THE USE OF FINANCIAL RATIOS AS MEASURES OF RISK IN THE DETERMINATION OF THE BIDASK SPREAD
Textonly Preview
Volume 9 Number 2 Summer 1996
THE USE OF FINANCIAL RATIOS AS MEASURES OF RISK
IN THE DETERMINATION OF THE BIDASK SPREAD
Huldah A. Ryan*
Abstract
The effect of financial reports on stock market behavior is a central issue of research in accounting and finance.
A number of studies investigate how financial information becomes impounded in security prices and affects
investment decisions. Prior studies on the determinants of the bidask spread investigate the effect of market risk
measures, and provide evidence that the bidask spread is a positive function of risk. Other studies report on an
association between market risk measures and accounting risk measures. The present study extends this line of
research by examining the effect of risk, proxied by accounting risk measures, on the bidask spread. The results of
ordinary least squares (OLS) regressions provide evidence of a statistically significant association between certain
accounting ratios and the bidask spread, and indicate that accounting risk measures account for more variability in
the bidask spread than market risk measures. Most notably, the results indicate that a model which includes both
accounting risk measures and market risk measures is a better fitted model that one which includes either
accounting risk measures or market risk measures alone.
INTRODUCTION
During the past two decades a large body of financial research has focused on the complexities of the stock market and
specifically on the process by which prices are determined. Previous theoretical research proposes that the dealer’s bidask
spread (which is the compensation to dealers for providing immediacy to market traders), is comprised of three
components: inventory order costs, inventory carrying costs, and adverse selection costs (Demsetz, 1968; Tinic and West,
1972; Stoll, 1978; Copeland and Galai, 1983). Stoll (1989) reports that 43 percent of the bidask spread reflects the
dealer’s adverse selection costs  the costs of trading with investors who possess “superior” information about the value of
security.
The finance literature shows that security dealers can diversify their unsystematic risk by maintaining a
portfolio of stocks (Sharpe, 1965; Fama, 1976), and empirical studies which relate risk to the bidask spread utilize
marketbased measures of systematic and unsystematic risk in the analyses (Benston and Hagerman, 1974; Barnea
and Logue, 1975). Other studies report on an association between accounting ratios and market risk measures, and
propose that certain accounting ratios can be used as proxies in predicting future security betas (Beaver et al. 1970;
Elgers and Murray, 1982).
This study investigates the effect of financial ratios as measures of risk on the bidask spread. It proposes that
financial statement analysis yields valuable information that can aid in investor decisionmaking, and uses as a
theoretical basis, a simple valuation model expressed as follows:
E( FCF )
V = (1+ r)
where:
*City University of New York
33
34
Journal Of Financial And Strategic Decisions
V
= market value of the firm
E(FCF) = expected future cash flows
r
= discount rate
Most event studies on the information content of accounting numbers base the analyses on earnings
announcements as a proxy for the numerator of the model, expected future cash flows. Trading occurs when there
are differences in expectations among investors relative to expected future cash flows and expected discount rates.
In this study, the focus is on the denominator of the model, and the proposal is that accounting ratios improve
market efficiency by providing additional information, on the value of the firm, that is not reflected in market risk
measures. These ratios include dividend payout, asset size, asset growth, liquidity, leverage, earnings variability,
and earnings covariability.
Fortin et al. (1989) report on the curious behavior of the spread around the end of the year which may be partly
due to the release of firmspecific financial information. Lev (1986) proposes that the more “equitable” and
“broadly informative” the firm’s financial information disclosure is, the lower is the information asymmetry
between informed and uninformed traders. Thus, the quality of information can determine the level of information
asymmetry in the market.
The present study investigates the determinants of the spread by analyzing a model using both accounting risk
measures and market risk measures to determine if it is superior to a model using either accounting or market risk
measures alone. The empirical results give new insights on how to evaluate risk in relation to the bidask spread,
and indicate that financial data do convey new information to market traders about a firm’s risk, which is reflected
in changes in the spread.
The remainder of this study is organized as follows: Section I describes the research methodology, the sample
design, and the various statistical tests used in the study. Section II reports on the key empirical findings, while the
summary and conclusions are presented in Section III.
RESEARCH METHODOLOGY
The proportional bidask spread, which is represented by:
( ask price ? bid price)
( ask price + bid price)×.5
is determined for each security over the threeyear sample period, and then used as the dependent variable in
several ordinary least squares (OLS) regressions to determine the factors affecting the bidask spread.
The Explanatory Variables
Previous theoretical and empirical studies report that the determinants of the bidask spread include the price of
the security, its trading volume, the number of dealers, and the number of shareholders trading in the stock (e.g.,
Demsetz, 1968; Benston and Hagerman, 1975; Stoll 1976, 1978; and Glosten and Milgrom, 1985). Tinic and West
(1972) show that spreads are negatively related to trading volume, while Glosten and Milgrom (1985) propose that
the average spread tends to decline for large volumes of trade. These variables are examined in this study, and
consistent with the results of previous studies, negative signs on their coefficients are expected in the results.
Roll (1984) presents a formulation of the bidask spread and empirically tests his model to determine the effect
of firm size on the spread. His findings reveal an inverse relation between size and the bidask spread. The present
study uses market value as a proxy for size, and examines its effect on the spread.
Studies on the determinants of the bidask spread use the Capital Asset Pricing Model as the basis for selecting
the risk variables analyzed (Bagehot, 1971; Ho and Stoll, 1983; Copeland and Galai, 1983). The studies which
report on the relationship between accounting ratios and market risk measures identify the accounting risk
variables examined in this study, as surrogates for the total variability of return on a firm’s securities (Beaver et al.,
The Use Of Financial Ratios As Measures Of Risk In The Determination Of The BidAsk Spread
35
1970). The second analysis extends the basic model to include these accounting risk measures, since previous
research show that they are related to market risk measures for which there is a theoretical base.
The overall expected results are for a positive relation between risk and the bidask spread as proposed by
Copeland and Galai (1983) and Glosten and Milgrom (1985). However, certain accounting ratios (dividend payout,
asset size, and asset growth), despite being risk measures, are negatively related to the bidask spread. In terms of
dividend payout (the ratio of the sum of cash dividends paid to common stockholders to the sum of income
available for common stockholders), previous empirical studies report a positive correlation between stock prices
and cash dividends (Aharony and Swary, 1980). Eades (1982) finds a clearly significant and negative relation
between dividends and risk, consistent with that reported by Beaver et al. (1970), and Rozeff (1982) reports that an
increase in dividend payout is associated with a decline in risk. Thus, as the dividend payouts increase, prices
increase because this can be interpreted as “good news” by investors, with the expectation for the firm to generate
higher future cash flows. As the firm’s risk is reduced, the bidask spread decreases. The empirical findings are
expected to be consistent with these predictions.
In terms of the asset variables (asset size  the natural log of total assets, and asset growth  the ratio of the
natural log of total assets in time period t, to the natural log of total assets in time period t1), prior research
findings show that larger firms are usually more diversified in terms of lines of business and less susceptible to
failure than smaller firms (Ohlson, 1980). Even though firms with larger asset sizes and higher asset growth rates
are riskier than firms with smaller asset sizes and lower growth rates, these variables provide signals to investors
and creditors about higher future cash flows. If investors value cash flows, they will trade more frequently in the
stocks of firms with increasing asset growth rates and asset sizes, and the bidask spreads will decline. Since the
number of shareholders is included in the model to control for the frequency of trading in a stock, the effect of asset
growth and asset size on bidask spreads can be determined.
The other accounting risk variables (leverage  the ratio of total senior securities to total assets, liquidity  the
ratio of current assets to current liabilities, earnings variability  the standard deviation of the earningsprice
ratio, and earnings covariability  the accounting beta computed by regressing the earningsprice ratio of each
firm over an eightyear period on a proxy for the market earningsprice ratio) are chosen because previous
research show them to be good surrogates for risk. It is conceivable that investors use these ratios in predicting the
future risk potential of a security, and positive signs on the coefficients for these variables are predicted in this
study.
The third analysis examines the effect of market risk on the bidask spread. The use of both price variability and
market beta is intended to represent the total risk of a security as proxied by market variables. Positive coefficients
for these variables are expected in the results, consistent with the theoretical and empirical results of past research.
Based on the foregoing description of the explanatory variables, the model to be analyzed is presented as
follows:
BA = ?0 + ?1PS + ?2NS + ?3ND + ?4VOL + ?5MV + ?6DP + ?7G
?8AS + ?9DE + ?10L + ?11EV + ?12AC + ?13MR + ?14PV + e
where:
BA
= proportional bidask spread
AS
= asset size
PS
= closing price per share
L
= liquidity
NS
= number of shareholders
EV
= earnings variability
ND = number of dealers
AC
= accounting beta
VOL = trading volume
MR
= market beta
MV = market value
PV
= price variability
DP
= dividend payout
?0
= intercept term
G
= asset growth
?1, ?2, ..., ?14 = regression coefficients
DE
= leverage
e
= error term, assumed to be serially independent,
normally distributed, and independent of the
regressors
36
Journal Of Financial And Strategic Decisions
The Hypotheses
The first analysis tests the hypothesis which predicts a negative association between price, number of
shareholders, number of dealers, trading volume, and market value on the bidask spread.
H1: ?1, ?2, ?3, ?4, ?5 < 0
The second analysis examines the association between the spread and the accounting risk measures, and the
hypotheses are stated as follows:
H2.1: ?6, ?7, ?8 < 0
H2.2: ?9, ?10, ?11, ?12 > 0
The relationship between the market risk measures and the bidask spread is investigated in the third analysis
with the following hypothesis being tested:
H3: ?13, ?14 > 0
The model predicts that the coefficients on price, number of shareholders, number of dealers, trading volume,
market value, and certain accounting variables (dividend payout, asset size, and asset growth) will be negative,
while the coefficients on the other accounting risk variables (leverage, liquidity, earnings variability, and earnings
covariability), and the market risk variables, (beta and price variability) will be positive.
The Data
Data for a threeyear period, January 1, 1982 through December 31, 1984, were collected on a random sample
of 60 OvertheCounter (OTC) firms. Other data requirements for selection include the following:
(1) Each firm had financial statement data available in Moody’s Manuals for the period 1982
through 1984;
(2) Daily ask and bid prices, as well as volume data for each firm were available on Compuserve
Tapes for the period to be studied;
(3) Data on the number of shareholders and the number of dealers were available for each firm in
Moody’s Manuals and the National Stock Summary, respectively.
Summary statistics on the bidask spread were computed for each security. Several OLS regressions were then
estimated using the bidask spread as the dependent variable, and price per share, trading volume, number of
dealers, number of shareholders, market value, accounting risk measures, and market risk measures as predictor
variables. The variables in the model were measured either at the end of the year (number of shareholders, number
of dealers, market value of the firm, accounting risk measures, and market risk measures), or over the entire year (
average of daily bidask spreads, prices, and volume).
CROSSSECTIONAL REGRESSION RESULTS
The first analysis examines the effect of price, trading volume, number of shareholder, number of dealers, and
market value on the bidask spread. The correlation analysis indicates a high correlation between price and market
value (0.935), and the regression results do not show any significant difference in the effect using price or market
value. Accordingly, the results in Table 1 excludes the effect of market value. The regression coefficients for the
variables are all of the predicted signs with price and volume having statistically, significant effects on the
proportional spread. The Fratio indicates that the variables as a whole, significantly determine the size of the bid
ask spread.
The Use Of Financial Ratios As Measures Of Risk In The Determination Of The BidAsk Spread
37
TABLE 1
Regression Results For Hypothesis 1 Testing The Effect Of Price,
No. Of Shareholders, No. Of Dealers, And Trading Volume On The BidAsk Spread
Variable
Expected Sign
Coefficient
TValue
Alpha Level
Intercept
0.1622
6.686***
.0001
Price

0.0004
1.563
.1038
No. Of Shareholders

<0.0001
0.012
.9907
No. Of Dealers

0.0033
0.339
.7357
Volume

0.0147
4.937***
.0001
N
= 180
Adjusted RSquare = 0.3346
FStatistic
= 8.291***
***Significant at the 0.01 level with onetailed test.
The results of the regression including accounting risk variables in the basic model above are presented in
Table 2. The coefficients for dividend payout, asset size, and asset growth are negative as predicted, and in the case
of asset size and asset growth, the results indicate a statistically, significant effect on the bidask spread. In terms of
liquidity, leverage, and earnings variability, the signs are all positive, indicating that the bidask spread widens
with increases in risk, as proxied by these variables, consistent with the theory of risk that has been developed. The
negative sign on the coefficient for accounting beta may be due to the fact that several variables are proxying for
risk in the model.
TABLE 2
Regression Results For Hypothesis 2 Testing The Effect
Of Accounting Risk Measures On The BidAsk Spread
Independent Variable
Expected Sign
Coefficient
TValue
Alpha Level
Intercept
0.3398
5.162***
.0001
Price

0.0003
1.224
.2270
No. Of Stockholders

<0.0001
0.072
.9431
No. Of Dealers

0.0009
0.120
.9051
Volume

0.0084
2.857***
.0064
Dividend Payout

0.0196
0.765
.4479
Asset Growth

0.0681
2.294**
.0263
Asset Size

0.0293
3.361***
.0016
Leverage
+
0.0243
1.163
.2508
Liquidity
+
0.0026
0.966
.3389
Earnings Variability
+
<0.0001
1.089
.2818
Accounting Beta
+
0.0010
0.423
.6743
N
= 180
Adjusted RSquare = 0.6046
FStatistic
= 9.064
*Significant at the 0.10 level with onetailed test
**Significant at the 0.05 level with onetailed test
***Significant at the 0.01 level with onetailed test
38
Journal Of Financial And Strategic Decisions
The Fstatistic for the regression is statistically significant at the one percent level. The effect of the accounting
risk measures on the spread is evidenced by the increase in the adjusted Rsquare (from 0.3346 to 0.6046) with the
addition of these ratios to the first model. A partial Fratio test to determine the incremental explanatory power of
adding these variables to the basic model indicates a significant partial Fstatistic at the one percent level. Thus,
the addition of accounting risk variables enhances the basic model, and provides justification for including
financial ratios as measures of risk in explaining variability in the bidask spread.
One objective of this study is to determine whether accounting risk variables explain more variation in the bid
ask spread than market risk measures. The results of the regression analyzing the effect of market risk measures
are presented in Table 3, and they provide further evidence on the effect of risk on the bidask spread. A partial F
ratio test indicates that the increase in the proportion of variance accounted for by market risk variables is
statistically significant at the one percent level. Accordingly, as the market risk increases, the bidask spread
widens, indicating that dealers charge more for assuming higher risk.
TABLE 3
Regression Results For Hypothesis 3 Testing The Effect
Of Market Risk Measures On The BidAsk Spread
Variable
Expected Sign
Coefficient
TValue
Alpha Level
Intercept
0.1688
7.700***
.0001
Price

0.0028
2.123**
.0386
No. Of Stockholders

<0.0001
0.275
.7842
No. Of Dealers

0.0054
0.602
.5495
Volume

0.0137
5.033***
.0001
Market Beta
+
0.1131
1.752*
.0857
Price Variability
+
0.0043
1.478
.1453
N
= 180
Adjusted RSquare = 0.4612
FStatistic
= 9.273
*Significant at the 0.10 level with onetailed test
**Significant at the 0.05 level with onetailed test
***Significant at the 0.01 level with onetailed test
The adjusted Rsquare in Table 3 using market risk variables (0.5169) is lower than the results shown in Table
2 using accounting ratios as risk measures (0.6046). This suggests that accounting risk variables account for larger
variations in the bidask spread than market risk variables.
The regression results presented in Table 4 (adjusted Rsquare = 0.6327) indicate that the full model
outperforms all of the other models estimated in this study. These results confirm the hypotheses of a positive
relation between accounting risk measures, proxied by liquidity, leverage, and earnings variability, and the bidask
spread. The findings on market beta also indicate a positive, but marginally significant association (at the 10
percent level) with the spread. However, the signs on the coefficients for the number of dealers and market value
are positive instead of negative as predicted. Note must be taken of the inclusion of price, price variability, and
market value in the regression model, and also, of the finding of a high correlation between the price variables and
market value.
SUMMARY AND CONCLUSION
This study investigates the effects of several variables on the bidask spread by use of separate crosssectional
regression models. The empirical findings of the models support prior research that there is an inverse relation
between the bidask spread and price, volume, the number of stockholders, and the number of dealers, and a
positive association between the spread and risk as proxied by several accounting and market risk variables
(liquidity, leverage, earnings variability, market beta, and price variability). The results also indicate a statistically
significant association between certain accounting ratios (asset size and asset growth) and the bidask spread.
The Use Of Financial Ratios As Measures Of Risk In The Determination Of The BidAsk Spread
39
TABLE 4
Regression Results On The Effect Of A Reduced
Accounting Risk Variable Set On The BidAsk Spread
Variable
Expected Sign
Coefficient
Tvalue
Alpha Level
Intercept
0.1723
8.247***
.0001
Price

0.0020
2.960***
.0047
No. Of Stockholders

<0.0001
0.779
.4397
No. Of Dealers

<0.0001
0.009
.9932
Volume

0.0096
3.196***
.0001
Dividend Payout

0.0555
2.327**
.0240
Asset Growth

0.0852
2.798***
.0073
Earnings Variability
+
<0.0001
0.445
.6582
N
= 180
Adjusted RSquare = 0.5035
FStatistic
= 9.404
***Significant at the 0.01 level with onetailed test
**Significant at the 0.05 level with onetailed test
Prior studies which examine the effect of risk on the bidask spread utilize market measures as proxies for risk
in the analyses. The results of this study indicate that accounting risk measures, as proxies for risk, explain a
significantly higher proportion of the variance in the bidask spread than market risk measures. Further, the
explanatory power of a combined model with both accounting risk measures and market risk measures is higher
than that of a model using either accounting or market risk measures alone, and higher than any of the models
tested in previous studies.
This study attempts to build on prior research on the usefulness of accounting information. The findings
indicate that the use of financial ratios as risk measures enhances the predictive power of a model explaining
variability in the bidask spread, and illustrate that a model with both accounting and market risk measures is
better fitted than one using either accounting or market risk measures alone. The evidence presented in this study
suggests that financial statement data provide information that reduces information asymmetry in the market, and
indicates that investors should fully utilize this information set in assessing the potential riskiness of a security, and
accordingly, in their investment decisionmaking.
The theoretical literature on the bidask spread focuses on market risk, but not on accounting risk. The findings
of this study highlight a need for the development of a theory that incorporates accounting ratios as they affect the
underlying risk characteristics of a security. Another possible avenue for future research is an examination of the
spread around the report release date. A comparison of the results obtained from analyzing the effect of these
variables around a window period surrounding a December 31 report date, with the results of their effect around a
window period surrounding a nonDecember 31 report date would be a natural extension of this study.
40
Journal Of Financial And Strategic Decisions
TABLE 5
Regression Results On The Determinants Of
The BidAsk Spread The Full Model
Variable
Expected Sign
Coefficient
TValue
Alpha Level
Intercept
0.3379
5.142***
.0001
Price

0.0031
2.161**
.0362
No. Of Stockholders

<0.0001
0.168
.8673
No. Of Dealers

0.0001
0.017
.9863
Volume

0.0093
3.210***
.0025
Market Value

<0.0001
1.618
.1128
Dividend Payout

0.0056
0.210
.8347
Asset Growth

0.0636
2.206**
.0327
Asset Size

0.0274
3.134***
.0031
Leverage
+
0.0163
0.797
.4295
Liquidity
+
0.0036
1.334
.1889
Earnings Variability
+
<0.0001
0.672
.5053
Accounting Beta
+
0.0001
0.041
.9672
Market Beta
+
0.2103
1.358
.1815
Price Variability
+
0.0015
0.356
.7235
N
= 180
Adjusted RSquare = 0.6327
FStatistic
= 8.138
*Significant at the 0.10 level with onetailed test
**Significant at the 0.05 level with onetailed test
***Significant at the 0.01 level with onetailed test
REFERENCES
[1] Aharony, J. and I. Swary, “Quarterly Dividends and Earnings Announcements and Stockholders’ Returns: An
Empirical Analysis,” Journal of Finance, March 1980, pp. 112.
[2] Amihud, Y. and H. Mendelson, “Dealership Market: Market Making With Inventory,” Journal of Financial
Economics, March 1980, pp. 3153.
[3] Bagehot, W., “The Only Game in Town” Financial Analyst Journal, March/April 1971, pp. 1214.
[4] Barnea A. and Logue, D., “The Effect of Risk on the Market Maker’s Spread,” Financial Analyst Journal,
November 1975, pp. 4549.
[5] Beaver, W., P. Kettler, and M. Scholes, “The Association Between Market Determined and Accounting
Determined Risk Measures,” The Accounting Review, October 1970, pp. 654682.
[6] Benston, G. and R. Hagerman, “Determinants of BidAsked Spreads in the OTC Market,” Journal of
Financial Economics, December 1974, pp. 353364.
[7] Copeland, T. and D. Galai, “Information Effects of the BidAsk Spread,” Journal of Finance, December
1983, pp. 14571469.
[8] Demsetz, H., “The Cost of Transacting,” Quarterly Journal of Economics, February 1968, pp. 3353.
The Use Of Financial Ratios As Measures Of Risk In The Determination Of The BidAsk Spread
41
[9] Eades, K., “Empirical Evidence on Dividends as a Signal of Firm Value,” Journal of Financial and
Quantitative Analysis, November 1982, pp. 471500.
[10] Elgers P. and D. Murray, “The Impact of the Choice of Market Index on the Empirical Evaluation of
Accounting Risk Measures,” The Accounting Review, April 1982, pp. 358375.
[11] Fama, E., “The Behavior of Stock Market Prices,” Journal of Business, January 1965, pp. 34105.
[12] Fortin, R., R. Grube, and O. Joy, “Seasonality in NASDAQ Dealer Spreads,” Journal of Financial and
Quantitative Analysis, September 1989, pp. 395407.
[13] Foster, G., “IntraIndustry Information Transfers Associated with Earnings Releases,” Journal of Accounting
and Economics, March 1981, pp. 201232.
[14] Garman, M., “Market Microstructure,” Quarterly Journal of Financial Economics, June 1976, pp. 257276.
[15] Glosten, L. and P. Milgrom, “Bid, Ask and Transaction Prices in a Specialist Market with Heterogeneously
Informed Traders,” Journal of Financial Economics, March 1985, pp. 71100.
[16] Ho, T. and H. Stoll, “On Dealer Markets Under Competition,” Journal of Finance, May 1980, pp. 259267.
[17] Lev, B., “Toward a Theory of Equitable and Efficient Accounting Policy,” The Accounting Review, January
1988, pp. 122.
[18] Morse, D. and N. Ushman “The Effect of Information Announcements on the Market Microstructure,” The
Accounting Review, April 1983, pp. 247258.
[19] Ohlson, J., “Financial Ratios and the Probablistic Prediction of Bankruptcy,” Journal of Accounting
Research, Spring 1980, pp. 109131.
[20] Ou, J. and S. Penman, “Financial Statement Analysis and the Prediction of Stock Returns,” Journal of
Accounting and Economics, November, 1989, pp. 295329.
[21] Roll, R., “A Simple Implicit Measure of the Effective BidAsk Spread in an Efficient Market,” The Journal
of Finance, March 1984, pp. 11271139.
[22] Rozeff, M., “Growth, Beta and Agency Costs as Determinants of Dividend Payout Ratios,” Journal of
Financial Research, Fall 1982, pp. 249259.
[23] Stoll, H., “The Supply of Dealer Services in Securities Markets,” Journal of Finance, September 1978, pp.
11331151.
[24] Tinic, S. “Inferring the Components of the BidAsk Spread: Theory and Empirical Tests,” Journal of
Finance, March 1989, pp. 115134.
[25] Tinic, S. and R. West, “Competition and the Pricing of Dealer Service in the OvertheCounter Market,”
Journal of Financial and Quantitative Analysis, June 1972, pp. 17071728.
[26] Woodridge, J., “Dividend Changes and Security Prices,” Journal of Finance, December 1983.