Finance Research Letters xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Finance Research Letters

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

The effect of liquidity on non-marketable securities☆

⁎
Menachem (Meni) Abudya, Hadar Binskyb, Alon Raviva,
a
Bar-Ilan University, School of Business Administration, Ramat-Gan, 5290002, Israel
b
Eitan Berglas School of Economics, Tel Aviv University, Tel-Aviv, 6997801, Israel

AR TI CLE I NF O AB S T R A CT

Keywords: We generalize the prevailing theoretical models that estimate the discount on securities for lack
Non-marketability discount of marketability, by considering the discrete trading frequency of the securities. The general-
Illiquidity ization shows that accounting for the illiquidity of securities may significantly reduce their non-
Thin-traded securities marketability discount. Further, the method reconciles the approaches of Longstaff (1995) and
JEL classification: Finnerty (2012a), which are special solutions of our method.
G01
G12
G13

1. Introduction

A non-marketability discount is the compensation that investors demand for being unable to trade a security compared to an all-
else-equal security with no such constraint. Legal and contractual restrictions on trading are common in financial markets. Examples
include stock lockups in IPOs (Brav and Gompers, 2003; Ding et al., 2009), merger and acquisition agreements (Li et al., 2016),
equity-based compensation (Abudy and Benninga, 2013), and private placements of publicly traded companies, under Rule 144 of the
Securities and Exchange Commission (SEC) (see Bajaj et al., 2001). Longstaff (1995) calculates the upper bound for the non-mar-
ketability discount by assuming that without trading restrictions, an investor with perfect market-timing ability can sell a security at
its maximum price during the restricted trading period. Thus, the upper bound for the non-marketability discount is priced as the
difference between this maximum price during the restricted trading period and the security price at the end of this period.
Finnerty (2012a) assumes that in the absence of trading restrictions, an investor is equally likely to sell a security at any time during
the restricted trading period. Hence, the discount is calculated relative to the average value of the security during this period.1
Both Finnerty's (2012a) and Longstaff's (1995) models make the unrealistic assumption that securities are continuously traded in
frictionless markets. However, in practice, in most cases in which a limitation on marketability exists, securities are already illiquid.2
We bridge this gap by adjusting the non-marketability discount for illiquidity risk. First, we relate to the frequency of trading, where a
security may be traded in various trading frequencies (Kalay et al., 2004). Exogenous factors, such as the trading mechanism, could
result in lower frequency trading, which may affect the marketability discount (Kalay et al., 2002). Allen et al. (2012) find that the
presence of dual market makers in the equity market and in the over-the-counter (OTC) loan market improves the liquidity of the

☆
We thank Olaf Korn, Sharon Peleg-Lazar and Efrat Shust for helpful suggestions and the participants of the 6th International Conference of the Financial
Engineering and Banking Society (FEBS).
⁎
Corresponding author.
E-mail addresses: menachem.abudy@biu.ac.il (M. (Meni) Abudy), hadarbin@post.tau.ac.il (H. Binsky), alon.raviv@biu.ac.il (A. Raviv).
1
The paper refers to the marketability discount of the firm's shares, and therefore the terms security and stock are used interchangeably. For the case of a corporate
bond a structural model should be introduced as in Abudy and Raviv (2016).
2
For example, in the sample analyzed by Finnerty (2012b), one-third of the placements were quoted in the National Association of Securities Dealers Automated
Quotation (NASDAQ) small cup, the OTC Bulletin Board, or Pink Sheets markets.

https://doi.org/10.1016/j.frl.2017.12.017
Received 14 November 2017; Accepted 29 December 2017
1544-6123/ © 2018 Elsevier Inc. All rights reserved.

Please cite this article as: Abudy, M., Finance Research Letters (2018), https://doi.org/10.1016/j.frl.2017.12.017
M. (Meni) Abudy et al. Finance Research Letters xxx (xxxx) xxx–xxx

more competitive and transparent equity market, but widens the spread of the less competitive OTC market. Second, we introduce a
limit on the traded quantity, where the informed investor with perfect market-timing ability has to sell his holding in several trans-
actions to avoid a potential impact on its fundamental price, as described in Dyl and Jiang (2008) and Albuquerque and
Schroth (2015).3 In this respect, the result of Jacoby and Zheng (2010) is consistent with the idea that higher ownership dispersion
improves market liquidity, indicating that stocks concentrated ownership firms are more illiquid.
We show that when illiquidity is introduced, the contributions of Longstaff (1995) and Finnerty (2012a) are special, corner
solutions of our general method. In the case where there are no limits on the traded quantity, a perfect-timer investor, in the absence
of trading restrictions, will sell a security at its maximum price during the restricted trading period. For this case, Longstaff (1995)
provides an upper bound for the non-marketability discount. At the other extreme, if an investor needs to sell his holding in a security
in an infinite number of transactions, the solution converges to the model of Finnerty (2012a), and the method provides a lower
bound for the non-marketability discount. While these two models may accurately describe the non-marketability discount of a
security with very specific liquidity conditions, our method covers a wider array of liquidity conditions that can fit a variety of real-
world circumstances. Moreover, our approach presents the difference between Longstaff's (1995) and Finnerty's (2012a) discounts as
a result of different assumptions on the security's liquidity and not due to the timing abilities of the investors.
Using our general method, one can accurately estimate the non-marketability discount while considering the liquidity of a se-
curity. For example, Dyl and Jiang (2008) analyze a case study of an estate owning 396,000 unregistered shares of the Elite Financial
Corporation, which had a mean daily trading volume of 4113 shares during the study period. They estimate that the non-market-
ability discount using Longstaff's (1995) model is 59%. However, as we shall see below, applying our method and considering the
limit on the traded quantity and on the frequency of trading yields a substantially lower non-marketability discount of 36.6%.

2. The framework of analysis

2.1. An upper bound for non-marketability discount

Our method generalizes the prevailing models, which assume that securities are traded continuously in frictionless markets, to a
discrete process. In a continuous framework, the value of the firm's security VT is governed by a geometric Brownian motion process
under a risk-neutral measure, as in Eq. (1) below:
dVT = rVT dt + σVT dWT (1)
where r is the risk-free rate, σ is the volatility of the firm's security, and WT is a standard Brownian motion process. Longstaff (1995)
assumes a hypothetical investor who is restricted from trading a specific security during a predetermined period t. With no trading
restrictions, an investor with perfect market-timing ability will sell the security at its maximum price during the restricted trading
period, at time τ, and will reinvest the proceeds at the risk-free rate till the end of the period. Hence, the payoff to the perfect-timer
investor at the end of the period denoted by Mt is
Mt = Max (e r (t − τ )·Vτ )
0≤τ≤t (2)
where Vτ denotes the maximum price of the security at time τ during the restricted trading period t (i.e., 0 ≤ τ ≤ t).
If the perfect-timer investor is restricted from trading, the result is a loss that is equal to the incremental value Max(0, Mt − Vt) ^
Mt − Vt at time t. The present value of this cash flow is the upper bound for the non-marketability discount:
e−rt ·(E [Mt − Vt ]) (3)

2.2. The impact of illiquidity on the non-marketability discount

A limit on the frequency of trading is introduced by shifting from the continuous process as in Eq. (1) to a discrete setting, where VT,
is expressed as
ΔVT = rVT Δt + σ ΔVT WT (4)
1
where Δ stands for a discrete frequency of trading. For example, assuming that the security is traded twice daily, then Δt = where 2·s
,
s is the number of trading days per year. The drift and volatility of the stock are set accordingly.
A limit on the traded quantity accounts for the case where an investor who trades a large quantity may impact the security price or
may not be able to execute the entire quantity in a single trade. Therefore, we consider the case where the trade is divided into several
transactions. An investor with perfect market-timing ability will choose to sell at the points of time that have the highest price during
the restricted trading period. The payoff, at the end of the period, is expressed as:

Mt = E ⎡ Max (e r (t − τ )·Vτ ) ⎤
⎣0≤τ ≤t ⎦ (5)

3
Our approach considers infrequent trading and not trade that results from asymmetric information. In fact, one can view Longstaff (1995) as relating to an extreme
case of illiquidity, since the investor with perfect market-timing ability is perfectly informed while all the other traders in the market have the same information. This is
the reason that Longstaff's model yields an upper bound for the non-marketability discount.

2
M. (Meni) Abudy et al. Finance Research Letters xxx (xxxx) xxx–xxx

Panel A:

36
Max

34

32

30
Stock price ($)

28

26

24

22

20
0 10 20 30 40 50 60
Restricted trading period

Panel B:

36

Max
34

32

30
Stock price ($)

28

26

24

22

20
0 10 20 30 40 50 60
Restricted trading period

Fig. 1. Simulating security price with one transaction each trading day. The figure plots the stochastic process of a security price. The security has one transaction each
trading day. The simulation is executed using a Geometric Brownian Motion volatility of 30%. Panel A shows the highest price of the security. Panel B shows the six
days with the highest security prices.

where τ denotes a vector from 1 to n that includes the times of the highest n values of V during the restricted trading period (e.g., n
can be regarded as the number of transactions), and Vτ is a vector that includes the maximum values of the security in this period. For
example, if the entire trade splits into three transactions, and the security receives its maximum price at times τ1 τ2 τ3, then
τ = {τ1, τ2, τ3} , and Vτ = {Vτ1, Vτ 2, Vτ3} includes the maximum values of the security. The payoff at the end of the period is

3
M. (Meni) Abudy et al. Finance Research Letters xxx (xxxx) xxx–xxx

Table 1
The effect of the frequency of trading on the relative non-marketability discount of an illiquid security
The table presents the relative non-marketability discount of an illiquid security as a function of the security's frequency of trading and the restricted trading period.
The relative non-marketability discount is the non-marketability discount of an illiquid security relative to the discount calculated using Longstaff's (1995) model (in
percentage points). The calculations assume a security volatility of 30%. Frequency of trading is the number of daily transactions of the illiquid security. Restricted
trading period is the non-marketability period in days.

Frequency of trading (No. of daily transactions in the security)

1 2 4 8 24 48

Restricted trading period 1 49.7% 60.1% 69.2% 77.6% 86.0% 89.6%

2 60.5% 68.9% 77.4% 83.2% 89.5% 93.1%
3 65.7% 74.0% 80.9% 85.9% 91.2% 94.2%
5 71.7% 78.7% 84.8% 89.0% 93.4% 95.2%
7 75.7% 81.9% 87.0% 90.5% 94.5% 96.2%
10 78.9% 84.2% 88.8% 91.9% 95.6% 96.7%
20 84.3% 88.5% 92.4% 94.2% 96.7% 97.1%
30 87.0% 90.9% 93.2% 95.2% 97.7% 98.3%
40 88.4% 91.3% 94.1% 96.0% 97.2% 98.1%
50 89.6% 92.7% 94.7% 96.2% 97.6% 98.3%
100 93.0% 94.2% 95.7% 97.4% 98.1% 99.5%
180 94.1% 95.6% 97.0% 98.1% 99.1% 99.4%
365 95.4% 97.1% 97.8% 98.2% 99.3% 99.7%

Table 2
The effect of the quantity of trading on the non-marketability discount of an illiquid security.
The table presents the effect of a limit on the traded quantity on the non-marketability discount of an illiquid security as a function of the security's frequency of
trading and the restricted trading period. The first row in the table presents the difference between the non-marketability discount of an illiquid security to the discount
calculated using Longstaff's (1995) model (in percentage points). The second row in the table presents the relative non-marketability discount, which is the non-
marketability discount of an illiquid security relative to the discount calculated using Longstaff's (1995) model (in percentage points). The calculations assume a
security volatility of 30%. Frequency of trading is the number of daily transactions of the illiquid security. Restricted trading period is the non-marketability period in
days. The effect of the limit on the traded quantity is expressed by the number of transactions that the investor trades in the illiquid security. The table refers to an
investor that needs to execute ten transactions to trade in her quantity. The missing cells represent scenarios with less than ten available transactions.

Frequency of trading (No. of daily transactions in the security)

1 2 4 8 24 48

Restricted trading period 1 0.7% 0.4%

(47.6%) (64.7%)
2 1.2% 0.6% 0.4%
(32.9%) (64.5%) (76.4%)
5 2.6% 1.7% 1.1% 0.6% 0.4%
(7.4%) (41.4%) (61.0%) (78.6%) (85.2%)
10 3.7% 2.4% 1.6% 1.1% 0.6% 0.4%
(7.5%) (40.8%) (60.7%) (73.2%) (85.6%) (89.8%)
20 3.4% 2.3% 1.5% 1.1% 0.6% 0.4%
(40.7%) (60.4%) (73.6%) (81.5%) (89.7%) (92.2%)
40 3.3% 2.3% 1.5% 1.0% 0.6% 0.4%
(59.8%) (72.3%) (81.3%) (87.2%) (92.2%) (94.7%)
50 3.2% 2.2% 1.5% 1.1% 0.6% 0.4%
(64.7%) (76.0%) (83.3%) (88.4%) (93.2%) (95.2%)
100 3.2% 2.3% 1.6% 1.1% 0.7% 0.4%
(75.9%) (82.6%) (87.7%) (91.8%) (95.0%) (97.3%)
180 3.3% 2.3% 1.6% 1.1% 0.5% 0.5%
(81.5%) (87.0%) (91.0%) (94.0%) (97.0%) (97.4%)
365 3.5% 2.4% 1.7% 1.3% 0.6% 0.4%
(86.5%) (90.9%) (93.5%) (95.1%) (97.5%) (98.5%)

Mt = E [e r (t − τ1)·Vτ1 + e r (t − τ2)·Vτ 2 + e r (t − τ3)·Vτ3]. Since there is no closed-form solution for the pricing equations, we solve them by using
a Monte Carlo simulation.4
Fig. 1 demonstrates the two trading effects of the perfect timer investor. The figure presents one specific realization of the value of
a security during a restricted trading period of 60 days. Panel A relates to the case of a limit on the frequency of trading and Panel B
relates to the case of a limit on both the frequency and the traded quantity. In Panel A, the investor can execute the trade in a single
transaction. However, since we assume that the stock is traded in a daily frequency, the maximum value of the stock is below its
continuous maximum and is equal to $34.49. Since the value of the security at the end of the period is $30.00, the potential loss of the

4
We run at least 50,000 simulation paths, till we reach a stable solution in the first four digits.

4
M. (Meni) Abudy et al. Finance Research Letters xxx (xxxx) xxx–xxx

Table 3
The non-marketability discount as a function of the frequency of trading and the limit on the traded quantity.
The table presents the non-marketability discount of an illiquid security as a function of the security's frequency of trading and the limit on the traded quantity. The
calculations use the real-world case of Dyl and Jiang (2008).

Number of daily transactions of the security

1 2 4 8 24 48

Number of quarterly transactions of investor during the non-marketability period 1 49.7% 51.8% 53.3% 54.0% 54.9% 55.8%
2 47.3% 50.1% 52.1% 53.2% 54.5% 55.4%
3 45.0% 48.5% 51.0% 52.4% 54.0% 55.1%
4 43.4% 47.4% 50.3% 51.9% 53.7% 54.9%
5 41.7% 46.2% 49.4% 51.3% 53.4% 54.7%
6 40.4% 45.3% 48.8% 50.9% 53.2% 54.5%
7 39.0% 44.3% 48.1% 50.4% 52.9% 54.3%
8 37.9% 43.5% 47.6% 50.0% 52.7% 54.2%
9 36.6% 42.6% 47.0% 49.6% 52.4% 54.0%
10 35.6% 41.9% 46.5% 49.3% 52.2% 53.9%
18 28.1% 36.7% 42.9% 46.8% 50.8% 52.9%
27 21.3% 31.9% 39.7% 44.6% 49.6% 52.0%
36 15.5% 28.0% 37.0% 42.7% 48.5% 51.2%

perfect timer investor is equal to $4.49. In Panel B, which demonstrates the effect of a limit on the traded quantity as well, the
investor must divide the trade into six different transactions, each made on a different trading day. Therefore, instead of having a
single maximum, there are six different transactions executed on these six days with the highest stock price. The average price of the
security on these days is $33.86, and the potential loss of the investor on this price path is $3.86, which is 14% less than in the case of
a single transaction. Note that the example shows only one simulation path of the security, and to calculate the effect of illiquidity on
the marketability discount one should simulate a large enough number of realizations over the restricted trading period.
The previously mentioned contributions of Longstaff (1995) and Finnerty (2012a) are special cases of our general method. First,
with continuous trading in the security, where Δt → dt, and an investor who can execute the sell in a single transaction, n = 1, the
model converges to the solution of Longstaff (1995). Second, with continuous trading, Δt → dt, and an investor who has to sell his
holding in an infinite number of transactions, n → ∞, the model converges to the solution of Finnerty (2012a). All other cases, which
represent different trading frequencies, are in between the upper bound of Longstaff (1995) and the lower bound of Finnerty (2012a).

3. Numerical analysis and implementation

3.1. The effect of the frequency of trading

Table 1 presents the non-marketability discount as a function of the security's frequency of trading and the length of the restricted
trading period. We defined the relative non-marketability discount as the ratio of the non-marketability discount of an illiquid security
relative to the discount of the non-marketability discount of a continuously traded security. As the frequency of trading increases, the
relative non-marketability discount, approaches a ratio of 1, regardless of the length of the restricted trading period. However, the
effect of the frequency of trading on the relative non-marketability discount decreases with the length of the non-marketability
period. The model results are in line with the empirical literature (Feng, Hung, & Wang, 2016) and demonstrate that prevailing
models overestimate the non-marketability discount when used for securities that trade infrequently.

3.2. The effect of a limit on the traded quantity

Table 2 examines the effect of a limit on the traded quantity, expressed as the number of transactions that the investor needs to
make to complete a trade in an illiquid security. We show that for a limit on the traded quantity that requires 10 transactions, the
marketability discount exhibits a similar pattern to the case where there is no limit on the traded quantity – as it increases with the
length of the restricted trading period and decreases with the frequency of trading. Thus, for a given restricted trading period, as the
frequency of trading increases, the relative non-marketability discount approaches a ratio of 1. However, the impact of thin trading is
much stronger in the presence of limits on the traded quantity, and this impact depends on the length of the restricted trading period.
The frequency of trading (i.e., the number of daily transactions) has a much lower effect on the relative non-marketability discount
for long restricted trading periods than for short restricted trading periods.

3.3. Applying the method: a real world case study

The effect of the limit on the traded quantity and on the frequency of trading is demonstrated by referring to a case study analyzed
by Dyl and Jiang (2008), who estimate the discount due to the lack of marketability of the unregistered shares of the Elite Financial

5
M. (Meni) Abudy et al. Finance Research Letters xxx (xxxx) xxx–xxx

Corporation for tax estate purposes in 1998.5 The estate owned 396,000 common shares of the firm. In 1998, the stock was traded on
NASDAQ, with 3.6 million outstanding shares, and with a mean (median) daily trading volume of 4113 (1900) shares.6 Given the SEC
regulation on unregistered stocks, the maximum number of shares that the estate was allowed to sell was 36,000 shares per quarter.
Dyl and Jiang (2008) estimate that the average time that the estate's shares were restricted from trading was 1.375 years and that the
stock's volatility was 60.5%. Using Longstaff's (1995) model the authors calculate a non-marketability discount of 59%.
The use of Longstaff's (1995) model overestimates the non-marketability discount since it assumes that the entire holdings of the
estate can be sold in a single transaction; however, under the SEC regulation and given the low daily trading in the stock, a sale of
396,000 securities would need to be divided into several transactions. With an average daily volume of 4113 shares, this means that
at least 9 transactions per quarter would be needed to sell the entire holdings of the estate, which leads to a non-marketability
discount of 36.6% (see Table 3). If the stock is traded four times a day, then 36 transactions would be needed per quarter, which
yields a discount of 37.0%. These levels are significantly lower than the discount of 59% calculated by Longstaff (1995) and sig-
nificantly higher than the 15% discount calculated by Finnerty (2012a), where, in both cases, the estate's holdings are sold in an
infinite number of transactions.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.frl.2017.12.017.

References

Abudy, M., Benninga, S., 2013. Non-marketability and the value of employee stock options. J. Banking Finance 37, 5500–5510.
Abudy, M., Raviv, A., 2016. How much can illiquidity affect corporate debt yield spread? J. Financial Stab. 25, 58–69.
Albuquerque, R., Schroth, E., 2015. The value of control and the costs of illiquidity. J. Finance 70, 1405–1455.
Allen, L., Gottesman, A.A., Peng, L., 2012. The impact of joint participation on liquidity in equity and syndicated bank loan markets. J. Financial Intermediation 21,
50–78.
Bajaj, M., Dennis, D.J., Ferris, S.P., Sarin, A., 2001. Firm value and marketability discounts. J. Corporate Law 27, 89–116.
Brav, A., Gompers, P., 2003. The role of lock-ups in initial public offerings. Rev. Financial Stud. 16, 1–29.
Dyl, E.A., Jiang, G.J., 2008. Valuing illiquid common stock. Financial Anal. J. 64, 40–47.
Feng, S.P., Hung, M.W., Wang, Y.H., 2016. The importance of stock liquidity on option pricing. Int. Rev. Econ. Finance 43, 457–467.
Finnerty, J.D., 2012a. An average-strike put option model of the marketability discount. J. Derivatives 19, 53–69.
Finnerty, J.D., 2012b. The impact of stock transfer restrictions on the private placements discount. Financial Manage. 42, 575–609.
Ding, B., Getmansky, M., Liang, B., & Wermers, R. (2009). Share restrictions and investor flows in the hedge fund industry, SSRN working paper.
Jacoby, G., Zheng, S.X., 2010. Ownership dispersion and market liquidity. Int. Rev. Financial Anal. 19, 81–88.
Kalay, A., Wei, L., Wohl, A., 2002. Continuous trading or call auctions: revealed preferences of investors at TASE. J. Finance 57, 523–542.
Kalay, A., Sade, O., Wohl, A., 2004. Measuring stock illiquidity: an investigation of the demand and supply schedules at the TASE. J. Financial Econ. 74, 461–486.
Li, W., Cao, T., Feng, Z., 2016. Heterogeneous venture capital, M&A activity, and market response. Emerg. Markets Rev. 29, 168–199.
Longstaff, F.A., 1995. How much can marketability affect security values? J. Finance 50, 1767–1774.

5
Unregistered shares are subject to selling limits under SEC Rule 144. The Rule sets that the number of restricted shares that can be sold during any three-month
window cannot exceed more than 1% of the firm's outstanding shares, and sets the average weekly trading volume of the stock (calculated over the 4 weeks prior to the
week in which the sale occurs).
6
Considering the market price of Elite in September 1998, the total value of the estate's shares was slightly more than $6 million.