WHAT'S AN OSCAR WORTH?

NELSON, RANDY A. ; DONIHUE, MICHAEL R. ; WALDMAN, DONALD M. 等

CALBRAITH WHEATON [*]

This article examines the impact of an Academy Award nomination and award for best picture, best actor/actress, and best supporting actor/actress on a film's (i) market share of theaters, (ii) average revenue per screen, and (iii) its probability of survival. The model is estimated using weekly box-office data for a matched sample of nominated and non-nominated films. The results indicate substantial financial benefits for a nomination and award for best picture and best actor/actress. The structure of rewards is consistent with that found in two-stage, single-elimination tournaments. (JEL L1, L8)

"When Oscar Talks, the [Box Office] Listens"

-- Variety, April 14, 1994, p. 7

I. INTRODUCTION

Every year in mid- to late February members of the Academy of Motion Picture Arts and Sciences vote to nominate five films for an Academy Award (an Oscar) in each of 23 major categories. [1] Approximately six weeks later the winners, determined by a second round of ballots by members of the Academy, are announced during the nationally televised awards ceremony. An Oscar nomination or award means different things to different people. For an actress, actor, or director, an Oscar nomination signifies the industry's recognition of significant professional achievement and may result in higher salaries and greater choice of roles or films in the future. [2] For moviegoers, an Oscar nomination serves as a signaling device, indicating which films are viewed by industry experts as being worthy of recognition, whereas for film distributors an Oscar nomination or award often means additional box office revenues.

In this article we suggest that the annual competition for an Oscar may be viewed as a tournament, in which films compete successively for a nomination and an award in a given category, and for more prestigious nominations and awards across categories. The competition clearly satisfies the first requirement for a tournament, as there is no absolute standard for an Oscar nomination or award; nominees (winners) simply must be among the five (single) best films in a given year. Our goal is to determine whether the films that receive an Oscar nomination or award possess what Rosen (1981, 846) describes as "the elusive quality of 'box office appeal,' the ability to attract an audience and generate a large volume of transactions," that characterize a superstar and thus guarantee the convex reward structure found in tournaments.

In estimating the value of an Oscar nomination or award, we employ a panel, data set and an event study framework. Our data set consists of weekly observations on box-office revenues during the period 1978 through 1987 for a sample of non-nominated films and for every film that received an Oscar nomination in one or more of five categories: best picture, best actor/actress, or best supporting actor/actress. The framework we employ enables us to examine gross box-office receipts before and after the announcement of the nominations and awards, using both the non-nominated films and data on films prior to the nominations and awards as a control. This approach allows us to identify the impact of a nomination or award on a film's gross box-office receipts and the probability that a film will remain in distribution following the nominations.

II. DATA

To determine the impact of an Oscar nomination or award on box-office receipts, data were obtained from Variety, a weekly trade journal for the entertainment industry. [3] Throughout the period 1978-87 Variety published data for each of the top 50 grossing films in a given week. [4] Although the number of screens included in the sample has changed over time, the sample consistently covers approximately 11% of all theaters in the United States, which produce approximately 26% to 27% of the yearly motion picture box-office revenues according to U.S. Department of Commerce Reports and 24% of the Motion Picture Association of America annual totals. [5]

During the period of our sample, 1978-87, a total of 131 films were nominated for best picture, best actor/actress, or best supporting actor/actress awards. [6] Of this total, 62 films received one nomination, 39 received two nominations, 14 received three nominations, 12 received four nominations, and the remaining 4 films received five nominations. Thirty-eight films received Oscars in these five categories, with 28 films receiving a single award, 8 films winning two Oscars, and 2 films receiving three awards. Of the 131 nominated films, 10% were released in the first quarter of the year, 13.9% in the second quarter, 16.9% in the third quarter, and 59.2% in the fourth quarter. The distribution of release dates for the nominated films is consistent with the belief that a late release date enhances both the probability of receiving a nomination and the financial rewards associated with the nomination.

In addition to the 131 nominated films, data for a control sample of 131 non-nominated films were also collected. For each nominated film, a non-nominated film released in the same week was chosen; the highest ranked non-nominated film that remained on the top 50 chart for a minimum of five weeks was selected. In some cases it was necessary to move forward or back a maximum of two weeks from the release date of the nominated film to find a non-nominated film that met these criteria. The decision to select the highest ranked non-nominated film may impart a downward bias to our estimates of the value of an Oscar, as our control sample is taken exclusively from the upper tail of the universe of non-nominated films. Our results may thus represent a lower bound estimate of the value of an Oscar.

For each film the sample begins in the first week in which the film appears on the top 50 chart, and ends when the film disappears from the chart for the last time. Since the majority of films were released in the fourth quarter, it was often necessary to track a film over two calendar years; no film remained on the charts for more than two calendar years. The combined sample of nominated and non-nominated films totals 4,544 observations. For each film we compute the percentage of total screens on which the film appears (SHARE) and the average revenue per screen (ARPS); the average values of SHARE and ARPS for the first 25 weeks following a film's release are presented in Figures 1 and 2, respectively, for both the nominated and non-nominated films. [7] All monetary values are converted to 1997 dollars using the CPI for entertainment expenditures. In the first week the nominated films average $55,726 per screen, whereas the non-nominated films average only $21,312. The ARPS for both types of films declines a lmost continually following the first week, reaching an asymptote of approximately $7,000-$8,000 (in the ninth week for the non-nominated films and in the twentieth week for the nominated films). Presumably, when a film's ARPS reaches the $7,000-$8,000 level, theaters stop showing the film and select another film with a higher expected ARPS.

As shown in Figure 2, the non-nominated films appear on an average of 8.52% of all screens in the first week, decreasing to 3.67% of screens in the fifth week, and only 0.69% of all screens in the twentyfifth week. The nominated films open on an average of 2.24% of all screens in the first week, increasing to a market share of 3.76% in weeks 6 through 8, before beginning a gradual decline to an average of 1.81% of all screens in week 25. The gradual increase in SHARE for the nominated films is consistent with the informational dynamics discussed by De Vany and Walls (1996), in which consumers discover which films they prefer over time and transmit the information to other consumers, creating information cascades. This same process may explain the rapid decline in SHARE for the non-nominated films if initial consumers provide negative rather than positive feedback.

DeVany and Walls (1997) stress the fact that most films have short and unpredictable lives in which they constantly compete with existing and newly released films for screens and audiences. Successful films are able to extend their stay in the theaters, thus generating additional revenues, whereas unsuccessful films often remain in the theaters for a relatively brief period. It has often been argued that films receiving Oscar nominations remain in the theaters longer than non-nominated films, or in the words of an industry executive, "For the studios, it [a nomination] meant you could extend the life of a film in release. . . . It became very clear, over the years, that an Oscar translated into money." [8] The difference between the longevity of the two groups is illustrated by the survival curves presented in Figure 3, which represent the percentage of films remaining on the top 50 chart in any given week following their release. The average non-nominated film remained on the Top 50 chart for 11.1 weeks, an d the average nominated film remained on the chart for 24.0 weeks, a difference of 117%. Among the nominated films, approximately 75% remained on the Top 50 chart for 16 weeks or longer, approximately 50% for 22 weeks or longer, and approximately 25% for 29 weeks or more; the corresponding figures for the films receiving no nominations are 7 weeks, 9 weeks, and 13 weeks, respectively. The survival curve for the non-nominated films is very similar to the survival curve for the 350 randomly selected films studied by De Vany and Walls (1997).

An Oscar nomination also increases the probability that a film will be in the theaters in the weeks following the nominations. Only 22.3% of the non-nominated films were still in the theaters at the time of the nominations or were rereleased following the nominations; only 6.2% of the non-nominated films were in the theaters in the weeks following the announcement of the winners. For the films receiving nominations, 61.5% were in the theaters at the time of the nominations, an additional 14.6% were rereleased following the nominations, and 56.2% were still in distribution in the weeks following the announcement of the winners. The probability that a nominated film would still be in the theaters at the time of the nominations is a function of the film's release date; 81.8%, 63.6%, 16.7%, and 7.7% of the films released in the fourth, third, second, and first quarters, respectively, were still in distribution when the nominations were announced.

III. MODEL SPECIFICATION AND ESTIMATION

The cumulative box-office revenues taken in by any film are a function of three distinct factors: (i) the number of screens on which the film appears each week, (ii) the average revenue per screen, and (iii) the number of weeks that the film remains in distribution. As demonstrated by the brief review of the data in section II, an Oscar nomination or award can affect each of these factors. To estimate the value of an Oscar, it is thus necessary to determine what impact, if any, an Oscar nomination has on SHARE, ARPS, and the probability of a film surviving a given number of weeks in the theaters.

We begin our analysis by focusing on the ARPS and SHARE equations. It is assumed that both ARPS and SHARE are a function of the number of weeks the film has been in distribution, seasonal fluctuations in movie attendance, film-specific factors, and whether or not the film has received an Oscar nomination or award. Both ARPS and SHARE will vary across films because of differences in the quality of the film's producer and director, screenwriters, actors and actresses, and so on. Given the panel nature of the data, the impact of these film-specific factors will remain constant over time, allowing us to control for them by including a separate dummy variable for each film, [F.sub.i]. To accommodate the possibly nonlinear relationship between both ARPS and SHARE and the number of weeks a film has been in circulation, the number of weeks a film is in circulation ([WKS.sub.t]) is entered as a fourth-degree polynomial in both equations.

Movie attendance is subject to seasonal fluctuations, generally reaching a peak around Christmas, declining in the late winter and spring months, increasing during the summer vacation months, and then declining again in early fall. [9] Three dummy variables are included to capture these seasonal effects; [Q.sub.2] = 1 for weeks in late winter and spring, [Q.sub.3] = 1 for the summer weeks, and [Q.sub.4] = 1 for weeks during the fall; each variable equals zero otherwise. The period surrounding the Christmas holidays is represented by the omitted variable. A priori, it is difficult to predict the signs of the coefficients for the seasonal dummy variables. The reduction in movie attendance in the fall or spring would be expected to reduce ARPS relative to the summer and Christmas peaks, but it might also lead studios to release fewer movies during these periods, thus reducing competition and possibly increasing ARPS. In the SHARE equation, the increase in movie attendance during the Christmas and summer seasons may lead to an increase in the number of films released, thus increasing the competition for venues and reducing the number of screens per film.

To capture the effects of an Oscar nomination or award we employ two alternative specifications, varying in complexity and flexibility. In Model I the impact of a nomination is captured by [SUP.sub.N], [ACT.sub.N], and [PIC.sub.N], which represent the number of nominations a film receives for best supporting actor/actress, best actress/actor, and best picture, respectively; the corresponding number of awards that a film receives in a given category are designated as [SUP.sub.W], [ACT.sub.W], and [PIC.sub.W]. The values of [SUP.sub.N], [ACT.sub.N], and [PIC.sub.N] are set equal to zero in the weeks prior to the nominations and equal to the number of nominations a film receives for [T.sub.N] weeks following the announcement of the nominations. The values of [SUP.sub.W], [ACT.sub.W] and [PIC.sub.W] are set equal to zero in the weeks prior to the announcement of the awards and equal to the number of awards a film receives for [T.sub.W] weeks following the announcement of the awards. The method used to determine the values of [T.sub.N] and [T.sub.W] is discussed below.

Using the variable definitions given above, Model I may be written as

(1) [Y.sub.i,t] = [[[sigma].sup.361].sub.i=1][[alpha].sub.i][F.sub.i] + [[beta].sub.2][Q.sub.2] + [[beta].sub.3][Q.sub.3] + [[beta].sub.4][Q.sub.4] + [[gamma].sub.1][WKS.sub.t] + [[gamma].sub.2][[WKS.sup.2].sub.t] + [[gamma].sub.3][[WKS.sup.3].sub.t] + [[gamma].sub.4][[WKS.sup.4].sub.t] + [[theta].sub.S][SUP.sub.N] + [[theta].sub.A][ACT.sub.N] + [[theta].sub.P][PIC.sub.N] + [[varphi].sub.S][SUP.sub.W] + [[varphi].sub.A][ACT.sub.W] + [[varphi].sub.P][PIC.sub.W] + [[epsilon].sub.i,t]

where [Y.sub.i,t] = ARPS or SHARE.

Although straightforward, Model I suffers from three shortcomings. First, the effects of a nomination or award are constrained to be constant in the weeks following the announcement dates, ruling out any possible lags experienced by distributors in expanding the distribution of the film, and the information cascades discussed by De Vany and Walls (1996), in which the transmission of information regarding the quality of the film may develop slowly over time. Second, it is not possible to determine the maximum number of weeks that a nomination or award impacts ARPS and SHARE. Finally, in Model I the impact of a nomination or award is assumed to be independent of the film's release date. In general, films released earlier in the year will command a smaller market share of screens and have a lower ARPS at the time of the nominations than films released later in the year. The percentage increase in ARPS or SHARE resulting from a nomination or award is thus expected to be inversely related to the film's release da te.

To accommodate these various factors we develop a more flexible model, Model II. Define [N.sub.k] to be a dummy variable equal to one in the kth week following the announcement of the nominations, and zero otherwise; similarly, [W.sub.m] is defined to be a dummy variable equal to one in the mth week following the announcement of the winners, and zero otherwise. [10] To allow the impact of a nomination or award to vary according to a film's release date, we include WR, defined as the week of the year in which the film was released (WR = 1 in the first week of January and WR = 52 in the last week of December). Employing the above notation, Model II may be written as:

(2) [Y.sub.i,t] = [[[sigma].sup.361].sub.i=1] [[alpha].sub.i][F.sub.i] + [[beta].sub.2][Q.sub.2] + [[beta].sub.3][Q.sub.3] + [[beta].sub.4][Q.sub.4] + [[gamma].sub.1] [WKS.sub.t] + [[gamma].sub.2][[WKS.sup.2].sub.t] + [[gamma].sub.3][[WKS.sup.3].sub.t] + [[gamma].sub.4] [[WKS.sup.4].sub.t] +(1 + [pi]WR){([[phi].sub.S][SUP.sub.N] + [[phi].sub.A] [ACT.sub.N] + [[phi].sub.P][PIC.sub.N]) * ([[[sigma].sup.[T.sub.N]].sub.k=1] [[delta].sub.k][N.sub.k])) + (([[varphi].sub.S][SUP.sub.W] + [[varphi].sub.A][ACT.sub.W] + [[varphi].sub.P][PIC.sub.W]) * ([[[sigma].sup.[T.sub.W]].sub.m=1] [[omega].sub.m][W.sub.m]))} + [[epsilon].sub.i,t].

In Model I the impact of a nomination for best picture, for example, is given by the coefficient [[phi].sub.P]. The resulting change in ARPS or SHARE is assumed to be constant throughout the six-week period between the nominations and awards and independent of the release date. In Model II the impact of a nomination for best picture is given by the term (1+[pi]WR)[[phi].sub.P][[delta].sub.k]. As discussed above, the effect of a nomination or award is expected to be inversely related to the release date, implying that [pi] [less than] 0. To ensure identification of the remaining parameters, we impose the restriction [[delta].sub.1] = [[omega].sub.1] = 1; estimates of [[delta].sub.k] and [[omega].sub.m] greater than 1 imply an increase in the impact of a nomination and award relative to the first week, whereas estimates of [[delta].sub.k] and [[omega].sub.m] less than one imply the opposite. If the impact of a nomination or award is independent of the release date, and constant in the weeks following the announc ement dates, then [pi] = 0 and [[delta].sub.k] = [[omega].sub.m] = 1 for all k and m. Model 1 may thus be viewed as a restricted version of Model II, implying that it is possible to choose between the two on the basis of standard hypothesis tests.

To determine the appropriate functional form for each model we employ a test developed by MacKinnon et al. (1983) to choose between a linear and a log-linear model. In both cases the test results support the use of the log-linear functional form. In Model II a nomination or award is assumed to have an impact on ARPS and SHARE a maximum of [T.sub.N] and [T.sub.W] weeks, respectively. If [T.sub.N] [greater than] 6, the impact of a nomination extends beyond the six-week period between the announcement of the nominations and awards. Initially, both [T.sub.N] and [T.sub.W] are set equal to 20; we then use a Wald test to test the hypothesis that (1 + [pi]WR)[[phi].sub.j]([[delta].sub.k][N.sub.k]) = 0 for [N.sub.k] = 20 and j = P (picture), A (actress/actor), and S (supporting actor/actress). If the null hypothesis is not rejected, we reduce [N.sub.k] by one and continue testing until the null hypothesis is rejected at the 10% confidence level. Using this approach we conclude that the effect on both SHARE and ARPS of a nomination does not exceed the six-week period between the nominations and awards. This result is consistent with the industry's prevailing wisdom that "the party ends on Oscar night. With one exception, every best-picture nominee (in 1993) that was completely shut out at the awards saw its box office immediately head south." [11] Using the same framework we find that the effects of an award do not exceed 11 weeks in the SHARE equation and 4 weeks in the ARPS equation.

IV. EMPIRICAL RESULTS

The estimated coefficients for both Models I and II are presented in Table 1. The parameter estimates common to both models are generally similar in sign, magnitude, and statistical significance, as are the adjusted [R.sup.2]s. A nomination or award for best supporting actor/actress appears to have little if any impact on ARPS or SHARE, as the estimated coefficients for [SUP.sub.N] and [SUP.sub.W] are not statistically significant in either model. A nomination for best actress/actor has a positive and statistically significant impact on SHARE in both models and a positive but insignificant effect on ARPS. An award for best actress/actor has a positive and statistically significant (at the 10% level or better) effect on ARPS and SHARE in both models. Finally, a nomination and award for best picture has a positive and significant impact on ARPS and SHARE in Model I; in Model II the estimated coefficients for [PIC.sub.N] and [PIC.sub.W] are positive and significant at the 1% level in the SHARE equation and posi tive but insignificant at the 10% level in the ARPS equation.

The impact of a nomination or award in the weeks following the announcement date is constrained to be constant in Model I, but is allowed to vary freely in Model II. The estimates of [[delta].sub.k] and [[omega].sub.m] in the SHARE equation are generally consistent with the information cascades discussed by De Vany and Walls (1996), in which the transmission of information regarding the quality of the film develops over time. The increase in the market share of screens resulting from a nomination or award reaches a peak in the third week following the announcement date and then declines thereafter. In the ARPS equation the estimates of [[delta].sub.k] are not statistically significant, while the estimates of [[omega].sub.m] reach a peak in the second week following the Oscars ceremony and decline thereafter. In Model II the impact of a nomination or award is in part a function of WR, the week in which the film was released. The estimated coefficient of WR is negative in both equations, and statistically signi ficant in the SHARE equation at the 1% level. These results are consistent with the hypothesis that films released earlier in the year will generally command a smaller market share of screens and have lower revenues per screen at the time of the nominations than films released later in the year, implying a larger percentage increase in SHARE and ARPS.

We test the null hypothesis that [[delta].sub.k] = [[omega].sub.m] and [pi] = 0 in Model II using a likelihood ratio test. The resulting Chi-squared test statistics are 10.66 and 20.4 for the ARPS and SHARE equations, respectively, which are both below the critical values at the 95% confidence level. Based on these results we are unable to reject the null hypothesis that the impacts of a nomination and award are independent of the release date and constant in the weeks following the announcements of the nominations and awards. On the basis of these test results we employ the estimates from Model I in calculating the predicted values of an Oscar in section VI.

V. NOMINATIONS, AWARDS, AND THE PROBABILITY OF SURVIVAL

The final step in calculating the value of an Oscar is to determine the impact of a nomination or award on the film's probability of survival. As demonstrated in Figure 3, films that receive Oscar nominations or awards generally remain on the Vareity Top 50 chart longer than non-nominated films; an Oscar nomination or award thus enhances a film's probability of survival. To assess the impact of a nomination or award on the probability of a film remaining on the Top 50 chart a given number of weeks or longer, we estimate a parametric survival function. Following De Vany and Walls (1997), we employ a parametric survival function based on the assumption of a Weibull distribution. The Weibull distribution allows for duration dependence [12] and facilitates the incorporation of covariates. The survival function for the Weibull distribution with heterogeneity, which represents the probability of a film remaining on the charts [t.sub.k] weeks, is given by

(3) P([t.sub.k]) = [[1 + [theta][([lambda][t.sub.k]).sup.[rho]]].sup.-1/0],

where t represents the length of a film's run on Variety's Top 50 chart and [lambda] [rho], and [theta] are parameters to be estimated. The coefficient [theta] is included to allow for unobserved film attributes and is a measure of heterogeneity. If [theta] = 0, the assumption of heterogeneity may be rejected, and the model reduces to the standard Weibull model. [13] The coefficient p is included to allow for duration dependence. The probability that a film will leave the Top 50 chart at time t + 1, conditional on having survived until time t (the hazard rate) is independent of time if [rho] = 1, increasing with respect to time if [rho] [less than] 1, and decreasing with respect to time if [rho] [greater than] 1. Finally, to incorporate the effect of the covariates, we redefine [lambda] as [lambda] = [e.sup.-[beta]'X], where the vector X includes indicators for nominations or awards.

To analyze duration, each film is followed from the time it first appears on the Top 50 chart until it leaves the theaters for the last time. A nomination or award might (i) extend the length of time a film remained in the theaters, or (ii) cause a film that had disappeared from the theaters to be rereleased. To accommodate the possibility of multiple spells, we employ a model specification developed by Hamerle (1989). We define a spell to be the number of consecutive weeks that a film remains on the Top 50 chart. In this framework a film's spell can end either when the film disappears from the chart or, for films still in distribution at the time of the nominations and awards, at the announcement date of the nominations or awards. In the latter case the spell is treated as a censored spell, because the conditions under which the film is in the theaters has changed. The log of the likelihood of a spell that crosses the nominations or award date therefore factors into the sum of the log likelihoods for the se gments of time before and after the nomination or award, since the value of the indicator variable is constant in those segments.

To understand the possible contributions of spells of different types to the likelihood, consider Figure 4. A spell could start and end before the nomination date, start and end between the nomination date and the award date, or start and end after the award date, as represented by spells #1, #5, and #6, respectively. In these cases the spell has only one contribution to the likelihood, equal to the number of weeks in length. Alternatively, a spell could cross the nomination date or both the nomination and award dates, as represented by spells #2, #3, and #4. Spells #2 and #4 make two contributions to the likelihood, a spell censored at the nomination or award date, and a subsequent, uncensored spell. Spell #3 makes three contributions to the likelihood, as two censored spells and one uncensored spell. The 262 films in our sample contributed a total of 456 spells, of which 292 were completed and 164 were censored.

The probability of survival in a given week is assumed to be a function of the nominations and awards a film receives, its release date, and the length of the prior spells. To model nominations and awards, indicator variables were used for best picture, best actress/actor, and best supporting actor/actress. The three 0/1 variables for nominations were set equal to zero until the time of the nominations, and then "turned on" if appropriate. They retained the value of 1 for the remainder of the film's run, or subsequent runs if the film had been nominated. The three award variables were set to 0 until the date of the awards (usually six weeks after the nominations) and changed to 1 if the picture won an award. Films released earlier in the year generally face less competition for screens than films released later in the year, thus enhancing their odds of survival in a given week. We thus include the release date of the film as an independent variable, with the expectation of a negative relationship between the release date and the probability of survival. Finally, a lagged duration variable (defined to be the length of the previous spell) is included in the belief that the probability of survival during a given spell is inversely related to the length of the previous spell. Because the values of these variables may change over time, the time varying covariate model developed by Petersen (1986) was employed.

The estimated parameters of the survival function, together with their asymptotic t-statistics, are presented in Table 2. The results indicate that a nomination or an award for best picture has a positive and statistically significant effect on the survival of a film. In the six-week period between the nominations and awards the nominees for best actor/actress and best supporting actor/actress appeared to be crowded out of the theaters by the nominees for the best picture award, as the coefficients of both variables are negative, although only the former is statistically significant. In the weeks following the announcement of the awards, however, the films receiving an Oscar for best actor/actress and best supporting actor/actress experience a positive and statistically significant increase in the probability of survival. The coefficient of the release date variable is negative and statistically significant, indicating that films released earlier in the year have a greater chance of survival than films relea sed later in the year. Finally, the negative and statistically significant coefficient for the lagged duration variable indicates that the probability of survival in the next spell is inversely related to the length of the previous spell. Thus, for example, films that were in the theaters for 20 weeks prior to the nominations have a lower probability of survival following the nominations than films that were in the theaters for 10 weeks prior to the nominations.

VI. PREDICTED VALUES OF AN OSCAR

Using the estimated coefficients from the SHARE, ARPS, and survival equations, we compute predicted values for an Oscar nomination and award for a representative film released in the middle of each quarter of the year. We derive the estimated number of screens per week for a non-nominated film by setting [Q.sub.2] equal to 1, setting each of the nomination and award variables equal to 0, and then calculating the predicted values of SHARE for 6 weeks following the nominations and 11 weeks following the awards. Multiplying the estimates of SHARE by the total number of movie screens in the United States during 1997 yields an estimate of the number of screens on which a non-nominated film would appear. We repeat the process for the ARPS and survival equations to derive weekly estimates of the average revenue per screen and the probability of survival in a given week. Multiplying the predicted probability of survival by the predicted number of screens and average revenues per screen yields an estimate of the pred icted box office receipts in the weeks following the nominations and awards. The entire process is repeated for the films receiving nominations or awards by setting the appropriate nomination or award variable equal to 1 for the relevant period. Finally, the predicted value of an Oscar nomination or award is derived as the difference in predicted box office receipts for the nominated and non-nominated films.

In the weeks following the nominations and awards films nominated for best picture and actress/actor significantly increase their market share of screens, as theater owners scramble to book the films with the greatest box office appeal. The results from Model I indicate that a nomination and award for best actress/actor increase the number of screens on which the film appears by 41.16% and 122.32%, respectively, whereas a nomination and award for best picture increases the number of venues by 84.95% and 200.76%, respectively. Films receiving only a nomination for best supporting actor/actress are apparently squeezed out of the theaters by films that receive nominations for more prestigious awards, as the number of screens on which they appear decline by 4.76% in the weeks following the nominations. An award for best supporting actor/actress, however, leads to a 23.99% increase in the number of screens. Using the same approach we derive estimates of the increase in ARPS resulting from a nomination or award. T he predictions obtained from Model I indicate that a nomination for best supporting actor/actress, best actress/actor, and best picture would increase ARPS by 3.45%, .96%, and 7.52%, respectively; the corresponding results for an award are 12.46%, 21.53%, and 14.79%, respectively.

We illustrate graphically the impact of a nomination or award on predicted box office revenues for a hypothetical film released in the middle of the fourth quarter. The impact of winning an Oscar is greatest in the first four weeks following the awards ceremony, as an award increases both ARPS and SHARE during this period; only the predicted values of SHARE are impacted by an award in the remaining seven weeks. As depicted in Figure 5, a nomination for best supporting actor/actress has a minimal impact on expected revenues, but an award increases predicted revenues by an average of 39.44% in the first four weeks and by 23.99% in the following seven-week period. A nomination for best actress/actor increases expected revenues by an average of 42.52% during the six-week period following the nominations; films receiving an award for best actress/actor experience an average increase in expected revenues of 170.18% in the first four weeks following the awards ceremony and by 122.32% in the remaining seven-week per iod. Finally, a nomination for best picture increases expected revenues by an average of 98.85% in the six-week period between the nominations and awards, whereas an award for best picture increases revenues by 245.23% and 200.76% in the first four weeks and last seven weeks, respectively, in the post-award period.

In Table 3 we present two different sets of estimates for the predicted value of an Oscar nomination or award; in each case it is assumed that the film was released at the midpoint of a given quarter. The values given for winning an award represent the incremental value of an award over and above that of a nomination. The values in parentheses represent the percentage of films in a given category released in the corresponding quarter; for example, 18% of the films nominated for a best picture award were released in the first quarter. In the top panel the estimates are calculated by multiplying the predicted probability of survival in week t by the predicted ARPS and number of screens for that week; in the lower panel the estimates are computed by multiplying the predicted value of ARPS by the predicted number of screens for week t. By comparing the two sets of estimates it is possible to assess the consequences of failing to control for the probability of survival when computing the value of an Oscar.

The estimated values of an Oscar nomination and award reported in the top panel of Table 3 are generally consistent with previous industry estimates. The fourth-quarter and weighted average estimates of $23,860,527 and $17,489,153 for a best picture nomination and award are well within the $10-$40 million range cited in industry trade publications. [14] The fourth-quarter and weighted average value of a nomination and award for best actress/actor of $6,308,783 and $4,511,640 are quite similar to the estimated value of $5,824,400 reported by Harmetz (1986). [15] Finally, the fourth-quarter and weighted average value of $2,229,737 and $1,465,939, respectively, for a nomination and award for best supporting actor/actress is consistent with the industry's belief that the lesser awards are worth substantially less than the awards for best actress/actor and best picture.

The estimates presented in the lower panel of Table 3 illustrate the consequences of failing to control for the probability of survival when computing the value of an Oscar nomination or award. Two generalizations may be drawn from these results. First, films released early in the year have a very low probability of being in distribution at the time of the nominations or being rereleased following the nominations. For example, based on the parameter estimates presented in Table 2, a film released in the first quarter that receives the award for best picture has only a 38.3% probability of being in the theaters during the first week following the awards ceremony; non-nominated films and films receiving lesser nominations or awards have substantially lower probabilities. Failure to account for these low probabilities of survival may lead to significant overestimates of the value of a nomination or award for films released early in the year. Second, our results indicate that a nomination for best actress/actor a ctually reduces a film's probability of survival in the period following the announcement of the nominations. Failure to control for this result leads to estimated values of a nomination or award for best actress/actor that greatly exceed the usual industry estimates.

Film studios and distributors must decide on the optimal release strategy to maximize the present value of the film's profits. [16] The prevailing wisdom in the industry is expressed by King (1992), who states, "Clearly, the lure of added box-office punch that normally comes with the best-picture Oscar is another reason studio executives have preferred to open promising movies at the end of the year." This practice may explain why every film to receive an award for best picture between 1980 and 1990 was released in the last four months of the year. [17] This theory is based in part on a belief that delaying a film's release will ensure that the film is fresh in the minds of the members of the Academy when they cast their ballots. Studios also delay the release of promising films until late in the year in an attempt to take advantage of the peak moviegoing period in the weeks surrounding the Christmas holidays and to improve a film's chances of remaining in the theaters at the time of the nominations in mid- to late February. A late release date also reduces the chances that a film will be in video distribution at the time of the nominations, which eliminates any box-office "boost" from a nomination or award. For example, after winning five Oscars in 1992 (including best picture), a rerelease of Silence of the Lambs produced per-screen revenues of less than $1,500 because the film was available on video and cable TV. [18]

The strategy of delaying a film's release until late in the year is not without risk, however, as it forces a film to compete with the large number of films released around the Christmas holidays for additional distribution or screens. Theater owners actively bid for the right to show the films they think will prove to be most popular, including those films nominated for the major Oscar awards. Films that fail to receive a nomination or are nominated for less prestigious awards may be losers in the battle for theater screens in the weeks following the nominations and awards. For example, using the predicted box office revenues for the week proceeding the nominations or awards as our benchmark, we find that weekly revenues for a non-nominated film decline by 51.88% in the 6-week period following the nominations and an additional 53.32% during the 11-week period following the awards. A nomination or award for best supporting actor/actress does little to diminish this rate of decrease, as box office revenues fo r these films decline by 52.85% and 42% in the 6-week period following the nominations and 11-weeks following the awards, respectively. These patterns may be compared with those experienced by films nominated or receiving more prestigious awards. Revenues for films receiving a nomination for best actress/actor or best picture decline by 31.41% and 11.0% in the 6-week period following the nominations; films receiving an award in these categories experience an increase of 3.99% and 40.68% in revenues in the 11-week period following the award ceremony.

The results presented in Table 3 illustrate the potential benefits of delaying the release of films with strong Oscar potential until late in the year to increase both their chances of receiving an Oscar nomination and the resulting financial rewards. Delaying the release date from the first to fourth quarter increases the predicted payoff for a nomination for best picture from $673,082 to $7,829,797, an increase of 1,063%, while the predicted payoff for winning the award for best picture increases from $2,737,124 to $16,030,730, an increase of 485.7%. Delaying the release of a film that is nominated for or receives the award for best actress/actor from the first to fourth quarter increases the predicted increase in box office revenues by 3,819% and 640%, respectively.

There are several reasons to believe that the estimates reported in Table 3 represent the lower bound of the true value of an Oscar. First, the sample used in this study is made up of an equal number of nominated and non-nominated films released at approximately the same time. In each case the highest ranked non-nominated film that remained on the top 50 chart for a minimum of five weeks was chosen as the control. The use of a control group comprised solely of the highest-ranked films may impart a downward bias to our estimates of the value of an Oscar. Second, the estimates in Table 3 ignore the impact of an Oscar on foreign box office receipts, which frequently equal or exceed those collected domestically. "While a best picture nod or a clutch of Oscars is most significant in terms of domestic box office, the awards impact is even greater internationally. Historically, the timing of the award is more likely to compliment the foreign release patterns of the majority of winners." [19] Finally, the estimates in Table 3 fail to include the impact of a nomination or award on the revenues received from video sales and rentals and from cable TV and pay-per-view movies. [20]

Markets organized as tournaments have two features in common. First, the compensation or prize awarded to each participant is based on relative not absolute performance. Second, the optimal compensation scheme is usually nonlinear and highly skewed in favor of the higher-ranked participants. For the 47% of the films in our sample that receive a single Oscar nomination, the annual competition for an Oscar may be viewed as a two-stage, single-elimination tournament, in which films first compete for a nomination with the successful nominees then competing for an award. For the majority of films that receive multiple nominations, the competition may be viewed as both a multistage tournament within a given category and a competition across categories for the more prestigious awards, like best and actress/actor and best picture.

For each of the four release dates we compute the absolute value of the percentage difference between the value of a nomination and award for all three categories. Based on the average of the ratios, the percentage difference between the value of a nomination and award is 1,664%, 2,451%, and 231% for best supporting actor/actress, best actress/actor, and best picture, respectively. This reward structure is consistent with Rosen's (1988) discussion of multistage tournaments, in which the prize for winning the final stage is disproportionately larger than the prizes awarded at the earlier stages. To explore the structure of rewards across awards we compute the absolute value of the percentage difference between the value of a nomination, award, and the combined value of nomination and award across categories. Averaging across the four release dates, a nomination for best picture is worth 2,366% as much as a nomination for best actress/actor, which in turn is worth 268% as much as a nomination for best supporti ng actor/actress, whereas an award for best picture is worth 235% as much as an award for best actress/actor, which in turn is worth 107% as much as an award for best supporting actor/actress. Finally, the combined value of a nomination and award for best picture is worth 318% as much as the combined value for best actress/actor, which in turn is worth 136% as much as the combined value for best supporting actor/actress. Ehrenberg and Boganno (1990) report that the prize money awarded to the first-place finisher in Professional Golf Association (PGA) tournaments is typically 1.67 times the prize awarded to the runner-up, which in turn is 1.59 times the prize awarded to the third-place finisher. If one views a nomination or award for best picture, best actress/actor, and best supporting actor/actress as the top three prizes in the Oscar competition, the prize structure in the competition for an Oscar appears to be more convex than the prize structure in PGA tournaments.

Our findings of a highly convex structure of rewards for the Oscar nominees and recipients are consistent with the results presented by De Vany and Walls (1997), who found that weekly box-office receipts were highly convex with respect to a film's ranking on the top 50 chart. They report that a decline in rank from first to second reduced weekly revenues by $2.4 million, while a drop from fourth to fifth reduced revenues by only $235,000. Both results are consistent with Rosen's (1981) theory of superstars, in which a relatively small number of firms or performers dominate their industry, earning a disproportionate share of the revenues or profits. [21] Rosen's theory applies to markets in which the goods or services provided are viewed by consumers as imperfect substitutes, and in which the "services rendered by any seller become more like a kind of public good the more nearly the technology allows perfect duplication at constant cost." [22] The industry characteristics cited by Rosen as being conducive to giving rise to superstars are consistent with the movie industry, given the ability of movie studios to create duplicate prints at a (presumably) constant marginal cost and the fact that most films are imperfect substitutes for one another.

VII. SUMMARY AND CONCLUSIONS

This article has employed weekly box-office data to examine the market's response to the yearly announcement of the Oscar nominations and awards. Our results indicate that a nomination or award for the "top" prizes, such as best picture and actress/actor, generally has a positive impact on a film's probability of survival, its market share of screens, and the average revenue per screen, while a nomination or award for "lesser" prizes, such as best supporting actor/actress, has little if any impact on these variables. Controlling for release date, we find that a nomination for best supporting actor/actress, best actress/actor, and best picture increases predicted box office revenues by $-147,131, $476,617, and $4,799,118, respectively; the corresponding increases for an award are $1,612,939, $4,035,023, and $12,690,035, respectively.

Our results indicate that failure to control for the probability that a film will be in distribution during the weeks following the nominations and awards may lead to overestimates of the value of an Oscar nomination and award, particularly for films released earlier in the year. In addition, our results offer support for the common industry practice of delaying a film's release until late in the year to enhance the probability of receiving a nomination and the resulting financial rewards. For example, the predicted payoff for a nomination for best picture increases from $673,082 to $7,829,797 as a result of delaying the release date from the first to fourth quarter of the year, while the predicted payoff for winning the award for best picture increases from $2,737,124 to $16,030,730, an increase of 485.7%.

Our findings indicate that Rosen's (1981) theory of superstars, in which small differences in talent or quality translate into large differences in earnings, may well be applicable to the motion picture industry. DeVany and Walls (1997) found that weekly box office receipts are highly convex with respect to a film's ranking on Variety's Top 50 chart, a finding that is consistent with the structure of rewards in the competition for an Oscar. Our findings also suggest that the competition for an Oscar may be viewed as a two-stage, single-elimination tournament, in which films first compete for a nomination with the survivors then competing for the award. In such tournaments the overall reward structure must be convex, and the reward for the survivor of the second stage must be significantly larger than the prizes awarded at the earlier stages. This latter requirement is satisfied in the competition for an Oscar, as the value of the award is several times the value of a nomination.

Nelson: Professor, Department of Economics, Colby College, Waterville, ME 04901. Phone 1-207-872-3567, Fax 1-207-872-3263, E-mail ranelson@colby.edu

Donihue: Associate Professor, Department of Economics, Colby College, Waterville, ME 04901. Phone 1-207-872-3115, Fax 1-207-872-3263, E-mail mrdonihu@colby.edu

Waldman: Professor, Department of Economics, University of Colorado, Boulder, CO 80302. Phone 1-303-492-6781, Fax 1-303-492-8960, E-mail waldman@spot.colorado.edu

Wheaton: Associate, Deutsche Banc Alex. Brown, Baltimore, MD 21202-3220. Phone 1-410-895-4570, Fax 1-410-895-4582. Email cal.wheaton@db.com

(*.)We are indebted to Scott Masten and two anonymous referees for helpful comments that greatly improved the final version of this paper, and to Michael Murray, Kerry Smith, and seminar participants at the Colby-Bates-Bowdoin seminar series and the Southern Economic Association Meetings in New Orleans for helpful comments on an earlier draft. Financial assistance from the Robert N. Anthony and Douglas Chair funds at Colby College is gratefully acknowledged.

(1.) The number of categories, and the categories themselves, have changed over time. In 1927-28, awards were first given in 12 categories; the number of categories have increased over time. In addition, each year a number of awards are given for special achievement, scientific merit, and so on.

(2.) Gumbel et al. (1998, W4), for example, estimate that an award for best supporting actor/actress could be worth anywhere between $250,000 and $2.5 million dollars in higher future compensation, depending on the attributes of the actor/actress.

(3.) Other forms of professional recognition, such as critics' ratings, Golden Globe awards, selection for the Cannes or Sundance film festivals, and so on will almost certainly affect box-office receipts. Studies that utilize cumulative box-office revenues to estimate the value of an Oscar, such as Littman (1983) and Smith and Smith (1986), may thus yield biased estimates unless they control for all such awards and evaluations. We minimize this problem by employing data on weekly rather than cumulative box-office receipts. Our results may be biased if the impact of these other forms of recognition are long-lived, and thus impact box-office receipts in the weeks following the Oscar nominations and awards. The results presented in section IV indicate that such a bias is unlikely, as the impact of a major award (such as an Oscar) affects box-office receipts for a relatively limited period of time.

(4.) The sample ends in 1987 because in 1988 Variety's chart of "50 Top-Grossing Films" was reduced to a sample of only the top 20 films. De Vany and Walls (1996, 1997) also employ the Variety Top 50 data in their studies of rank tournaments and adaptive contracting in the motion picture industry.

(5.) Between 1979 and 1988 the number of theaters in the sample increased from an average of 1,600 to 2,750, an increase of 72%. The total number of theaters in the United States increased from 16,755 in 1978 to 23,555 in 1987, or 41%.

(6.) It is possible for a film to receive multiple nominations in a single category; for example, a film could receive two nominations for best supporting actor or actress.

(7.) The averages are constructed for only those films still in the theaters; the number of films in the sample declines over time. For example, only two of the non-nominated films survived to the twentyfifth week.

(8.) Industry public relations executive Mark Angelotti, quoted in Dretzka (1998, 8).

(9.) Vogel (1990) contains a discussion of the seasonal fluctuations in movie attendance.

(10.) For films that win an Oscar, the effects of the award are assumed to supersede the effects of a nomination. To accommodate this, values of [N.sub.k] for k [greater than] 6 are set equal to zero for the films that win awards. This restricts the effects of a nomination to the six-week period between the announcement of the nominations and awards for the films that win an Oscar but allows the impact of a nomination to continue after the announcement of the winners for those films that fail to win an award.

(11.) Klady (1994).

(12.) DeVany and Walls (1997) investigate the number of weeks that films survive on Variety's top 50 list. Based on parametric and nonparametric analysis of the data, they conclude that the hazard function, which represents the probability that a movie will disappear from the charts between time t and t + 1 given that it is on the charts at time t, is an increasing function of time. The Weibull distribution is consistent with a hazard function that is increasing, decreasing, or constant with respect to time.

(13.) A similar approach has been employed by Hausman et al. (1984). The heterogeneity assumption has been criticized as being overparameterized by Heckman and Singer (1984), who introduced a nonparametric version. Kiefer (1988) suggests that the method is valid when the underlying hazard function is correctly specified.

(14.) For example, see Tobenkin (1992).

(15.) Harmetz (1986) estimated that an award for best actor/actress was worth a maximum of $4 million in 1986. We convert this estimate to 1997 dollars using the Consumer Price Index for entertainment expenditures to obtain the figure of $5,824,400.

(16.) In addition to picking the date of release to movie theaters, film distributors must also determine when to release the film to foreign markets, domestic and overseas video markets, cable and pay-per-view TV, and network and local television stations. The order of release may be viewed as a form of intertemporal price discrimination, in which distributors release the film to consumers with the highest willingness to pay or profit margin first, and so on down the line. For example, Waterman (1987) estimates that theatergoers provide distributors with a net profit of $1, while consumers who rent videocassettes or see the film on pay-per-view cable or network TV contribute $.35, $.20, and $.05, respectively. See Waterman (1985), Owen and Wildman (1992), or Frank (1994) for further discussion of the optimal release strategy for film distributors.

(17.) See King (1992). The practice of employing the release date as a competitive tool has spread to the publishing industry. Publishers now time the release of books in an effort to avoid competition from major authors. "More and more, the three rules of publishing seem to be timing, timing, timing. Savvy publishers spend as much time analyzing the best possible month to bring out a book as analyzing the attractiveness of the jacket and the promotibility of the author" (Kaufman [1996]).

(18.) See Klady (1994).

(19.) Variety, April 3, 1995, 20. Films must be released domestically by December 31 of a given year to be considered for a nomination in February of the following year. Because the majority of nominated films are released in the fourth quarter and foreign releases almost always follow domestic release dates, there is a high probability that nominated films will he showing in foreign markets when the nominations and awards are announced. This release pattern increases the foreign box-office boost of an Oscar nomination or award. See Groves (1994) or Harris (1994) for the impact of an Oscar on overseas box-office revenues.

(20.) This omission may be significant, as video rentals and cable TV are important sources of revenue for film distributors. For example, Harmetz (1984) reports that in 1983 theatrical film rentals, pay cable TV, and home video cassettes contributed 63%, 24%, and 13%, respectively, of domestic film revenues. We are unable to obtain any evidence indicating the impact of an Oscar nomination or award on cable TV or home video revenues.

(21.) We are indebted to Ernie Berndt for suggesting this interpretation.

(22.) Rosen (1981, 849).

REFERENCES

De Vany, A., and W. D. Walls. "Bose-Einstein Dynamics and Adaptive Contracting in the Motion Picture Industry." Economic Journal, November 1996, 1493-1514.

-----. "The Market for Motion Pictures: Rank, Revenue, and Survival." Economic Inquiry, October 1997, 783-97.

Dretzka, G. "Chasing Oscar: Campaigns to Land an Academy Award Nomination Are Expensive, Political, and Occasionally Bizarre." Boston Globe, 4 January 1998, section 7, p. 1.

Ehrenberg, R., and M. Boganno. "Do Tournaments Have Incentive Effects?" Journal of Political Economy, December 1990, 1307-24.

Groves, D. "Oscarites Get Foreign B.O. Boost." Variety, March 28--April 3, 1994, p. 16.

Gumbel, P., J. Lippman, L. Bannou, and B. Orwall. "What's an Oscar Worth?" Wall Street Journal, March 20, 1998, sec. W, p. 1.

Hamerle, Alfred. "Multiple-Spell Regression Models for Duration Data." Applied Statistics 38, 1989, 127-38.

Harmetz, A. "Hollywood Thriving on Video-Cassette Boom." New York Times, May 7, 1984, sec. A, p. 1.

-----. "Will Oscar Help Low-Budget Films?" New York Times, March 27, 1986, sec. C, p. 23.

Harris, M. "Oscar Noms Boost Pix O'Seas." Variety, March 7-13, 1994, p. 16.

Hausman, J., B. Hall, and Z. Griliches. "Economic Models for Count Data with an Application to the Patents--R&D Relationship." Econometrica, July 1984, 909-38.

Heckman, J., and B. Singer. "A Method for Minimizing the Impact of Distributional Assumptions in Econometric Models for Duration Data." Econometrica, March 1984, 271-320.

Kaufman, J. "Landing on the Bestseller List: It's All in the Timing." Wall Street Journal, April 23, 1996, sec. A, p. 20.

Kiefer, N. "Economic Duration Data and Hazard Functions." Journal of Economic Literature, June 1988, 646-79.

King, T. R. "Oscar Victory by 'Silence of the Lambs' Has Studios Rethinking Release Strategy." Wall Street Journal, April 1, 1992, p. B4.

Klady, L. "When Oscar Talks, the B.O. Listens," variety, April 4-10, 1994, p. 7.

Littman, B. R. "Predicting Success of Theatrical Movies: An Empirical Study." Journal of Popular Culture, Spring 1983, 159-75.

MacKinnon, J. G., H. White, and R. Davidson. "Tests for Model Specification in the Presence of Alternative Hypotheses: Some Further Results." Journal of Econometrics, January 1983, 53-70.

Owen, B. M., and S. S. Wildman. Video Economics. Cambridge, MA: Harvard University Press, 1992.

Petersen, T. "Estimating Fully Parametric Hazard Rate Models with Time-Dependent Covariates." Sociological Methods and Research 14, 1986, 219-46.

Rosen, S. "The Economics of Superstars." American Economic Review, December 1981, 845-58.

-----. "Promotions, Elections, and Other Contests." Journal of Institutional and Theoretical Economics 144, 1988, 73-90.

Smith, S. P., and V. K. Smith. "Successful Movies: A Preliminary Empirical Analysis." Applied Economics, May 1986, 501-7.

Tobenkin, D. "Winning a Major Oscar Can Mean up to $40 Million--and Hollywood Knows It." Gazette (Montreal), February 1, 1992, sec. E, p. 8.

Vogel, H. Entertainment Industry Economics: A Guide for Financial Analysis, 2nd edition. New York: Cambridge University Press, 1990.

Waterman, D. "Prerecorded Home Video and the Distribution of Theatrical Feature Films," in Video Media Competition: Regulation, Economics, and Technology, edited by B. Owen. New York: Columbia University Press, 1985.

-----. "Electronic Media and the Economics of the First Sale Doctrine," in Entertainment, Publishing, and the Arts, edited by R. Thorne and J. D. Vierra. New York: Clark Boardman, 1987. Parameter Estimates Model I Model II Variable SHARE Equation ARPS Equation SHARE Equation ARPS Equation Intercept -7.2037 10.6426 -6.0309 10.2944 (27.81) (111.17) (30.70) (107.61) [Q.sub.2] .2647 -.0840 .2736 -.0848 (4.52) (3.88) (4.623) (3.92) [Q.sub.3] -.0452 -.0685 -.0047 -.0673 (.64) (2.62) (.06) (2.57) [Q.sub.4] .1732 -.1437 .1723 -.1430 (2.74) (6.16) (2.71) (6.13) LnWKS -.0991 .0073 -.0567 .0116 (.40) (.08) (.23) (.13) Ln[WKS.sup.2] .6050 -.2935 .5289 -.3013 (2.29) (3.02) (1.97) (3.10) Ln[WKS.sup.3] -.3271 -.0210 -.2858 -.0169 (3.25) (.57) (2.78) (.46) Ln[WKS.sup.4] .0387 .0147 .0319 .0141 (3.10) (3.22) (2.47) (3.07) WR -.0077 -.0074 (2.97) (1.36) [PIC.sub.N] .6149 .0725 .7956 .1602 (5.54) (1.77) (3.41) (1.52) [ACT.sub.N] .3448 .0096 .4248 .0807 (4.33) (.33) (2.91) (1.08) [SUP.sub.N] -.0487 .0340 -.0354 .0730 (.62) (1.18) (.35) (1.07) [N.sub.2] 1.0912 .7146 (3.63) (2.14) [N.sup.3] 1.2546 .3967 (3.78) (1.31) [N.sub.4] 1.2319 .1546 (3.76) (.54) [N.sub.5] 1.0193 .4024 (3.30) (1.22) [N.sub.6] .8306 .4676 (2.92) (1.41) [PIC.sub.W] 1.1011 .1379 1.5060 .2155 (7.32) (1.79) (3.224) (1.40) [ACT.sub.W] .7990 .1950 1.0887 .2236 (7.43) (3.72) (3.76) (1.94) [SUP.sub.W] .2150 .1175 .1916 .1187 (1.61) (1.91) (1.11) (1.39) [W.sub.2] 1.2907 1.83 (3.76) (2.24) [W.sub.3] 1.3602 1.1963 (3.80) (1.97) [W.sub.4] 1.3421 .6689 (3.77) (1.40) [W.sub.5] 1.1542 (3.55) [W.sub.6] 1.2638 (3.64) [W.sub.7] 1.1302 (3.44) [W.sub.8] 1.1435 (3.31) [W.sub.9] .9274 (3.09) [W.sub.10] .9518 (2.98) [W.sub.11] .8176 (2.44) [R.sup.2] .4087 .7360 .4017 .7174

Notes: The estimates were obtained by estimating a fixed-effects model containing dummy variables for each film; the film-specific parameters are not reported here, but are available from the authors on request. Values in parentheses are t-statistics for Model I and asymptotic t-statistics for Model II. Parameter Estimates Survival Model Estimated Asymptotic Variable Parameter t-Statistic Constant 3.7745 35.65 Release date -0.0113 4.27 Nomination: best picture 0.4672 3.46 Nomination: best actor/actress -0.4239 5.07 Nomination: best sup. actor/actress -0.0514 0.72 Award: best picture 0.5216 2.32 Award: best actor/actress 0.4020 2.75 Award: best sup. actor/actress 0.4516 2.60 Lagged duration -0.0255 6.74 ln([sigma]) -0.5340 11.53 ln([theta]) 0.9750 15.11 Mean log-likelihood: -3.8631 Predicted Values of an Oscar Nomination and Award Release Date 1st Quarter 2nd Quarter 3rd Quarter Estimates adjusted for probability of survival Best supporting actor/actress Nomination -$20,677 -$26,374 -$58,159 (10%) (9%) (25%) Award $383,336 $482,085 $872,727 (15%) (15%) (15%) Best actress/actor Nomination $19,059 $27,675 $87,365 (13%) (7%) (19%) Award $751,737 $957,244 $1,796,299 (20%) (0%) (15%) Best picture Nomination $673,082 $857,981 $1,906,171 (18%) (8%) (20%) Award $2,737,124 $3,402,309 $5,998,613 (0%) (0%) (33%) Estimates not adjusted for probability of survival Best supporting actor/actress Nomination -$19,372 -$24,468 -$41,062 (10%) (9%) (25%) Award $719,805 $811,410 $1,132,047 (15%) (15%) (15%) Best actress/actor Nomination $561,733 $709,527 $1,190,700 (13%) (7%) (19%) Award $3,435,082 $3,862,363 $5,366,594 (20%) (0%) (15%) Best picture Nomination $1,269,787 $1,605,363 $2,698,171 (18%) (8%) (20%) Award $5,378,222 $6,035,574 $8,360,194 (0%) (0%) (33%) Weighted 4th Quarter Average Estimates adjusted for probability of survival Best supporting actor/actress Nomination -$228,839 -$147,131 (56%) Award $2,458,576 $1,612,939 (55%) Best actress/actor Nomination $746,889 $476,617 (61%) Award $5,561,894 $4,035,023 (65%) Best picture Nomination $7,829,797 $4,799,118 (54%) Award $16,030,730 $12,690,035 (67%) Estimates not adjusted for probability of survival Best supporting actor/actress Nomination -$129,436 -$86,889 (56%) Award $2,455,762 $1,750,158 (55%) Best actress/actor Nomination $3,753,359 $2,638,474 (61%) Award $11,548,494 $8,998,527 (65%) Best picture Nomination $8,531,495 $5,503,632 (54%) Award $17,879,921 $14,704,488 (67%)