convertible call policiesdownload.xuebalib.com/5crd3lyoxtkr.pdf · received november 1985, final...
TRANSCRIPT
Journal of Financial Economics 19 (1987) 91-108. North-Holland
CONVERTIBLE CALL POLICIES* An Empirical Analysis of an Information-Signaling Hypothesis
Aharon R. OFER
Northwestern Uniuersi<v. Ecansron, IL 60208, USA Tel Aoic University, Tel Am. Isruel
Ashok NATARAJAN
?lorthwestern Unioersgv. Evanston. IL 60208. USA
Received November 1985, final version received December 1986
This paper tests an information-signaling hypothesis as a potential explanation for corporate convertible bond call policies and for the negative share price reaction to the announcement of the calls. We test this hypothesis by trying to ascertain whether the information signaled is realized. Our results show an unexpected decline in the firm’s performance.subsequent to the call. We also find significant negative cumulative returns during a sixty-month period following the calls.
1. Introduction
A number of recent studies have examined corporate policies with regard to calls of convertible bonds. Yet the motivation for the calls of convertible securities and the reaction of the firm’s shareholders to the call announcements are still not well understood. Ingersoll (1977a) and Brennan and Schwartz (1977) show that to minimize the value of the liability, convertible bonds
should be called as soon as the firm can force conversion, which is when the conversion value equals the call price. Ingersoll (1977b), however, reports that
of the 179 companies in his sample all but nine wait too long, according to his model, to call the bonds. Furthermore, this discrepancy is not explained by transaction costs or corporate taxes. Additional perplexing evidence is pro- vided by Mikkelson (1981), who finds statistically significant negative excess returns associated with the call announcements.
Brennan and Schwartz (1982), Dunn and Eades (1984), and Constantinides and Gundy (1985) discuss several reasons why management might rationally delay calling convertible bonds. These explanations are based on arguments such as: (i) management compensation schemes are a function of earnings per
*We have benefited from the comments and suggestions of Ravi Jagannathan. Robert Korajczyk, David Larcker, Artur Raviv. Daniel Siegel, Eduardo Schwartz (the referee). and Clifford W. Smith (the editor). Research assistance was provided by Stephen V. Thomas. Financial support from the Banking Research Center and the Accounting Research Center at Northwestern University is gratefully acknowledged. The usual disclaimer applies.
0304-405X/87/$3.50 c 1987. Elsevier Science Publishers B.V. (North-Holland)
92 A. R. Ojer and A. Naturu~an, Concernble call pohcies
share: (ii) the loss of bondholders’ goodwill has an effect on the pricing of future bond issues; (iii) there is a preference for voluntary conversion induced through dividend increases; and (iv) some owners of convertible securities follow less than optimal voluntary conversion strategies. Mikkelson suggests that the negative excess returns associated with a call announcement could be related to the lost interest tax shield.
Another explanation for the observed negative common-share return and the delayed calls is that the announcement of the call signals ‘bad news’. A signaling model of convertible call policy is presented by Harris and Raviv (1985). In their model, investors perceive the decision to call as a signal that conveys unfavorable information. As a result of the call, investors revise their expectations about the firm’s future prospects, and these revisions cause a decline in the share price. Smith (1986) also discusses the relationship between information asymmetry and observed negative stock price changes associated with the announcement of convertible bond calls.
Other than the common shares’ negative return (a phenomenon that can be attributed to other factors), however, there is no documented evidence of bad news associated with calls of convertible debt. Thus an unresolved issue is whether the common shares’ negative returns reflect the direct effects of the conversion (such as the lost interest tax shield) or an information effect. A related issue is whether in delaying the call, management acts in its own interest (trying to maximize its compensation by avoiding dilution of earnings per share) or whether the call is delayed because of favorable information. The results of previous studies do not allow one to differentiate among the competing hypotheses regarding management motivation for the timing of the call and the share price reaction to the announcement of the call. In this study we provide evidence that can distinguish between the signaling hypothesis and other plausible explanations.
If investors are rational and calls signal bad news, the information conveyed should be realized, and we should be able to observe this realization as an unexpected decline in the firm’s performance following the call.’ Thus if calls of convertible bonds are motivated by insiders’ information, two hypotheses follow:
(1) The announcement of the call is associated with a decline in common share price since investors perceive the call as signaling bad news.
(2) The bad news is manifested in the performance of the firm after the call.
Results supporting the first hypothesis are provided by Mikkelson and are replicated in this study. The second hypothesis has not previously been examined, however. It should be emphasized that an unexpected change in performance is consistent with the signaling hypothesis, but inconsistent with
‘SC~ Schwert (1981) for a discussion of the problems in using financial data to identify changes in economic performance.
A. R. Ofer and A. Natarajan. Concemble call policies 93
the other proposed explanations. In this study, we find that bad news is manifested in that firms calling their convertible bonds experience an unex- pected decline in performance in the years following the call. We also examine the returns to common shares after the call and find significant negative cumulative abnormal returns. These results are consistent with the unexpected decline in economic performance and with the hypothesis that calls of convert- ible bonds signal bad news. They also imply market inefficiency, however. since they indicate that the signaled information is not fully incorporated into
share prices at the announcement of the call. In section 2 we describe our sample, outline our methods, and present the
results of our analysis of firm performance subsequent to the call. Our analysis of the equity returns is presented in section 3. A summary concludes the paper.
2. Economic performance subsequent to call
2. I. Sample
A sample of 232 calls in the ten-year period between January 1971 and December 1980 is identified using Standard and Poor’s Bond Guide.’ From this sample we eliminate fifty-four calls for firms whose common stock is not
listed on the New York or American Stock Exchanges, as well as twenty-two calls for companies for which no data are available on the COMPUSTAT or Center for Research in Security Prices (CRSP) tapes. Further examination reveals that in fifteen instances the calls were made in connection with an impending or a completed merger or acquisition. Since these calls are not considered voluntary, they are excluded from the analysis. The result is a sample of 141 voluntary calls, listed in the appendix.
For the analysis of firm performance after the call, two calls are omitted
from the sample because they were made by two companies that had called other convertible debt earlier in the same fiscal year. Eleven other calls, for
which data are not available on the COMPUSTAT tapes (although return data are available on the CRSP tapes), are also omitted. This results in a sample of 128 calls.
The 128 calls are distributed over a ten-year period, with some concentra- tion toward the latter half of the decade.3 No industry concentration is evident among the companies making calls (seventy-seven 4-digit SIC industries are represented). Summary characteristics of the sample are presented in table 1. The amounts of convertible debt called are not trivial, either as a proportion
‘Calls of more than one series of convertible debentures on the same day were treated as a single call.
‘The distribution of calls by calendar year is as follows: 1971 - 15. 1972 - 17. 1973 - 1% 1974 - 5, 1975 - 3. 1976 - 13, 1977 - 12. 1978 - 17, 1979 - 11. and 1980 - 26. There is no concentration of calls among calendar months.
J.F E.- D
94 A. R. Ofer and A. Natarqan, Conrerrzble call pohcies
Table 1
Summary characteristics of 128 calls of convertible debt betaeen 1971 and 1980 (millions of dollars).
Mean .Median Range
Cull amount charactenstrcs Call amount
Cull antount 0s a percentage o/: Common equity at beginning of call year Convertible debt at beginning of call year Long-term debt at beginning of call year
Company characrerisrics Sales in call announcement year Total assets in call announcement year
$31.8 $13.1 $0.3-$250.0
18.5 13.4 0.2-81.0 79.1 98.4 6.2-100.0 30.2 16.6 0.2- 100.0
$1057.2 $334.4 $7.2-$X1,209.0 $1119.9 $408.7 $33.2-$11.631.0
of the debt claims on the corporations calling the debt or as a proportion of the common equity.
2.2. Methods and results
To test the information-signaling hypothesis we examine whether there is an unexpected change in the firm’s performance after the call. Performance is defined as a relative change in profitability.4 One problem in measuring unexpected changes in performance following the call is the separation of cause and effect. For example, if profits are measured as earnings per share, an observed change in performance following the call could have been either the motivation for the call or the result of the conversion increasing the number of outstanding shares (and thus not represent an unexpected change in perfor- mance). To insure the correct identification of cause and effect, we use several measures of profitability, classified in relation to the impact of the conversion: 5
earnings before interest and taxes (EBIT), earnings before taxes (EBT), and earnings per share (EPS).
The first measure, EBIT, is not affected by the conversion: the second, EBT, is affected by the reduction in interest payment; and the third, EPS, is affected by both the reduction in interest payment and the increase in the
4Smith (1986) discusses the implied changes in net operating cash Rows associated with changes in financial policy.
‘These profitability measures are not independent.
A. R. Ofer and A. .ValaraJan. Concemble call pohnes 95
Table 7
Growth rakes in earnings before interest and titxeh ( EBIT). ramings before tames ( EBT). and earnings ptx share ( EPS). for a portfolio of tirms uith convertible bonds that were called between
1971 and 1980.
Year’ EBfT .v lJ EBT :vh EPS ,vh
-5 18.122 110 19.529 110 23.5lC; 105 -4 15.86% 112 18.15% 112 24.51% 108 -3 18.89% 114 22.09% 114 10.354 110
1 -; 17.238 17.81% 11s 116 18.797 18.14ac 115 116 24.03% 21.969
110 112
Cl 12.30% 107 10.12% 107 - 1.14% 103 +2 13.82% 103 9.4jq 103 -O.jjC, Y9 -3 4.927 104 3.2jaC 104 - 8.685 100 +4 4.50% 92 1.33% 92 - 8.43nC Y6 +5 12.48% 77 12.07% 17 7.537 72
‘Year in relation to call year. h,V = number of calls.
number of shares outstanding. To explore cause and effect in the measurement of unexpected changes in performance further, we create a fourth variable that measures the profitability of the firm in the years following the call as if the conversion had not occurred, by subtractin, 0 the interest on the convertible
bonds from earnings before taxes in the years following the call. For the period following the call the following variable is calculated:
AEBT,=EBT,-R. t=l 5 t..., ,
where R is the interest payment on the convertible debt.6 This variable, AEBT, represents the adjusted earnings before tax during the period after the
call, assuming the conversion did not occur. We first examine the relative change in the profit variables for a portfolio
that includes the remaining companies in our sample. The relative change in each profit variable for the portfolio is calculated as follows:
p,.,= k E,.,i- t E,.,,t-1 i
(2) r=l 1=1
where E,.;., is the profit of company i in year t, measured in terms of the profit variable j, and n is the number of firms in the sample.
The results in table 2 indicate that there is a change in performance subsequent to the call. For example, during the fi,ve years before the call,
‘The interest payment is calculated as equal to the coupon rate times the face value of the bonds that were converted.
96 A. R. Ofer and A. :VamraJan, Conrerrible call policres
growth in earnings per share for the portfolio ranges from 24.5% to 10.3%; in the five-year period following the call it ranges from 7.5% to -8.6%. More important, this change in performance is noticeable even when performance is measured in terms of EBIT, a variable not affected by the conversion. The growth rate in EBIT before the call ranges from 18.9% to 15.8%; after the call it ranges from 13.8% to 4.5%.
To determine whether these changes are unexpected and significant, we have
to identify changes in performance for individual firms. The performance of
the firm in terms of the profit variable E,,,., is measured as the relative change
of profit:
P r.,.r= (E,,,,,-E,.,.,-*)/IE,.,,l-~l~
where 1.1 defines the absolute value of the variable. Dividing by the absolute value of the profits introduces noise into the
analysis, especially if earnings in a particular year are relatively small or close to zero. To ensure that our results are not ilriven by outliers, we eliminate observations where the relative change in profit is extreme in comparison with its average over the sample period. The criterion used is’
IP,.,.k/E.,I’ 5, (4)
where p, , is the average change in profit variable j for firm i.
To identify unexpected changes the expected performance has to be defined.
Three models are postulated to describe the stochastic behavior through time of the performance indices and to identify their expected values. Model 1 assumes that the firm’s performance is stationary through time. Model 2 assumes that the expected performance of the firm in any given year is a function of the average performance of all the firms in the economy. Model 3 defines the expected performance of the firm in any given year as a function of the average performance of all other firms in the industry during the same
year. To identify unexpected changes in performance, these models are estimated
with the inclusion of a shift variable, defined as
D,. r = 0 for t -C call year,
(5) = 1 for t > call year.
‘The number of calls that were eliminated as a result of this criterion ranges from six when performance is measured as the relative change in earnings before interest and taxes to twenty-one for the earnings per share variable.
A, R. Ofer and A. ,VaararaJan. Convertible call polrcws 97
The following three regressions corresponding to the three models are estimated:
P !.I.’ =QI ,./.l+ljr.,.I~,.r+‘,.,.r.l, (6)
P 1.J.l =(Y ,.,.1+Y,.,.2PM,.,+P,.,~,D,.,+e,~,.,.2, (7)
P !.,.I =CX l,J.3+~~.,.3PK~,J.~+~~.,.3D,.~+E~,,,~.3’ (8)
where PM,., is the average performance of all the firms on the COMPUSTAT tapes measured in terms of performance measure j, in year t, and PI,. ,, I is the average performance of all other firms in the same industry as firm i, measured in terms of performance measure j, in year t. The regressions are estimated for all the companies in the sample with at least eight observations; the maximum number of observations is fifteen.
Under the null hypothesis that the calls are not motivated by new unfavor- able information, calls should not be followed by an unexpected change in economic performance and the coefficients /3,, J. k should equal zero. Thus the following hypothesis can be tested:
H,: P,.,.k=O’ for all i, for all j, and for all k.
H,: b,.J.kfo’
To test this hypothesis we use the maximum likelihood ratio test. Define
x ,.J.k= [SSR(u),,,,k/SSR(r),.J.~]~.‘.X’2, (9)
where SSR( u)~. J. k is the sum of squared residuals in the unrestricted regres- sion for company i, performance index j, and model k; SSR(r),. J,k is similarly the sum of squared residuals in the restricted regression for company i, performance index j, and model k; and T,,., is the number of observations used in the regression.
Then asymptotically,
- 2 ln X,.,,k - ~~(1)~ (10)
and the test statistic is
,$t - (21nX,,,,,) -XZ(nj>, (11)
under the null hypothesis, where n, is .the number of companies for which enough data points are available to estimate /I;. j,k.
The likelihood ratios in table 3 indicate that firms that call their convertible bonds have an unexpected change in performance subsequent to the call. All the likelihood ratios are significant at the 1% level, although it should be noted that the performance measures are not independent.
98 A R. Ofer und A. ~Vucoru~an. concerrihie cd pohaes
The direction of the unexpected changes in performance can be identified by analyzing the average coefficient of the shift variable D, for each performance index. If the calls are motivated by new unfavorable information, the coet% cient ,B,. ,. k should be negative. To test this we first standardize each p,, ,, k by its standard error and define
SP‘.,.k = PL,.k/SW!.,J, and
The following hypothesis can be tested:
(12)
(13)
H,: SP/,L=O, H,: J7u,. k < 0.
Under the Central Limit Theorem s$,.~ is distributed normal and the follow- ing test statistic can be used:
z = ~,.k/(“,)1’2. (14)
In table 3 we report the average standardized coefficients of the shift variable, s$,. k, and their ;-values. For all of the performance indices. ST,,., is negative and significant at the 1% level, indicating a significant decline in performance subsequent to the call of the convertible bonds. This decline is observed whether the benchmark is the performance of these firms in the
period preceding the call, the average performance in the economy, or the average performance in the industry.
The results also show an unexpected decline in performance for variables that are not affected by the conversion, such as earnings before interest and taxes and the adjusted earnings before taxes variable. These results suggest that the decline in performance after the call is not caused by the conversion of the bonds; rather. the expectation of such a decline is the reason for the call.
The significant unexpected decline in the firms’ performance after the calls supports the hypothesis that the calls act as signals of bad news. The consistency of the results across three models that use different data sets to estimate the unexpected change in performance and the fact that the unex- pected decline in- performance is identified for performance indices that are not affected by the conversion validate the robustness of the results.
3. Rates of return on common shares
The common share returns of the companies in our sample are analyzed during two periods. First, we analyze the daily returns during the days surrounding the announcement of the call. This is done to verify that the calls
A R. Ofer and A. Yararajan. Concertlble call policies 99
Table 3
Tests for changes in performance subsequent to calls of convertible bonds between 1971 and 1980: likelihood ratios. Xa, and the average standardized coefficient of the shift variable. 3,. k.
Model la
1. Earnings before interest and taxes 2. Earning before taxes 3. Earnings per share 4. Adjusted earnings before taxes
X Zb
143.82’ 118.41’ 126.76’ 117.95’
q. kc
-0.441 - 0.286 - 0.243 - 0.390
; d Ne
- 4.34’ 97 - 2.67’ 87 - 2.29’ 88 - 3.57’ 83
XZb
Model 2’
$. kc :d N’
1. Earnings before interest and taxes 168.47’ - 0.395 -3.84’ 97 2. Earnings before taxes 135.54f - 0.297 - 2.60’ 87 3. Earnings per share 151.75’ - 0.317 - 2.99’ 88 4. Adjusted earnings before taxes 133.21’ - 0.381 - 3.48’ 83
Model 3”
XZb Ne
1. Earnings before interest and taxes 144.23’ - 0.392 - 3.84’ 96 2. Earnings before taxes 131.40’ - 0.257 - 2.38’ 86 3. Earnings per share 148.32f - 0.306 - 2.87’ 87 4. Adjusted earnings before taxes 130.89’ - 0.343 - 3.12’ 83
‘Model 1. Model 2, and Model 3 refer to the following three regressions. respectively:
(I) P,.,.,=*,.,,1 +P*.,.rDi.*+EI.,.r.I,
(2) P~.,.r=~,,/.2+~,.,.2~~,.,+B,.,.2D,.,+~,,,,,.z.
(3) ~.,.,=~L,.3+YL,.3PI,.,.l+8,.,.34,+~L,.~.3.
where P, ,,I is the performance of company i measured in terms of performance index j in year 1. P,M,,, and Pl,, are the corresponding performance measures for the economy and for the other firms in the it&&. respectively. D,,, IS a shift variable that takes the value of zero for the years prior to the call and one for the years subsequent to the call.
hThe likelihood ratios are used to test whether the coefficient of the shift variable fi,,,,k is different from zero. Under the null hypothesis, these ratios have a chi-square distribution.
‘P,. ,, !. is standardized by its standard error and these values are then summed and averaged across all firms to produce $,, i(, which. under the Central Limit theorem. is distributed normal.
‘The z-statistic.
‘V = number of calls.
‘Significant at the 1% level
are indeed unexpected and that the market perceives them as signals of bad news. Second, we analyze the common stock monthly returns during the sixty months following the call. We do this to determine whether the unfavorable information is captured immediately in share prices or whether more unex- pected bad news is revealed in subsequent periods.
The Wall Street. Journal Index is examined for the formal call announce- ment date and for any prior announcements. The trading day before the
100 A.R. Ofer and A. NataraJan, Concwtlble call pokes
earliest announcement date is specified as the call announcement date or the event date and is identified as day 0 in our analysis. In thirty cases, no date is ascertainable from either the Wall Street Journal Index or Standard and Poor’s Bond Record. and these calls are dropped from this part of the analysis. The final sample for the evaluation of the market reaction to call announcements consists of 111 calls. Elimination of seventeen companies with missing data during the estimation period before the call announcement, as well as two calls made by companies that had called other convertible debt earlier in the same fiscal year, reduces the sample used in the analysis of monthly returns to ninety-two companies.
Daily and monthly stock return data for the companies are obtained from the CRSP tapes. The CRSP daily and monthly equal-weighted indices are used as a proxy for the market return. The statistical procedure used to identify excess returns is described below for both daily and monthly returns.
The single-factor market model is taken as the characterization of the rate of return on a stock:
(15)
where t is the time unit over which the return is calculated, either a month or a day; R,, is the continuously compounded rate of return on stock i for period t; R,, is the continuously compounded rate of return on the market for period f; and E’ is the random error for stock i for period t.
The parameters (Y, and p, are estimated for each stock over a period before the call. For the daily returns analysis, the estimation period begins at day - 80 and ends with day - 21. For the monthly returns analysis, the estimation period begins at month -61 and ends with month -2. The estimated market model parameters are then used to calculate the residuals over the estimation period and the excess returns over the event period. To test for significance the procedure described in Pate11 (1976) is used. The excess returns for each firm are standardized by two factors: (i) the estimated standard deviation of the residuals during the estimation period and (ii) a factor that adjusts for predicting returns outside the estimation period. Assuming that the individual standardized excess returns are cross-sectionally independent, the average standardized excess return, ASR,, has a unit-normal distribution. To test the null hypothesis of a zero mean excess return, the statistic it is used:
zl,= ASR, f (7’- 2)/(7--4) “’ ,
r-l 1 I 06)
where T is the number of observations in the estimation period. The squared standardized excess returns can be used to test for a change in
the variance of the excess returns given that the mean of the excess returns is zero. The statistic zZ, which is used to test for a shift in variance, has a
A R. Ofer und A. NararaJun, Concerrrbie call polrcm 101
unit-normal distribution and is calculated as
=2r =(ASSR,-1) f 2(7--3)/(T-6) (17) r=l
where ASSR, is the average squared standardized excess return for period t. The cumulative average excess returns are defined as
CAR,= i AR,, (1%) r=j
where AR, is the average excess return in period t and / and / are the first and last month in the cumulation period, respectively. To test for significance, the cumulative standardized excess returns for firm i are normalized by dividing by the number of months in the cumulation period. The normalized cumulative standardized excess returns are summed across firms. Assuming that the individual standardized excess returns are independent, the statistic z3 is used to test the null hypothesis of zero cumulative excess returns:
zJc= CSR, 5 (T- 2)/(T- 4) ‘[ l/2
, 1=1 I
(19)
where CSR, is the sum of the normalized cumulative standardized excess returns across nc firms with cumulative returns available in period c“.
Table 4 presents the results of the analysis of share-price reaction to call announcements. The table reports the average excess returns, AR, average standardized excess returns, ASR, the average squared standardized returns, ASSR, and their corresponding z-statistics. In addition, to provide a distri- bution-free test for the variables of interest, ASR and ASSR, a jackknife procedure is employed. The jackknife procedure splits the total sample into arbitrary equal-sized subsets by deleting one observation at a time and computes n estimates of the sample mean. These sample means are then used to calculate a jackknifed mean, standard error, and t-statistic.
The cumulative price drop on the announcement day and the following day is 0.09%. The average cumulative excess return over the forty-one-day test period is - 0.6%. whereas it is - 1.7% over the five-day period from day - 1 to day +3. An examination of the squared residual statistics reveals only two days with significant t-statistics at a 1% level for the two-tailed test: these are day 0 (the call announcement date) and day +l (the day following the call announcement). This indicates that there is a market reaction to the call announcements. The nature of the market reaction can be determined through
107 A. R. Ofer and A. .Yafaru~an. Concernble call policres
Table 4
A\eragc excess returns (.A R 1. average standardized excess returns ( .-ISR ). and aterage squared standardized excess returns ( ASSR) for forty-one days surrounding the announcements of 111
calls of convertible bonds between 1971 and 1980.
Day” ARh ASRh I’ Id AS.SRh :’ P
- 20 - 19 - 18 - 17 - 16 - 15 - 14 -13 - 12 - 11 - 10
-9 -X -7 -6 -5 -4 -3 -2 -1
0 1 2 3 4
: 7 8 9
10 11 12 13 14 15 16 17 18 19 20
0.002 0.152 1.589 1.350 1.238 0.004 0.176 1.840 1.702 1.211 0.001 0.068 0.711 0.657 1.188 0.001 0.056 0.585 0.542 1.180 0.002 0.095 0.996 0.943 1.136
- 0.003 -0.12Y - 1.342 - 1.537 0.7Y2 0.00’ 0.024 0.251 0.2’6 1.244 0.000 0.034 0.353 0.374 0.905 0.005 0.230 2.403 2.019 1.487
- 0.001 - 0.063 - 0.656 - 0.681 0.942 - 0.001 - 0.107 - 1.107 - 1.023 1.238
0.002 0.138 1.446 1.489 0.964 - 0.002 -0.110 - 1.152 - 1.153 1.021 - 0.003 - 0.222 - 2.312 - 2.380 1.004
0.001 0.071 0.738 0.624 1.422 0.003 0.139 1.450 1.640 0.810 0.004 0.167 1.746 1.586 1.254 0.000 0.011 0.118 0.120 0.964 0.001 0.080 0.834 0.506 2.754
- 0.002 -0.018 -0.190 -0.112 2.901 - 0.002 -0.114 - 1.188 - 0.841 2.057 - 0.009 - 0.499 - 5.200 - 3.465 2.527 - 0.004 -0.183 - 1.904 - 1.822 1.137 - 0.001 -0.100 - 1.041 - 0.943 1.243 - 0.002 -0.148 - 1.545 - 1.590 0.977 - 0.002 -0.113 - 1.179 - 1.043 1.306
0.000 - 0.049 - 0.514 - 0.574 0.813 0.001 0.095 0.990 0.876 1.301 0.002 0.035 0.362 0.3Y5 0.854
- 0.001 - 0.032 - 0.334 - 0.321 1.096 0.000 0.005 0.052 0.045 1.362
- 0.002 - 0.090 - 0.930 - 0.930 1.039 0.000 0.036 0.378 0.437 0.773
- 0.003 -0.121 - 1.270 - 1.229 1.096 0.001 0.027 0.286 0.245 1.382
- 0.001 - 0.005 - 0.049 - 0.048 1.088 - 0.003 -0.135 - 1.406 - 1.322 1.161
0.002 0.085 0.883 0.816 1.190 - 0.001 0.003 0.032 0.030 1.112
0.004 0.205 2.137 1.888 1.338 - 0.001 - 0.048 - 0.502 - 0.539 0.879
1.559 0.643 1.388 0.856 1.201 1.050 1.141 0.920 0.828 0.498
- 1.641 - 1.549 1.605 0.541
- 0.833 - 0.699 3.350 1.252
- 0.567 - 0.431 1.495 1.308
- 0.410 - 0.236 0.004 0.134
-0.124 0.027 2.885 1.136
- 1.517 - 1.270 1.676 1.028
- 0.409 - 0.205 12.452 1.009 13.509 1.190 7.465 2.953
10.820 3.377 0.836 0.837 1.59Y 0.997
-0.317 -0.111 2.048 1.213
- 1.492 - 1.430 2.010 1.200
- 1.195 - 1.151 0.542 0.465 2.454 1.450 0.131 0.189
- 1.771 - 1.607 0.541 0.358 2.591 1.183 0.481 0.538 1.009 0.766 1.216 0.888 0.651 0.623 2.276 1.472
- 1.020 - 0.890
“Trading day prior to the earliest announcement is specified as day 0. hExcess returns are based on market model parameters estimated over days - 80 through - 21. ‘The :-statistic. ‘The r-statistic is generated with a jackknife method.
A. R. O/H and A. .Vuturapn. Conwrtrble call pokes 103
-.o.o 4. /
+:::::.~::-+::‘::C_CC::::::x3
1 5 9 13 17 21 25 29 33 37 41 45 49 53 57
Yonth in relation to Call Announosmrnt Date
Fig. 1. Average excess returns following the announcement of calls of convertible bonds between 1971 and 1980. Excess returns are based on market model parameters estimated over months - 61
to -2.
the average standardized residuals ( ASR). On day + 1 and day + 2, the ASRs are negative and significant at the 1% an d 10% levels, respectively. In terms of magnitude, the ASR on day +I is more than twice the size of the same statistic on any of the other forty days. The significant negative market reaction to the call announcement supports the hypothesis that the calls signal unfavorable information about the firms’ prospects. The results in table 4 are consistent with those reported by Mikkelson (1981)’
The analysis of the returns to common shares during the sixty months following the call indicates significant negative abnormal returns. During most of the period analyzed, the average excess returns, AR, and the average standardized excess returns, ASR,‘are negative. Whereas eleven months out of the sixty have a negative and significant (at a 5% level or better) ASR, none of the ASRs are positive and significant. The average monthly residuals are presented in fig. 1. Table 5 reports the cumulative average excess returns, CARS, and the z-statistics for the normalized cumulative standardized excess returns, CSRs, for different holding periods. The CARS for all the holding
“Mikkelson’s event date is the date on which the announcement appears in the Wall Srreet Journul and is equivalent to our day + 1.
104 A. R. Ofer and A. NararaJan. Conoerrrble cull pohcies
Table 5
Cumulative averrtge excess returns (C.4 Rs) and :-statistics during different holding periods. following the announcement of calls of comeruble bonds between 1971 and 1980 (in percentages).
Months”
l-12 l-24 l-36 l-48 l-60
CARb
- 11.15 - 24.16 - 39.44 - 57.65 - 72.61
.L @
3.96’ 87 6.21’ 81 7.39e 79 8.54’ 76 7.21e 56
‘Month 1 is the first moc;:l following the call announcement. ‘Excess returns are based on market model parameters estimated over months -61 through
_ 2. The percentage of negative residuals in a given month ranges from 437 to 70%. with a mean value of 53.9%.
‘The z-statistic is equal to the sum of the normalized cumulative standardized excess returns across all firms divided by (I:_, (7 - 2)/( T - 4))“‘. where ir is the number of obsemations in the estimation period and n IS the number of tirms.
dN = the number of firms in the sample with returns available during cumulation period. “Significant at the 1% level.
periods- are negative. More important, the z-statistics for the cumulative standardized excess returns are negative and significant at the 1% level.
To ensure that these results are not caused by outliers, we examine the distribution of the normalized cumulative standardized excess returns for individual firms. Out of the ninety-two calls, sixty-six have a negative CSR.
Twenty-two firms have a CSR less than - 1.96, and only three firms have a CSR greater than 1.96. The distribution of the CSRs indicates that these results are not driven by a few outliers. 9 Furthermore, replicating this analysis, using the mean-adjusted and market-adjusted models to estimate the excess returns, yields similar results.”
The significant negative monthly excess returns during the sixty-month period indicate that more unexpected and unfavorable information is revealed to the market after the calls. These results present a puzzle. The significant negative CARS are consistent with the unexpected decline in accounting performance these firms experience during the same period, and they are also consistent with the hypothesis that calls of convertible bonds signal bad news. These results also imply, however, that the information signaled by the calls is not fully incorporated into the share price at the time of the call announce- ment.” To this extent, these results suggest a market inefficiency. An altema- tive explanation for the negative CARS is that there is a shift in the mean
‘The results cannot be attributed to a shift in beta. The average beta in the estimation period was 1.037 and the average beta in the test period was 1.090.
“See Brown and Warner (1985) for a discussion of these models.
‘t In this respect, it is interesting to note that results in table 2 indicate that the largest drop in profit growth rates occurred in the third and fourth years following the call.
A. R. Ofer and A. NararaJan. Concertlble call policies 10s
return and that the benchmarks we use in the analysis are inappropriate. resulting in an over-estimate of the normal rate of return for these securities.
4. Summary
Previous studies of convertible bond calls showed that firms tend to delay such calls. and that call announcements are associated with significant nega- tive common stock returns. Several explanations have been offered for this phenomenon. The results in previous studies are not sufficient. however. to
differentiate among different hypotheses regarding the motivation for the calls and the share price reaction to their announcement.
This paper presents evidence that supports the information-signaling hy- pothesis. This hypothesis postulates that calls of convertible bonds signal new and unfavorable information to investors in the market. For a signal to be effective, the information it conveys must materialize at some point. If calls of convertible debt are motivated by, and convey. new unfavorable information about the firm’s prospects, we should observe an unexpected decline in the firm’s performance in the years after the call.
Performance is defined as the relative change in a profit variable, and four profit variables are used in the analysis. To ensure that the results are independent of model specification, three models are used to identify the unexpected change in performance. All three models identify statistically unexpected declines in performance for all four profit variables after the calls
of convertible debt. Although the various tests are not independent. the consistency of the results across three models and the fact that the unexpected decline in performance is identified also for performance indices that are not affected by the conversion attests to the robustness of our results.
Taken together. the significant decline in performance after the call of
convertible bonds. the significant negative common stock returns associated with the announcement of the call, and the significant cumulative abnormal returns following the call support the hypothesis that calls are motivated by new unfavorable information. The significant negative cumulative residuals subsequent to the calls suggest market inefficiency, They imply that the information signaled by the calls is not fully incorporated into share prices at the time of the announcements and that more unexpected and unfavorable information is revealed to the market after the calls. Alternatively the negative cumulative returns can be the outcome of overestimating the normal rate of return for these securities during the period after the call.
In recent years the notion of asymmetric information and signaling equi- librium has become important in economic and financial theory. Few studies have examined the empirical content of this concept. Our study provides a direct evaluation of a signaling hypothesis, since an unexpected decline in performance after the call is inconsistent with other proposed explanations and is consistent with the information-signaling hypothesis. The results of our analysis support this hypothesis for calls of convertible bonds.
Ap
pe
nd
ix
Nam
e of
com
pany
m
akin
g ca
ll
Tab
le
6
Lis
t of
sam
ple
com
pani
es
calli
ng
conv
ertib
le
debt
be
twee
n 19
71
and
lYX
0.
Ann
ounc
emen
t da
te
Nam
e of
com
pany
m
akin
g ca
ll
Aer
o-Fl
ow
Dyn
amic
s A
kzon
a In
c.
Ala
ska
Air
lines
A
mer
ada
Hes
s A
mer
ican
A
ir
Filte
r A
mer
ican
H
ospi
tal
Supp
ly
AM
R
Cor
p A
pco
Oil
APL
C
orp
AV
CO
C
orp
Ran
iste
r C
ontin
enta
l B
anis
ter
Con
tinen
tal
Bau
sch
and
Lom
b B
axte
r L
abs.
B
caun
i t (
El
Paso
C
o)
Bee
ch
Air
craf
t B
ig T
hree
In
dust
ries
B
urlin
gton
N
orth
ern
Bur
roug
hs
Cor
p Il
uttc
s G
as
and
Oil
Col
lins
and
Aik
man
C
oppe
rwel
d C
orp.
C
ore
Lab
orat
orie
s C
ryst
ui
Oil
Dev
elop
men
t C
orp.
of
Am
eric
a D
igita
l E
quip
men
t C
orp
Eas
tern
A
irlin
es
Eas
tern
A
irlin
es
Ech
lin
Man
ufac
turi
ng
Eck
erd
(Jac
k)
Co
Ehr
enre
ich
Phot
o-O
ptic
E
l Pa
so
Nat
ural
G
as
El
Paso
N
atur
al
Gas
197X
O
S/12
/71
OX
/l s/
79
04/2
6/71
06
/W/7
1 08
/07/
X0
12/2
1/71
05
/1Y
j77
1976
0X
/02/
79
02/0
2/72
02
/26/
71
1974
06
/l 3/
75
04/0
5/77
12
/21/
7Y
&I/
24/7
2 19
x0
01/1
x/73
12
/l l/8
0 19
7X
1976
19
76
06/2
0/77
11
/20/
x0
01/1
7/x0
IY
71
OY
/OX
/7X
1 l
/01/
73
OX
/l s/
7’)
01/l
X/7
2 19
78
0X/2
6/77
Mac
y (R
.H.)
In
c.
Mac
v (R
.H.)
In
c.
Map
;; In
c.
Mar
indi
que
Min
ing
Mar
tin
Mar
ietta
M
asco
C
orp
McC
ullo
ch
Oil
McC
ullo
ch
Oil
McD
erm
ott
(J.
Ray
) M
esa
Petr
oleu
m
MG
IC
Inve
stm
ent
MG
M
Inc
Mic
rodo
t M
isso
uri
Paci
fic
Moh
awk
Dat
a Sc
ienc
es
MSL
In
dust
ries
N
atio
nal
Can
N
atio
nal
Hea
lth
Ent
erpr
ises
N
atio
nal
Med
ical
E
nter
pris
es
Nat
iona
l St
arch
an
d C
hem
ical
s N
CR
C
orp
New
hall
Lan
d an
d Fa
rmin
g N
orth
rop
Oak
In
dust
ries
O
ccid
enta
l Pe
trol
eum
Pa
n A
mer
ican
W
orld
A
irw
ays
Pan
Am
eric
an
Wor
ld
Air
way
s Pa
tric
k Pe
trol
eum
Pa
hall
Inc
Peng
o ln
dust
ries
Pe
nney
(J
.C)
Phili
p M
orri
s Pi
llsbu
ry
Co
Ann
ounc
emen
t da
te
I973
06
/l 3/
7X
I Y72
05
/01/
74
z
07/2
X/7
X
‘?
06/l
5/72
Q
02
/10/
72
2
12/1
1/7Y
%
12
/lY/7
3 ra
i
05/1
4/71
$
12/2
2/72
02
/06/
76
:
1Y71
%
12/0
1/7x
P
0x/3
0/7x
07
/03/
74
? 2 0x
/17/
7’)
G
0X/0
6/X
0 s
03/2
0/7Y
2
03/0
3/77
=
z %
0x
/23/
7x
%
01/1
7/x0
S’
I Y77
5
02/l
3/x0
05
/10/
77
0x/3
0/7x
O
Y/l2
/7X
07
/l 7/
X0
06/l
X/7
6 12
/30/
X0
1971
1Y
74
07/l
l/75
Elg
in
Nat
ion
al
Ind
ust
ries
E
mp
ire
Gas
E
nst
ar
Co
rp
Ess
ex
Inte
rnat
ion
al
Est
erlin
e C
orp
E
van
s P
rod
uct
s F
irst
U
nio
n
Rea
l E
stat
e F
isch
er
and
P
ort
er
Fis
cher
S
cien
tifi
c F
lori
da
Gas
C
o.
Fo
rest
C
ity
En
terp
rise
s F
R
Lia
uid
atin
r F
uq
ua’
Ind
us&
G
elco
ca
rp
Gen
eral
In
stru
men
t (;
corg
ia
Pac
ific
(;
rorg
ia
Pac
ific
G
rain
ger
(W
.W)
Gro
w
Gro
up
(;
rtm
mla
n
Car
p
(;u
lf
&
Wcs
tcrn
(i
ulf
&
W
cbtc
rn
Ind
ust
rica
(i
ulf
L
ife
llold
ing
s G
ulf
R
eso
urc
e an
d
Ch
emic
als
Har
rah
s I I
elm
eric
h
and
P
ayn
e H
cub
lin
Inc
flo
crn
er-W
ald
orl
H
ug
hes
T
oo
l ln
silc
o
Co
rp
ltel
C
arp
I c
cl c
arp
K
M
arl
Kau
fman
B
road
In
c K
err-
McG
ee
Co
rp
Les
lie
F;l
y In
c.
LT
V
Co
rp
1976
02
/22/
7Y
11/3
0/7Y
19
76
06/2
4/X
0 09
/21/
77
06/2
2/7Y
19
72
19x0
05
/l 3/
71
Oh
/l 3/
7Y
06/2
3;X
O
05/0
1/72
‘1
97X
05/0
4/79
01
/27/
76
I l/2
3/76
07
/18/
72
11/2
4/X
()
ol/3
o/x
o
01/2
2/76
IY
XI
Ol/j
17/7
3 06
/20/
75
0X/1
6/78
12
/04/
75
03/1
4/72
lY
77
11/2
0/X
()
09/2
5/X
0 02
/16/
77
06/2
0/77
01
/l l/7
3 04
/24/
72
0X/2
2/12
1Y
71
0X/0
6/76
Piz
za
Hu
t I’n
eum
o
Car
p
Pri
me
Co
mp
ute
r R
epu
blic
N
atio
nal
B
ank
(N.Y
) R
oh
r In
du
stri
es
Ro
wan
C
os.
R
yder
S
yste
m
Inc.
S
abin
e C
orp
S
anta
F
e In
tern
atio
nal
S
anta
F
e In
tern
atio
nal
S
axan
In
du
stri
es
Sco
tty’
h
Inc.
S
eisc
om
D
elta
Sh
ears
on
L
och
R
ho
crd
cs
So
uth
east
l)
anco
rp
Sp
erry
C
orp
S
tan
dar
d
Oil
(ln
dia
mi)
S
tau
ffer
C
hem
ical
S
un
Ele
ctri
c
Sim
dst
ran
d
Car
p
Su
mh
ine
Min
ing
T
and
y C
orn
. T
and
; C
O&
J.
Ten
nec
o
Co
rp.
Tid
ewat
er
Tyl
er
Co
rp
Tyl
er
Co
rp
UA
L
US
AIR
W
ain
oco
O
il W
alla
ce-M
urr
ay
Wah
nar
t S
tore
5 W
arn
er
Bro
ther
+7
Art
s W
eld
ed
Tu
be
Co
rp
of
Am
eric
a W
este
rn
Un
ion
C
arp
W
illi
amr
Bra
s W
illia
ms
Co
s.
Zap
ata
Car
p
07/2
2/76
05
/ 1 Y
/77
02/0
4/X
0 O
X/O
Y/7
X
1913
I2
/ I x
/x0
IY71
07
/03/
X(1
03
/10/
72
11/2
6/X
0 1Y
72
12/1
4/72
O
X/2
Y/X
O
19x0
10/0
4/73
01
/30/
x0
10/3
1/74
10
/22/
15
OY
/I2/7
7
10/2
0/x0
19
x0
1Y71
19
x1
0X/2
4/7’
) 10
/31/
77
05/2
7/76
12
/20/
71
12j1
4;7x
08
/25/
7X
06;2
1;78
06
/l 5/
7X
II/IX
/77
OY
/21/
72
01/2
6/76
03
/2Y
/72
02/O
Y/7
I
I l/0
5/73
01
/2x/
x0
108 A. R. Ofer and A. ,Vumru~an. Conwrtible call pohes
References
Brennan. Michael J. and Eduardo S. Schwartz, 1977. Convertible bonds: Valuation and optimal strategies for call and conversion. Journal of Finance 32, 1699%17:j.
Brennan. Michael J. and Eduardo S. Schwartz. 1982, The case for convertibles. Chase Financial Quarterly. 27-46.
Brown. Stephen J. and Jerold B. Warner. 1985. Using daily stock returns: The case of event studies. Journal of Financial Economics 14. 3-31.
Constantinedes. George M. and Bruce D. Grundy, 1985. Call and conversion of convertible corporate bonds: Theory and evidence. Working paper (University of Chicago. Chicago. IL).
Dunn. Kenneth B. and Kenneth M. Eades. 1984, Voluntary conversion of convertible preferred stock and the optimal call strategy. Unpublished manuscript (Carnegie-Mellon University. Pittsburgh. PA).
Harris, Milton and Artur Raviv. 1985. A sequential signaling model of convertible debt policy, Journal of Finance 40. 1263-1281.
Ingersoll. Jonathan E. Jr.. 1977~1. A contingent claims valuation of convertible securities, Journal of Financial Economics 4. 289-322.
Ingersoll. Jonathan E. Jr., 1977b. An examination of corporate call policies on convertible securities, Journal of Finance 32. 463-478.
Mikkelson, Wayne H.. 1981, Convertible calls and security returns. Journal of Financial Econom- ics 9. 237-264.
Mikkelson, Wayne H., 1983, Capital structure changes and decreases in stockholders’ wealth: A cross-sectional study of convertible security calls. Working paper no. 1137 (National Bureau of Economic Research, Cambridge, MA).
Patell, James M., 1976. Corporate forecasts of earnings per share and stock price behavior: Empirical tests. Journal of Accounting Research 14, 246-276.
Schwert. G. William, 1981. Using hnancial data to measure the efi’ects of regulation. Journal of Law and Economics 24. 121-158.
Smith, Clifford W. Jr., 1986. Investment banking and the capital acquisition process. Journal of Financial Economics 15. 3-29.
本文献由“学霸图书馆-文献云下载”收集自网络,仅供学习交流使用。
学霸图书馆(www.xuebalib.com)是一个“整合众多图书馆数据库资源,
提供一站式文献检索和下载服务”的24 小时在线不限IP
图书馆。
图书馆致力于便利、促进学习与科研,提供最强文献下载服务。
图书馆导航:
图书馆首页 文献云下载 图书馆入口 外文数据库大全 疑难文献辅助工具