What is the value of educational broadcast?∗ Jo˜ao Pedro Azevedo World Bank First Version: March 2008 Revision: February 2009 Abstract Over the years, several developed and developing countries have used television broadcasting as a tools to enhance the dissemination of specific educational programs, as well as information and values to a large segment of their population. This paper sets out to measure the social value of one of this initiatives in the context of a developing country, namely the Futura TV channel in Brazil. The Futura is a communication project, ran by a nonprofit foundation financed exclusively by private sector donations in Brazil. With almost 10 years of existence, Futura broadcast reaches almost 50% of the Brazilian population, and has approximately 33 million regular viewers in the whole country. Contingency valuation (CV) techniques using the referendum design were used to estimate the willingness to pay of a representative sample of the Brazilian adult population. Using both parametric and non-parametric techniques this paper is able to estimate the population willingness to pay for Futura, even in a sample in which a 60% of the population choose not to consume the good. The analysis suggest that the willingness to pay for the channel is positive and at least 16 times greater than the channel operation cost. The average minimum willingness to pay estimated using the Turnbull nonparametric model was R$ 3.7 per month, while the parametric average willingness to pay estimated with the Spike model suggested a value of R$ 5.9. When multiplied by the 5 million households with at least one member willing to pay something for the Futura TV channel, the total annual social value of the channel ranges from 330 to 550 million reais. Moreover, the complementary social mobilization activities promoted by the Futura channel seems to have a positive impact on its social value, suggesting that the method was successful in terms of disentangling embedded effects. Key-words: Cultural public goods, Television, Contingent valuation by referendum, Negative willingness to pay,Spike, Turnbull, Canal Futura, Brazil. JEL classification: C14; C35; H41; L31; Z10 ∗ The present work was greatly improved by the several comments and suggestions, in particular the ones from L´ıgia Vasconcellos, Na´ercio Menezes, Walter Belluzzo, Andr´e Portela, and Ricardo Paes de Barros. The author would also like to thank L´ ucia Ara´ ujo, Monica Dias Pinto, Marisa Vasssimon, Carlos Fraga and Marcelo Pessoa. The comments and suggestions of Mariana Iootty and Francisco Ebeling are also greatfully acknowledged. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author. They do not necessarily represent the views of the International Bank for Reconstruction and Development / World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. 1 Electronic copy available at: http://ssrn.com/abstract=1332786 1 Introduction We live in the world where old and new problems coexist. The persistency of poverty and extreme poverty; sever food security crises; political freedom violations; significant gender disparities; and substantial treats to the environment and the sustainability of social and economic achievements (Sen, 1999). On the other hand, constant technological advancements, creativity, and knowledge have been increasing the number, range and depth of social programs available to policy makers, decision takers, and society as whole, to chose from. Faced with this situation, it is important to assure that the solutions on the table are both theoretically and empirically sound, and benefit from the analysis of past failures and successes. In this aspect, it is particularly relevant for policy makers to focus on the application of rigorous evaluation techniques of their social projects, in order to better understand the key leverage points for future program improvement, as well as the overall value of their efforts. Over the years, several developed and developing countries have used television broadcasting as a tools to enhance the dissemination of specific educational programs, as well as information and values to a large segment of their population. This paper sets out to measure the social value of one of this initiatives in the context of a developing country, namely the Futura TV channel in Brazil. The Futura is a communication project, ran by a non-profit foundation financed exclusively by private sector donations in Brazil. With almost 10 years of existence, Futura broadcast reaches almost 50% of the Brazilian population, and has approximately 33 million regular viewers in the whole country. The methodological choice made was to apply an economic technique, known by the specialized literature as contingent valuation (CV) to measure the willingness to pay (WTP) of the Brazilian population for the Futura project. This technique has been traditionally applied on valuation exercises of public goods with significant externalities, such as natural resources, national parks, energy conversation initiatives, and environmental disasters (Portney, 1994). More recently, economists have started to apply these methods on the valuation of cultural goods, such as museums, national parks, and TV channels, such as the BBC in the UK, the Irish broadcast network in Ireland, the American PBS, and NHK, the Japanese broadcast company (Noonan, 2003a). It is important to note that economic valuations in general, and cost benefit-analysis in 2 Electronic copy available at: http://ssrn.com/abstract=1332786 particular, can be, and often are useful tools for the evaluation of social projects. Nevertheless, it is also undeniable that methodological assumptions and reductions on the multiple dimensions of the subject are often needed. This work does not attempt to solve this limitation, however it does contribute by estimating the social value of a television based, educational and cultural project, using a transparent, replicable, and refutable methodology. Among the main findings of the paper it is worth noting that the social value of Futura is at least 16 times greater than its operational cost; in addition it seems that the mobilization initiatives implement by the channel have a positive and statistically significant impact on the social value of the channel. In addition to this introduction, the current paper is organized in five section. The next section presents the contingent valuation method and the estimations used to estimate the willingness to pay for the channel. The third section presents the Futura project, characterizing in broad lines its mission and main programs and activities, as well as a brief summary of the literature of contingent valuation of other cultural goods. The forth section presents the field work, and a few key elements of the instrument used in this exercise. The fifth section presents the nonparametric and parametric estimation of the population willingness to pay for the channel The last and sixth section summarizes in a nutshell the main results of the paper. 2 Methodology The first published reference to contingent valuation method (CV) apparently occurred in 1947, when Ciriacy-Wantrup wrote about the benefits of preventing soil erosion, however it wasn’t until almost two decades latter that the methods began to be applied in academic research (Portney, 1994). In March 2008, a quick search at Google Scholar yield the reference of almost 21,000 studies that use CV to elicit consumers willingness-to-pay (WTP) for non-market goods1 . In recent years, increasing attention has been given to the statistical aspects of CV survey design and data analysis. The main reason has been the shift in CV practice from open-ended format to using a a closed-ended question, also known as the referendum model (Hanemann and Kanninen, 1999). The referendum format was introduced into CV by Bishop and Heberlein (1979), and since 1 Google Scholar search the terms: ”contingent valuation” and ”willingness to pay” (20th of March of 2008). 3 the mid-1980s gained widespread acceptance as the preferred way to cast the valuation questions, a position that has latter been endorsed by the NOAA Blue Ribbon panel on CV lead by two Nobel prizes laureates (Arrow et al., 1993). One of the main advantages of this format is that it appears to be incentive-compatible, as least in concept, and thus eliminates strategic behavior as a major concern with contingent valuation . The standard referendum model for CV can be interpreted as the difference of an indirect utility function. In this theoretical model, originally developed by Hanemann (1984) which utilizes the random utility function developed by Mcfadden (1973) and others, individuals have a utility function which can be expressed by V1 (y; s) if they accept a project, and V0 (y; s) with they reject it; where y is the respondent’s income, and s represents a vector of socioeconomic characteristics. When V is composed of a linear deterministic part and a random element, such as V (y; s) = αj (s) + βy + εj for j = 0, 1, and the coefficients have been estimated, then WTP is given by W T P = α(s)/β + ε/β where α(s) = α1 (s) − α0 (s) and ε = ε1 − ε0 . Hence the mean WTP can be expressed as E(W T P ) = α(s) β (1) When ε has a mean zero, symmetric distribution such as logistic or normal, the range of WTP is the real line. However, nothing guarantees that E(W T P ) is positive and in fact for particular values of s it is often negative. For most problems addressed by CV, negative WTP is simply wrong (Haab and Mcconnell, 1997). CV typically deals with public goods or public dimensions of private goods. For most public goods, negative WTP is not correct because the goods can simply be ignored, if it does not provide utility to the respondent. More often than not, negative estimates of WTP are a consequence os statistical fit and functional form, not true preferences. This section briefly presents two alternative ways, one parametric and one nonparametric, to estimate WTP in referendum-style contingent valuation when a substantial fraction of the respondents choose not to consume the good. All estimations were implemented in Stata and algorithms written are available from the author upon request. 4 2.1 Nonparametric Estimation When the pattern of response is well behaved, estimates of WTP will not be especially sensitive to the choice of distribution for the unobserved random component of preferences, or for the functional form of the preference function. However, there are many cases when the distribution or the functional form can have a substantial effect of the estimates of WTP. Because of this sensitivity of WTP for some CV studies, it is useful to sometimes implement a least restrictive approach to estimating WTP. 2.1.1 The Turnbull Estimator When samples are large and as the offered price increases, the probability of observed no responses to each bid should increase (Fj ≤ Fj+1 ). In other words, as the bid prices increase, we would expect the distribution function to monotonically converge to one for large sample sizes. In such cases a few authors have suggested to impose a monotonicity restriction on the distribution-free estimator. 2.1.2 The Estimation A procedure to calculate the Turnbull Distribution-Free estimator can be summarized in the following manner: 1. For bids indexed j = 1, ..., M , calculate Fj = Nj Nj +Yj , where Nj is the number of no responses to tj and Yj is the number of yes responses to the same bid, and Tj = Nj + Yj . 2. Beginning with j = 1, compare Fj and Fj+1 . 3. If Fj+1 > Fj then continue. 4. If Fj+1 ≤ Fj then pool cells j and i into one cell with boundaries (tj , tj+2 ], and calculate Fj∗ = Nj +Nj+1 Tj +Tj+1 = Nj∗ Tj∗ . That is, eliminate bid tj+1 and pool responses to bid tj+1 with responses to bid tj . 5. Continue until cells are pooled sufficiently to allow for a monotonically increasing CDF. ∗ 6. Set FM +1 = 1. ∗ . 7. Calculate the PDF as the step difference in the final CDF: fj∗ = Fj∗ − Fj−1 5 2.1.3 Lower Bounds for Mean and Median Willingness to Pay Once the distribution of the WTP has been estimated it is possible to estimate measures of central tendency for WTP. Because the Turnbull only uses the information contained in responses to provide estimates of the distribution function, and estimate of mean or median WTP will similarly only use the minimal amount of information (Haab and Mcconnell, 2002). Median Willingness to Pay The Turnbull proved an estimate of the range in which median WTP falls. Since the noresponse proportions are consistent estimates of the distribution point masses at each price, the price for which the distribution functions passes 0.5 is the lower bound on the range of median WTP. The next highest price represents the upper bound on the range of median WTP. In other words, the median represents the price for which the probability of a no response equals 0.5. Since the Turnbull only gives point mass estimates at a discrete number of points, the median can only defined within a range. A Lower Bound Estimate for Willingness to Pay ∗ ELB (W T P ) = M X ∗ tj (Fj+1 − Fj∗ ) (2) j=0 where Fj∗ = Nj∗ Tj∗ , ∗ F0∗ = 0, and FM +1 = 1. This lower bound estimate of WTP has a very intuitive interpretation. By multiplying each offered price by the probability of WTP falls between that price and the next highest price, we get a minimum estimate of WTP. The estimate proportion of the sample that has WTP failing between any two prices is assumes to have WTP equal to the lower of those two prices. This estimate is appealing because it offers a conservative lower bound on WTP for all non-negative distributions of WTP, independent of the true underlying distribution. Using a similar procedures, an upper bound on WTP can be define as ∗ EU B (W T P ) = M X ∗ t∗j+1 (Fj+1 − Fj∗ ) (3) j=0 The problem here, however, lies in the definition of tM ∗ +1 . Since tM ∗ is the highest offered bid it is necessary to define the upper bound on the range of WTP using an ad hoc method. Some authors suggest the use of the country income, but acknowledge that most values will be 6 arbitrary (Haab and Mcconnell, 2002). One advantage of the lower bound estimate os WTP is the distribution of the estimator. Since fj∗ are normal and tj are fixed, the ELB (W T P ) is also a normal. Hence the variance of the lower bound estimate is: ∗ V (ELB (W T P )) = M X Fj∗ (1 − Fj∗ ) Tj∗ j=1 (tj − tj−1 )2 ∗ = M X V (Fj∗ )(tj − tj−1 )2 (4) j=1 The variance can be used for constructing hypothesis tests and confidence intervals about ELB (W T P ). 2.2 2.2.1 Parametric Estimation The Spike Model Corner solutions arise in the theory of demand for private goods in a variety of circumstances. For example, goods may be mutually exclusive for logical or institutional reasons, they may be perfect substitutes, or the indifference curves may intersect the axis in some dimension (Kristrom, 1997). The Spike model proposed by Kristrom (1997), allows researchers to sort the respondents in two (or more) groups, without compromising the analysis. It is worth noting that the most popular parametric models to analyzes CV experiments are special cases of the spike model. In the Spike model it is assumed that the distribution function of WTP has the following form: 0 Fwtp (A) = p Gwtp (A) if A < 0 if A = 0 if A > 0 where p belongs to (0,1) and Gwtp (A) is a continuous and increasing function such that Gwtp (0) = 0 and lim → ∞Gwtp (A) = 1. Thus, there is a jump-discontinuity at zero. 7 Fwtp (A) = 0 if A = 0 2.2.2 if A < 0 if A > 0 The Estimation The spike model can be estimated with a variety of approaches (Kristrom, 1997), including parametric maximum likelihood methods. The spike model basically uses two valuation questions: one asks whether or not the individual would want to contribute at all to the project; the other suggests a price A. For each individual i, define and indicator that tells if the individual is in-the-market or not: Si = 1 if W T P > 0 (0 otherwise) (5) Similarly, let Ti indicate if the respondent is willing to pay the suggested price: Ti = 1 if W T P > A (0 otherwise) (6) The log-likehood for the sample is then given by l = N X Si Ti ln[1 − Fwtp (A)] + 1 + Si (1 − Ti )ln[Fwtp (A) − Fwtp (0)] + + (1 − Si )ln[Fwtp (0)]. (7) This likelihood was programmed using Stata ML function. 2.2.3 Calculating Mean and Median Willingness to Pay As the distribution of the WTP in the logistic model, the mean WTP for the spike model is given by the area under the survival function Z inf Z 0 1 − Fwtp (A)dA − E(W T P ) = Fwtp (A)dA − inf 0 8 (8) The latter part vanishes when negative WTP is not allowed, the the spike at zero has zero contribution. The equation 8 o calculate the mean WTP, one must solve the integral Z 0 inf exp(α − βA) 1 + exp(α − βA (9) The integral converges only if β > 0; the marginal utility of money must be positive in order the mean to exist in this model (Kristrom, 1997). If β is positive, mean WTP in this spike model is given by ln[1 + exp(α)]/β. The median is given by α/β if [1 + exp(−α)]−1 < 0.5 and 0 otherwise. A few authors note that the spike model appears to be ideally suited for cases when the distribution of WTP is asymmetric and when a sizeable fraction of the population has a zero WTP. 3 The contingent valuation of the Futura and other cultural goods 3.1 The Futura Channel Futura Channel is a communication technology for social change based on a television channel and methodologies for social mobilization. It aims to overcome the country’s territorial, infrastructure and access to knowledge obstacles in order to meet the need for information, education and entertainment of its population. It is based on television’s capacity to disseminate information in a large scale, of reaching communities with little or no access to cultural goods and of strengthening educational actions and resources. In addition to exhibiting its programs on a national level, Futura developed a technology for community mobilization to favor the appropriation and local use of its content by a growing number of institutions, people, projects and social movements. It also has been training a growing number of youths in audiovisual skills so they may become agents for social transformation and expression in their respective local communities. Futura is simultaneously a private, nonprofit, public interest, innovative technology for the production of educational television and for community mobilization. It has an annual budget 9 of R$ 35 million and it uses an operation model based on partnerships, network articulation and creative and cooperative production. Futura project is funded through partnerships with the Brazilian private sector2 , these institutions contribute with aproximately 30 million reais per year, the remaining 5 million reais are covered mostly by special project and program royalties. Futura is on the air 24 hours, every day, with a technological coverage that reaches 73 million Brazilians over different distribution environments, from the most orthodox, such as conventional parabolic antennae, cable TV and satellite dishes (NET and SKY systems) and network TV (generator and university TV networks), to the most heterodox such as passenger barges and trains. The Channel is regularly watched by 40 million people (most free of charge), and it articulates and mobilizes a significant social network: more than 13.300 institutions, including nongovernmental organizations, public schools, day care centers, hospitals, prisons, companies, social movements, universities, university TVs, national and international agencies develop different forms of relationship with Futura: they expand signal coverage, participate in program creation, appropriate and apply the programs in social projects according to local demands. Currently, 2 millions people use the Futura programming in their organizations. With 150 professionals and a little over one decade of existence, Futura has produced 83 thousand hours of programming, reached 40 million viewers and trained more than 425 thousand educators from social institutions. In 2006 alone, it carried out 250 hours of technology transfers in television processes with the exchange of knowledge and joint achievements responsible for 7.14% of Futura’s real production time. Besides that, it guaranteed free phone service help for operators of affiliated TVs for programming solutions. 3.2 Contingency Valuation of Other Cultural Goods The use contingent valuation to estimate consumer’s willingness to pay for cultural goods is becoming more popular, specially after the NOAA panel, for a comprehensive survey of the existing literature in the field please refer to Noonan (2003a) and Noonan (2003b). It is possible to find applications of this method in the cultural economics literature in areas as diverse as arts, heritage sites, theaters, museums, archeological sites, sports, and television. 2 The main funders are: Bayer Schering Pharma;CNI; CNN; CNT; FIESP; FIRJAN; Funda¸ca ˜o Bradesco; Funda¸ca ˜o Ita´ u Social; Funda¸ca ˜o Vale; Gerdau; SEBRAE; TV Globo; Votorantim. 10 In the field of arts, Thompson et al. (1983) evaluate what type of artistic manifestations the Australians were willing to pay. Morrison and West (1986) did the same in the province of Ontario, Canada, using a phone survey, and Thompson et al. (2002) adopted similar procedures, using a postal questionnaire survey, in the American state of Kentucky. In terms of the conservation of historical heritage sites, Pollicino and Maddison (2001) evaluate the potential benefit of the restoration of the British cathedral of Lincoln, while Grosclaude and Soguel (1994) did the same regarding the impact of pollution on historical sites. Another study by Chambers et al. (1998), looked at the preservation of an American historical site, namely the St. Genevieve academy, in the state of Missouri. There is also a growing number of studies which applied CV methods on television, either ´ or to specific programs or looking at the channel as a whole (i.e. BBC, NHK, PBS and RTE) TV shows. In 2004, the BBC during the process of reviewing its charter, released the result of a WTP exercise to measure its benefit. The study concluded that the annual value of the TV license of, at the time, 121 pounds, was inferior to the value that the respondents attributed to the channel, 240 pounds per year (BBC, 2004). Nakamura and Fontenay (2006) applied a similar method to estimate the WTP of the Japanese population for the NHK, as well as the willingness to accept for the same channel given its closure. Jennings and Curtis (2002) also evaluate the WTP for the Irish public television and concluded that the value attributed to the channel, 72,6 pounds, was lower than the TV license charged in that country. Bohm (1972), in a pioneer study, valuated the WTP for the Swedish comedy show. One important aspect of this exercise, was that immediately after the WTP exercise, the viewers were effectively requested to pay for the show, and the results of the WTP were very close to the actual final results. Finally, Finn et al. (2006) estimated the WTP of Canadians to an internet broadband service (which also streamed video). At the end, the study recommended that this project would only be of interest to the public if the final price was lower than 15 Canadian dollars. Table 1 summarizes the main characteristics of the studies using CV methods that looked at the public good television. 11 Table 1: Sum´ ario da Literatura sobre Aplica¸c˜oes no Campo da Televis˜ao: Pa´ıs Ano Canal Tipo de TV Valor atribu´ıdo Inglaterra Jap˜ ao 2004 2006 BBC NHK P´ ublica aberta P´ ublica aberta Su´ecia 1972 - Irlanda Canad´ a 2002 2006 ´ RTE - 20 libras por mˆes 16 d´ olares via antena e 11 via sat´elite (por mˆes) Dois resultados (grupos distintos): Entre 7,29 e 8,84 coroas suecas e entre 10,19 e 10,33 coroas suecas (por um determinado programa) 72,6 libras por mˆes Soma de todos os servi¸cos ofertados pela Internet banda larga inferior a 15 d´ olares canadenses por mˆes 3.3 Televis˜ ao fechada P´ ublica aberta Servi¸cos de banda larga (que incluem v´ıdeo) Specificities and challenges when applying contingent valuation methods to cultural goods The use of CV methods on the analysis of cultural goods has been growing in the past 15 years. However, there are a few specificities of the market of these goods that must be taken into account. First of all, there is often a significant difficulty on the definition of what are cultural goods, since the pay-back of these goods are often much more abstract. One possible consequence, is the creation of the hypothetical scenario which is presented to the respondent, as well as her or his understanding of it, which can jeopardize the credibility of the results. Moreover, cultural public goods sometimes sprang passionate or adverse reactions from the respondents (?). There are public culture goods which the WTP is even negative, given the strong adverse reaction that the respondent might have, or might be indifferent for the respondent. This possibilities have significant implications on the subsequent statistical analysis. One important additional point is that often cultural goods are not fully public, what poses substantial methodological problems in terms of the creation of the hypothetical market for this good. The economic value attributed by the population to a public good might not be sufficient to represent its real cultural value, and, in many situations might lead to the underestimation of the value of the cultural good being evaluated. 4 Data The current paper estimates the value of the Futura project considering both the social value of the Futura TV channel and the social mobilization activities. In order to do that a survey 12 representative of the Brazilian adult population was conducted, to collect face-to-face information. The main characteristics of the survey design were: i) men and women, with 16 or more years of age; ii) national coverage; and, iii) 2.511 face-to-face interviews. Regarding the instrument used, the field protocol was divided in two main moments. During the first part, the respondent was presented to a text explaining the Futura project, its financial needs, and how the program was planning to raise funds. Following this part, the responded was asked if she or he would be willing to pay a particular ‘bid’ value for the Futura project. Two different texts presenting the Futura project were used. The first, here after referred to as Model 1 (see Table 2), described both the TV Channel as well as its social activities and partnerships. The second, hereafter called Model 2 (see Table 3), present the Futura TV channel without explicitly mentioning its social mobilization activities. Table 2: Verbatim of the Futura presentation text - with social mobilization activities (Model 1) NOW I WOULD LIKE TO PRESENT YOU THE FUTURA CHANNEL.. (SHOW CARD WITH PRESENTATION TEXT AND READ IT TOGETHER) Futura is a TV channel with cultural and educational programs broadcasted 24 hours per day. The access to its scheduling can be made through parabolic antennas, UHF and VHF, University TV, and cable TVs (NET, Sky or Directv). There is not type of cost associated with the current access to its programs. In additional to its differentiated content, the channel promotes and support social mobilization activities in public schools, jails, hospitals, health centers, libraries, day care center, among several other social entities. This activities are related to the channel programs, including the joint production of programs in partnership with those institution, capacity building workshops, and other activities. Currently, Futura is financed solely by private companies, however Futura does not air product advertisements. It only acknowledge the companies and institutions which support and finance its activities. There is also no government participation and public funds on the maintenance of the channel or of the social services provided by it. Hence, it is very important for Futura to obtain funding so it can continue its broadcast and social mobilization activities. For this reason, we the societies contribution. This contributions can be done through a special agreement with the electricity company of your locality. This companies can charge a monthly fee on your electricity bill only of those individuals who choose to support Futura. Source: Questionnaire Futura’s Contingency Valuation/2007. The objective of using these two distinct hypothetical markets is to enable the evaluation of the value attributed to the social mobilization activities implemented by the Futura channel. Often, contingency valuations face difficulties when dealing with multiple aspects of a single public good. There are a number of reports in the literature which show that consumers/respondents are often not able disentangle the effects of multiple goods, which sometimes a bundled together. 13 Table 3: Verbatim of the Futura presentation text - without mobilization activities (Model 2) NOW I WOULD LIKE TO PRESENT YOU THE FUTURA CHANNEL.. (SHOW CARD WITH PRESENTATION TEXT AND READ IT TOGETHER) Futura is a TV channel with cultural and educational programs broadcasted 24 hours per day. The access to its scheduling can be made through parabolic antennas, UHF and VHF, University TV, and cable TVs (NET, Sky or Directv). There is not type of cost associated with the current access to its programs. Currently, Futura is financed solely by private companies, however Futura does not air product advertisements. It only acknowledge the companies and institutions which support and finance its activities. There is also no government participation and public funds on the maintenance of the channel or of the social services provided by it. Hence, it is very important for Futura to obtain funding so it can continue its broadcast. For this reason, we the societies contribution. This contributions can be done through a special agreement with the electricity company of your locality. This companies can charge a monthly fee on your electricity bill only of those individuals who choose to support Futura. Source: Questionnaire Futura’s Contingency Valuation/2007. Nevertheless, the characteristics of Futura argued for the relevance of this comparison, with a particular emphasis on the difference between the channel content and the social mobilization activities and special training activities promoted by the Project. As it will be seen in the next section, the results suggest that in the case of Futura, consumers were able to separate the two values. After the project presentation, the second part of the interview consisted in a specific referendum question, presented in Table 4, as well as several socio-demographic questions about the respondent and their household, a detailed module on TV habits, and a number of debriefing questions, in order to identify the motivations for accepting or not the offer. Table 4: Referendum question regarding the willingness to pay for the Futura project P10. (GO TO P8=2) Are you willing to pay R$ financially support the Futura? per month, charged on your eletricity bill, to Source: Questionnaire Futura’s Contingency Valuation/2007. As it can be seen, the methodological choice of this study was to apply the referendum model. As it was presented in section 2, in this design each respondent is faced with a single price (‘bid’ value), based on which she or he have to express their willingness to pay. In what follows several debriefing questions were made in order to shed light on the motivations and circumstances under which the respondent accepted or not to pay the offered value. The use of 14 Table 5: Desenho do Experimento modulo bid 1 2 5 8 10 15 20 30 Total 1 No. 2 No. Total No. 248 251 250 250 251 248 252 253 2003 64 64 65 63 62 64 63 63 508 312 315 315 313 313 312 315 316 2511 Fonte: Futura/DataFolha 2007 . Popula¸ca ˜o: Universo. the referendum model is justifiable on the grounds that it is the only survey format which allows the replication, in the closest possible manner, of the market conditions, in which consumers are always faced with a single price. It was precisely this reason that made the referendum approach the recommended option of the NOAA panel (Arrow et al., 1993). Table 5 shows the final design of the experiment and the distribution of the interviews. In total 16 types of instruments were created, given the interaction of the two basic criteria of the exercise: 8 different bid values; and two different presentations to the channel (with and without the social mobilization activities and training programs). As it can be seen, 2003 respondents were faced with Model 1, while 500 respondents answered Model 2. In both cases the number of respondents which face each bid value was evenly balanced. In order to assure the internal validity of the results, it is important to assure that field work was implemented correctly and that the questionnaires application was properly balanced. In other words, it important to check the quality of the randomization of the ‘bid’ values. Eventual biases in this process my lead to distortions on the results, which could jeopardize the credibility of the exercise. In order to test the quality of the randomization of the ‘bid’ values a multivariate approach was used. The hypothesis being tested is that if the instrument application was truly random, all characteristics of the respondents should be completely orthogonal to the ‘bid’ value. In order to test for this, a number of observable characteristics of the respondents, such as gender, age, region, income, among other were included in a multinomial logit model and ordered probit 15 model of the ‘bid’ values. Ideally, this exercise should suggest that these observable characteristic should not be correlated with the dependent variable. Table 6: Resumo do Modelo Logit Multinomial e Probit Ordenado: Determinantes do Valor Oferecido Vari´ avel chi2 pval1 N ratio pval Logit Multinomial 181.827 .8545914 2265.000 Probit Ordenado 26.831 .5808075 2265.000 164.204 0.672 Nota: DataFolha/Futura, 2007. Table 6 the summary statistics of the fitting of both model, multinomial logit and ordered probit. As it can be noted, both models presented a very poor fitting, with a joint significance test failing to reject the hypothesis that model is different from zero, p-values of 0.85 and 0.58, respectively. These results suggest that the field work was correctly implemented, assuring the internal validity of exercise. 5 Empirical Results 5.1 Descriptive statistics This section presents and discusses the main findings of the parametric and nonparametric estimations of the WTP for the Futura Channel. In all cases, two subpopulation will often be used. First, the total number of households which have declared a willingness to pay greater than zero, and second, the total number of viewers, who have also declared a willingness to pay greater than zero. This groups represent approximately a total of 13 and 5 million households, respectively. Table 7 illustrate the WTP for all households, including and not including the social mobilization activities of the channel. Given the qualitatively similarities between the two groups of results, the analysis in this section will focus on the results that consider the Futura and its social mobilization activities (Model 1). One important result of this analysis is the fact that at least 42% of all households are 16 Table 7: Propensao a Pagar Pelo Canal Futura bid 1 2 5 8 10 15 20 30 Disposto a pagar o valor oferecido (R$) com A¸c˜ ao Social (M´ odulo 1) Pagam % 42.4 37.2 22.6 20.9 18.0 9.7 10.7 5.2 N˜ ao Pagam % 57.6 62.8 77.4 79.1 82.0 90.3 89.3 94.8 sem A¸c˜ ao Social (M´ odulo 2) 1 2 5 8 10 15 20 30 41.3 35.7 15.1 18.9 19.7 9.0 8.7 10.1 58.7 64.3 84.9 81.1 80.3 91.0 91.3 89.9 com A¸c˜ ao Social (M´ odulo 1) apenas telespectadores 1 2 5 8 10 15 20 30 57.7 54.3 27.7 43.1 24.9 13.4 19.4 8.6 42.3 45.7 72.3 56.9 75.1 86.6 80.6 91.4 Fonte: Futura/DataFolha 2007 . Popula¸ca ˜o: Modulo 1 e 2. willing to pay at least R$ 1,00 per month for the channel. It is also clear that the proportion of households willing to pay decreases as the ‘bid’ value offered increases, indicating coherence with was is expected given the economic theory regarding the demand behaviors of common good3 . Table 7 also presents the WTP for Futura, considering only those households that have the habit of watching the program4 . The WTP of viewers is slightly higher 57,7% of the respondents are willing to pay at least R$ 1,00. These finding is also consistent with what is expect, since the suggests that those respondents that already consume this good, are more willing to pay for it. Figure 1 presents the Kernel-weighted local polynomial regression the same data used in 3 4 A good is considered ‘common’ if their demand presents a negative relationship with the price variation. 11,5 million households or 33 million of persons. 17 Table 7 to generate a continuum representation of the WTP. A couple of important results derive from the analysis of this very basic set of descriptive statistics. First of all, it becomes evident that the there are a few of inversion of the consumer’s preferences. In other words, the demand curve for Futura does not decrease monotonically. This aspect is inconsistent with economic theory, as it implies that in a few cases the demand for this good increases as the prices increases. It must be beared in mind this inversions are well within the bounds of the 95% confidence interval (gray area of Figure 1), and might be mostly driven by measurement errors due to the sample sizes used. Another important results is the fact a large share of the population would not be in the ‘market’ for Futura. In other words, they are indifferent towards the channel. There are two possible explanations for this. The first one related to the choice of the bid values, in particular the lowest ones. It might be the case that the lowest bid value was to big for the good being offered. While this is a possibility, which ultimately can only be tested empirically, the pre-tests of the instrument do not seem to support the idea that a lower initial bid value would dramatically increase the WTP. The second possible explanation, is related to the nature of the good been offered, and is not surprising in the context of cultural goods. ? point out that cultural goods sometimes generate passionate reactions from the respondents, which might lead to the polarization of the respondents, of those individuals that can not perceive themselves in the market of this good, and other that can. One important analytical implication of this truncation is that the traditional econometric methods will fail to adequately to fit this data, since they all assumes that there is no truncation, as a result the application of these methods will result in a negative WTP, which is inconsistent with the economic theory. In order to cope with this data characteristics, the literature on WTP was development alternative estimation procedures, such as the use of the nonparametric Turnbull estimation or the parametric Spike model, both described in detail in section 2. This section now turns to the presentation of the results using these two models. 5.2 Nonparametric estimation of the willingness to pay: Turnbull This subsection presents the results using the Turnbull estimation algorithm suggested by Haab and Mcconnell (2002). This estimation procedure generates the minimum average WTP for the 18 Figure 1: Estima¸c˜ ao N˜ ao Param´etrica da Fun¸c˜ao de Densidade de Probabilidade: propor¸c˜ao a pagar pelo Canal Futura com a¸c˜ ao social Futura channel. Table 8 shows the the main values, as well as their respective p-values. The minimum WTP for the total population is R$ 3,75 per month, considering the channel social mobilization activities (Model 1). The average minimum WTP without considering the social mobilization activities (Model 2) is substantially lower, R$ 2,23 per month. An additional analysis presented in Table 8 narrows to view to only those households who watch Futura regularly. As it can be seen in this subgroup the minimum WTP increases to R$ 5,56 and R$ 4,54 per month, for those face with and without the social mobilization activities, respectively. One important finding are the consistently higher values, when the social mobilization activities of the channel are included, suggesting that these activities, indeed add value to the channel. Moreover, all the results were statistically significant at one percent. 19 Table 8: Estima¸c˜ ao N˜ao Param´etrica da Disposi¸c˜ao a Pagar: Estimador de Turnbull Valor M´ınimo Pval Total Modelo 1 3.754 0.000 Total Modelo 2 2.233 0.000 Telespectadores Modelo 1 5.559 0.000 Telespectadores Modelo 2 4.544 0.000 Once the average minimum WTP is produced it is possible to estimate the aggregated social value based on a conservative estimation of the WTP. In order to do that two populations were considered. The first one looks to all households which are willing to pay a positive value for the channel (13,653 millions). The second one looks at a subgroup in which only those households with at least one viewer of the channel is included (5,018 millions). Table 9 presents the main results of this aggregation. Considering the total number of households with a positive WTP, the total social value of the channel is estimated in 615 millions reais per year; with the subgroup of households, with number drops to 340 millions per year. All the results are statistically significant at 1%. Table 9: Valor Total da Disposi¸c˜ao a Pagar M´ınima em R$ (Estimador de Turnbull) Total Telespectadores Popula¸c˜ ao 13,653,753.000 5,018,342.000 Modelo l 615,078,016.587 334,761,872.038 Modelo 2 365,862,973.812 273,615,254.703 Valor da A¸c˜ ao Social 249,215,042.776 61,146,617.335 40.518 18.266 Propor¸c˜ ao % 20 Table 10: Propensao a Pagar Pelo Canal Futura com Acao Social Disposto a pagar algum valor Disposto a pagar o valor oferecido (M´ odulo 1) Nao Sim Total Nao Sim Total 88.6 0.0 70.1 % linha 11.4 100.0 29.9 100.0 100.0 100.0 Disposto a pagar o valor oferecido (M´ odulo 2) Nao Sim Total Nao 88.5 0.0 71.0 Sim 11.5 100.0 29.0 Total 100.0 100.0 100.0 Assiste ou Conhece o Canal Futura (espontˆ aneo e estimulado) Nao Sim Total Nao 75.4 58.0 70.3 Sim 24.6 42.0 29.7 Total 100.0 100.0 100.0 Assiste o Canal Futura (espontˆ aneo e estimulado) Nao Sim Total Nao Sim Total 74.9 56.6 70.3 25.1 43.4 29.7 100.0 100.0 100.0 Fonte: Futura/DataFolha 2007 . Popula¸ca ˜o: Universo. 5.3 Parametric estimation of the willingness to pay: Spike As it has been noted, corner solutions arise in the theory of demand for private goods in a variety of circumstances. The Spike model proposed by Kristrom (1997), allows researchers to sort the respondents in two (or more) groups, without compromising the analysis. In order to show the robustness of the results, this paper also apply this method to estimate of the willingness to pay. Table 10 suggests that 11% of the 36 million households which responded that were not willing to pay the offered ‘bid’ value, also said on the debriefing questions that would be willing to pay a lower price than the one offered for the channel. In addition, Table 10 also shows that household with viewers are more willing to pay something for the channel, 42% e 43,4%, in the case of Model 1 and Model 2, respectively. As it was discussed in section 2 this parametric estimation is able to generate both the mean and median WTP for the channel, and not just the minimum WTP as in the case of the Turnbull estimator. Tables 11 and 12 present the estimations of the average WTP for the channel. Most results suggest that for a total of households, the WTP is R$5.93 per month in the 21 case of Model 1, and R$5.72 in the case of Model 2. Considering only the households with viewers, the value became even greater, reaching R$9.58 per month for the channel, with social mobilization activities (see Table 11). It is important to note that the WTP estimations without the social mobilization activities were omitted since the sample sizes is not sufficient to estimate the parameters. As it can be seen in Table 12, the total social value attributable to the channel for the population as a whole is approximately R$1 billion per year, and R$ 577 millions for the viewers. One important additional exercise is the comparison of the social value of Futura with the with amount of funds received by its donor institutions as well as the total operational costs of the channel. According to the latest external auditor’s report, in 2007, Futura received, approximately, R$ 30 millions from donors, and had a total operational cost of R$ 35 millions (Ernst & Young, 2008). Hence, it is possible to notice that the total social value of Futura (attributed by the total number of households) is 28 times higher than its operational cost. In other words, for each real spent with the operational cost of the program, it generates o social return of R$ 28 . If we consider only the households with viewers, lowers than to R$ 16. In terms of the donor funds, the returns are even greater, for each real invested, the social return is R$ 21 and R$37, depending on the population considered, total or only viewers, respectively. Finally, it is worth noting the R$ 34 million reais can be attributed to the social mobilization activities promoted by the channel - defined as the difference between the WTP, with and without the social mobilization using the total population. It worth noting that this difference is substantially lower than the one estimated in the previous section (Turnbull - R$ 249 millions). One possible explanation for this difference is simply the effect that minimum average value and the average value reflect different things, with the Module 1 and Module 2 values converging as they move from the minimum towards the average. An alternative explanation might be the simply the differences in the sample sizes between the two modules, which can create greater measurement errors on the estimation of minimum average values of Module 2, forcing the pooling of groups that actually were not to be pooled, and subsequently biasing downwards its WTP. In any case, for conservative reasons, our prefered comparison in this exercise is the later, since it uses the Spike model, which reports the average WTP and not the minimum average. 22 Table 11: Disposi¸c˜ao a Pagar M´edia em R$ Probit Probit ln bid Probit ln bid Spike Total Modelo 1 -6.256 19.457 0.759 5.936 Total Modelo 2 -11.450 30.285 0.550 5.727 Telespectadores Modelo 1 1.238 32.355 2.053 9.586 Telespectadores Modelo 2 -8.354 - 0.423 - Table 12: Valor Total da Disposi¸c˜ao a Pagar M´edia em R$ (Spike) Total Telespectadores Popula¸c˜ ao 13,653,753.000 5,018,342.000 DAP com A¸c˜ ao Social 972,503,264.828 577,260,452.986 DAP sem A¸c˜ ao Social 938,418,302.581 - Valor da A¸c˜ ao Social 34,084,962.248 - 3.505 - Propor¸c˜ ao % 6 Final considerations The Futura TV channel is a communication project, ran by a non-profit foundation financed exclusively by private sector donations in Brazil. With almost 10 years of existence, Futura broadcast reaches almost 50% of the Brazilian population, and has approximately 33 million viewers in the whole country. The main objective of this paper is to measures the social value of this initiative and compere it with its operational costs. Contingency valuation (CV) techniques using the referendum design were used to estimate the willingness to pay of a representative sample of the Brazilian adult population, with approximately 2,500 face-to-face interviews. 23 Using both parametric and non-parametric techniques this paper is able to estimate the population willingness to pay for Futura, even in a sample were from 40% to 60% of the population choose not to consume the good. The analysis suggest that the willingness to pay for the channel is positive and at least 16 times greater than the channel operation cost. The average minimum willingness to pay estimated using the Turnbull nonparametric model was R$ 3.7 per month, while the parametric average willingness to pay estimated with the Spike model suggested a value of R$ 5.9. When multiplied by the 5 million households with at least one member willing to pay something for the Futura TV channel, the total annual social value of the channel ranges from 330 to 550 million reais. Moreover, the complementary social mobilization activities promoted by the Futura channel seems to have a positive impact on its social value, suggesting that the method was successful in terms of disentangling embedded effects. References Arrow, K., R. Solow, U. S. N. Oceanic, and A. Administration (1993). Report Of The Noaa Panel On Contingent Valuation. BBC (2004, October). Measuring the value of the bbc. Technical report, BBC and Human Capital. Bishop, R. and T. Heberlein (1979). 1979: Measuring Values Of Extra-Market Goods: Are Indirect Measures Biased?,”. American Journal Of Agricultural Economics 6, 926–30. Bohm, P. (1972). Estimating Demand For Public Goods: An Experiment. European Economic Review 3 (2), 111–130. Chambers, C., P. Chambers, and J. Whitehead (1998). Contingent Valuation Of Quasi-Public Goods: Validity, Reliability, And Application To Valuing A Historic Site. Public Finance Review 26 (2), 137–154. Ernst & Young (2008). Demonstra¸c˜ oes financeiras auditadas do Canal Futura, 31 de dezembro de 2007 e 2006. Technical report, Funda¸c˜ao Roberto Marinho. 24 Finn, A., S. Mcfadyen, and D. Thomas (2006). Use And Non-Use Values Of Super-Net Enabled Broadband Content And Services. Thassociation Of Cultural Economics International Conference, Vienna, 6–9. Grosclaude, P. and N. Soguel (1994). Valuing Damage To Historic Buildings Using A Contingent Market: A Case Study Of Road Traffic Externalities. Journal Of Environmental Planning And Management 37 (3), 279–287. Haab, T. and K. Mcconnell (1997). Referendum Models And Negative Willingness To Pay: Alternative Solutions. Journal Of Environmental Economics And Management 32 (2), 251– 270. Haab, T. and K. Mcconnell (2002). Valuing Environmental And Natural Resources: The Econometrics Of Non-Market Valuation. Edward Elgar Publishing. Hanemann, W. (1984). Welfare Evaluations In Contingent Valuation Experiments With Discrete Responses. American Journal Of Agricultural Economics 66 (3), 332–341. Hanemann, W. M. and B. Kanninen (1999). The Statistical Analsyis Of Discrete-Response Cv Data. Oxford University Press, Usa. Jennings, M. and J. Curtis (2002). Assessing The Public Service Element Of Irish Public Broadcasting–The Validity Of The Contingent Valuation Method. The Contingent Valuation Of Culture Conference, Cultural Policy Centre, Chicago, February. Kristrom, B. (1997). Spike Models In Contingent Valuation. American Journal Of Agricultural Economics 79 (3), 1013–1023. Mcfadden, D. (1973). Conditional Logit Analysis Of Qualitative Choice Behavior. Institute Of Urban And Regional Development, University Of California. Morrison, W. and E. West (1986). Subsidies For The Performing Arts: Evidence On Voter Preference. Journal Of Behavioral Economics 15 (Fall), 57–72. Nakamura, K. and A. B. D. Fontenay (2006). A critical analysis of public service broadcasting in a digital environment -its changing role in japan from the international comparative viewpoint. Technical report. 25 Noonan, D. (2003a). Contingent Valuation And Cultural Resources: A Meta-Analytic Review Of The Literature. Journal Of Cultural Economics 27 (3), 159–176. Noonan, D. (2003b). Contingent Valuation Studies In The Arts And Culture: An Annotated Bibliography. Irving B. Harris Graduate School Of Public Policy Studies, University Of Chicago. Pollicino, M. and D. Maddison (2001). Valuing The Benefits Of Cleaning Lincoln Cathedral. Journal Of Cultural Economics 25 (2), 131–148. Portney, P. (1994). The Contingent Valuation Debate: Why Economists Should Care. The Journal Of Economic Perspectives 8 (4), 3–17. Sen, A. (1999). Development as Freedom. Oxford University Press. Thompson, B., C. Throsby, and G. Withers (1983). Measuring Community Benefits From The Arts. Macquarie University, School Of Economic And Financial Studies. Thompson, E., M. Berger, G. Blomquist, and S. Allen (2002). Valuing The Arts: A Contingent Valuation Approach. Journal Of Cultural Economics 26 (2), 87–113. 26
© Copyright 2024