal of Financial Markets, 20, 151–174.
2001 6 Lease, R. C., Masulis, R. W., & Page, J. R. (1991). An investigation of market
2002 7 microstructure impacts on event study returns. The Journal of Finance, 46(4),
2003 12 1523–1536.
2004 11 Lee, C. M. C. (1992). Earnings news and small traders: An intraday analysis. Journal
2005 3 of Accounting and Economic, 15(2-3), 265–302.
2006 5 Lee, C. M. C., & Ready, M. J. (1991). Inferring trade direction from intraday data. The
Journal of Finance, 46(2), 733–746.
2007 4
Levi, S., & Zhang, X. (2015). Asymmetric decrease in liquidity trading before
2008 4
earnings announcements and the announcement return premium. Journal of
2009 0
Financial Economics, 118, 383–398.
2010 2 Lo, K., & Coggins, R. (2006). Effects of order flow imbalance on short-horizon
2011 5 contrarian strategies in the Australian equity market. Pacific-Basin Finance
2012 2 Journal, 14(3), 291–310.
2013 3 Lummer, S. L., & McConnel, J. J. (1989). Further evidence on the bank lending
2014 2 process and the capital-market response to bank loan agreements. Journal of
2015 2 Financial Economics, 25(1), 99–122.
Strebulaev, I. A., & Yang, B. (2013). The mystery of zero-leverage firms. Journal of
Financial Economics, 109(1), 1–23.
Declaration of Competing Interest Zhang, S., & Gregoriou, A. (2019a). Initial bank loans, zero-leverage firms and stock
market liquidity: New empirical evidence from the UK. Journal of Economic
Studies, 46(5), 1028–1051.
The authors report no declarations of interest.
Zhang, S., & Gregoriou, A. (2019b). The price behavior around initial loan
announcements: Evidence from zero-leverage firms in the UK. Research in
References International Business and Finance, 50, 191–200.
Zhang, T., Gu, G., & Zhou, W. (2019). Order imbalances and market efficiency: New
evidence from the Chinese stock Market. Emerging Markets Review, 38,
Alzahrani, A. A., Gregoriou, A., & Hudson, R. S. (2013). Price impact of block trades
458–467.
in the Saudi stock market. Journal of International Financial Markets Institutions
and Money, 23(1), 322–341.

634