Return To Publications
Download This Publication
Marketability Discounts and Risk in Transactions Prior to Initial Public Offerings
by Philip Saunders, Jr., Ph.D.
It is generally accepted that marketability has value and that stocks lacking marketability are worth less than marketable shares. Studies of private transactions of unlisted securities prior to initial public offerings (IPO's) provide evidence of discount for lack of marketability. Studies have shown average discounts of the pre-IPO price from the offering price of around 40% to 45%. The studies have also shown substantial dispersion of the discounts around their sample means.
This article provides evidence that part of the dispersion is related to perceived risk that the IPO may not take place, or may not take place at the offering price anticipated at the time of the pre-IPO private transaction. Other things being equal, the longer before the initial public offering that the private transaction occurs or the more inherently risky the enterprise, the greater the discount.
The most extensive publicly available record of transactions in unlisted securities prior to IPO's has been provided by Emory in this journal. The Emory data cover eight separate periods between January 1980 and April 1997, inclusive. Emory has previewed an additional study of pre-IPO transactions in stocks of Dot-Com companies. A sample of Dot-Com companies is different from the more catholic samples in the prior studies examined herein and, therefore, deserves a separate analysis. To gather the data in the original studies Emory and his associates searched prospectuses for transactions in the five months prior to IPO's. The Emory data are the basis for this study. Certain adjustments to the data were made as described below. Otherwise no attempt was made to validate or augment the data.
In some Emory studies, dates were reported only by month and year. In each such case we assumed that the transaction occurred on the 15th of the month. In the January 1985 to June 1986 study some dates were reported by month and year in the prospectuses, and Emory had assigned each date to the first of the month. For consistency we changed these to the 15th.
Some observations were removed from the sample because the data were internally inconsistent or, at the very least, improbable. In some cases the pre-IPO transaction date was reported as occurring either after the offering date or well before the five-month pre-offering window from which the observations were supposed to have been drawn. Cases removed from the samples for these reasons were Flextronics, Inc. from the August 1987 through January 1989 sample; The Buckle and Wind River Systems from the February 1992 through July 1993 sample; and Piercing Pagoda, Inc. and Serologicals Corp. from the January 1994 through June 1995 sample. Document Services was removed from the November 1995 through April 1997 sample because its offering value was reported to be more than double its market capitalization, as was Hot Topic, which had a reported book value of almost four times market cap. Crest Industries, Inc. was removed from the August 1990 through January 1992 sample because it had already been included in the February 1989 through July 1990 sample. Trimble Navigation was moved from the August 1990 through January 1992 sample to the February 1989 through July 1990 sample, to place it in the appropriate sample period. Bird Medical Technologies, Inc. was moved from the February 1989 through July 1990 sample to the August 1990 through January 1992 sample, for the same reason.
Raw price data were used to recalculate the percentage discounts (1 – Private transaction price/Offering price). In some instances there were small differences between the discounts so calculated and the discounts reported by Emory. The means and other summary data reported herein are based upon our calculations.
The Emory data, adjusted as described above, are summarized in Table 1. There are 302 transactions in all, 67 sales of stock and 235 grants of options. Included in the sales is one payment made in stock. The options data include, in addition to grants of options, two determinations of fair market value for taxes, one stock dividend, one issue of shares, one issue of performance units, and two exercises of options.
Several points stand out from the table. The mean and median discounts have been around 40% to 45%, except in the 1980-81 study. The average and median discounts on sales (51%) have been slightly larger than the discounts on grants of options (43%).
There is enormous dispersion of the discounts. They range in each study from highs of 80% to 95% to lows generally in single digits or even negative, i.e., the pre-IPO prices were higher than the IPO prices. The standard deviations are consistently around 20%.
The existence of the discounts and their substantial sizes indicates that stocks which lack marketability are less valuable than marketable stocks. However, the wide variation in the sizes of the discounts indicates that the mere fact of lack of marketability says little about the amount of discount for any particular unmarketable share.
Each pre-IPO transaction occurred during the five months prior to the IPO. As it takes several months to bring a stock offering to market, it is likely that the participants were aware that the stocks which were traded or on which options were granted had a good chance of becoming marketable in a matter of months. To quote Emory, "Since an initial public offering often takes four or five months from conception to completion, the transactions mentioned in the prospectuses in the study would almost certainly have reflected the likelihood of marketability and any other value adjustment associated with being a public company."
Logically the size of each discount would be related to the degree of perceived risk that (1) the IPO might not take place and that the security might not become marketable and (2) even if the IPO did occur, the offering price might be less than anticipated. The IPO might not take place or the expected price might change for any number of reasons, of which a partial list follows. Anticipated financial results might not materialize. The fundamental condition of the company might deteriorate. The stock market or the new issue market might take a turn for the worse. The industry might fall from favor with investors. The controlling stockholders might change their minds. The owners and the investment bankers might not agree on a price.
Time could serve as a general proxy for risk. The more time between the pre-IPO transaction and the IPO the more things that can go wrong and the more uncertainty as to both the IPO price and whether the IPO will occur at all.
Two other proxies for risk might be company size and the ratio of equity (book value) to market capitalization. With respect to company size, the theory would be that small companies may be more risky than larger ones and therefore could have more go wrong between the date of the transaction and potential IPO. The theory behind the latter measure would be that companies with low equity/market capitalization ratios would be valued more on the basis of future earnings than on value in hand. Future earnings are inherently more uncertain and, therefore, the enterprise more risky.
We tested the theory, that the size of the discount of the pre-IPO transaction price from the IPO price is related to risk that the anticipated IPO price might change or that the IPO might not occur at all, using the three proxies of risk described: time, company size, and equity/market capitalization ratio. The three proxies of risk were independent variables in least squares regressions upon the discounts of pre-IPO prices from IPO prices, using the Emory cross-sectional data. Time and only one measure of company size, book equity, were available for all of the Emory samples. Market capitalization was available in only the most recent sample.
The results are summarized in Table 2. The correlation of discount size with time (number of days elapsed between the pre-IPO transaction and the IPO) was statistically significant and positive in data from five of the eight Emory studies and for the aggregate of all of the studies. See the Chart 1 for discount and time for the combined studies. Since the three samples in which the relationship was not statistically significant were from the early periods, it is tempting to infer that the correlation is a recent phenomenon and that there has been some fundamental shift in the market's assessment of risk. However, those three samples are also the smallest; the lack of demonstrated relationship may be a function of small sample size.
Table 3 provides the data on the regressions on time and equity/market capitalization, where the relationship was found to be statistically significant at the five percent level or better. The estimated equations "explain" anywhere from 7% to 20% of the variations about the means. The coefficients on the time variable for the first four sample periods are around 0.002. The coefficient in the latest period (1995-97) is 0.0013; the drop is in part due to one outlying observation of 18% and 184 days. The coefficient for all periods, including the earlier ones where no correlation was found, is 0.0016.
The intercepts in the equations indicate the estimated values of the discount when time = 0, i.e., in the Emory data when the pre-IPO transaction and the IPO occur in the same month. The intercepts range from 0.182 to 0.305. See Table 3. The intercepts indicate minimum values for a lack of marketability discount in roughly the 20% to 30% range.
The estimated equations indicate that, other things being equal, the lack of marketability discount on pre-IPO transactions starts at 20% to 30% and increases 15 to 20 basis points for each day between the private transaction and the IPO. Using the equations from the six sample periods for which a statistically significant relationship between the discount and time was found, discounts for pre-IPO sales occurring five months (150 days) prior to the IPO have been calculated. See Table 4. The calculated 150-day discounts range from 0.493 to 0.569, or about 50% to 55%.
The tests found little or no relationship between company size and size of discount. See Table 2.
The Emory data allow testing for a relationship between size of discount and another risk proxy, the ratio of equity to market capitalization, in the 1995-97 sample. The relationship is statistically significant and appropriately negative, i.e., the lower the ratio, the less risk and the smaller the discount. See Table 3. When included with time as a variable the ratio roughly doubles the explanatory power of the equation.
With the five samples, including the combined sample, in which there were seven or more sales observations, we tested the relationships on stock sales alone. In none of the individual period samples was there a statistically significant relationship between the discount and time. See Table 2. For the combined sample for all periods the relationship was statistically significant at the five percent level but explained only four percent of the variance in the discount. In the 1995-97 sample, the equity/market capitalization ratio was negatively correlated with the discount on stock sales, as expected. The relationship was significant at the five percent level or better and accounted for 20% of the variance of the discount around the mean. The results for stock sales are similar to the results for stock sales plus grants of options, when the independent variable is equity/market capitalization. The results differ when the independent variable is time. Why there is a difference when time is the explanatory variable is not obvious.
Despite the statistical significance of the correlations described above, the explanatory power of the estimated equations is relatively low, ranging from three to 20 percent of the variation in the discounts about the sample means. There clearly is much stock-specific risk which is not captured in the proxy variables, time and equity/market capitalization.
In general, the tests conducted on the Emory data support the notion that the greater the risk that an IPO will not take place or will not take place at an anticipated price the lower the stock price in a pre-IPO transaction. Time is the proxy for risk most extensively tested and demonstrated herein. Other things being equal, the minimum value for a discount on a pre-IPO transaction is around 25%. The discount increases 15 to 20 basis points for every day that separates a pre-IPO transaction from the IPO. When the anticipated liquidity event is five months away, the discount, other things being equal, is estimated to be about 50% to 55%. Although the analysis says nothing about what the discount should be if a liquidity event is more than five months away, logically the discount would be greater.
The implication for the analyst valuing a non-marketable share five months or more from a liquidity event is that the starting point should not be the average discounts in the Emory studies, which are around 40% to 45% (see Table 1) but rather 50% to 55%. This is not to say that all non-marketable stocks, with an anticipated liquidity event somewhere between five months hence and never, should be discounted 50% to 55% or more. There is still substantial unexplained stock-specific variance. However, these results shift the burden to the analyst who wants to use a lower discount. Use of a lower discount requires a very good explanation of why a liquidity event is more certain to occur than for the average stock with an anticipated IPO or why the stock's attributes would make a willing buyer and a willing seller less concerned than usual with the lack of marketability.
The analysis provides a partial answer to Emory's oft-stated rhetorical question, "[I]f the kinds of discounts found in the above studies are appropriate for promising situations where marketability is probable, but not a certainty, how much greater should discounts be for the more typical company's stock that has no marketability, little if any chance of ever becoming marketable, and is in a neutral to unpromising situation?"
The author wishes to thank Anna Lau and Elizabeth D. Saunders, Research Assistants at Philip Saunders Associates, who assisted with analysis of the data.
First published in Business Valuation Review, Vol. 19, No. 4, December 2000.