Consumer Preferences for Digital Cable Broadcasting Service in Korea: A Choice Experiment Study

Abstract

A digital cable broadcasting service is a multimedia broadcasting service that provides high definition and various supplementary services by using digital transmission. Korea implemented a complete digital broadcasting service by 2012. This study applied a choice experiment to investigate consumer preferences, and it calculated the marginal willingness to pay for this service. Moreover, we employed a multinomial probit model to relax the assumption that all respondents have the same preference for attributes being valued. The results suggest that respondents value channels, definition, video-on-demand (VOD) service, pay-per-view (PPV) service, and commerce based on TV (T-commerce). On the other hand, online gaming may be less important as an attribute for digital cable broadcasting service in Korea.

I. Introduction

A digital cable broadcasting service is a multimedia broadcasting service that provides high definition and various supplementary services by using digital transmission. The interest in digital broadcasting service had increased around the world. The United States completed its conversion to digital broadcasting service by June 2009. Japan and Europe phased out their analog broadcasting services by 2011 and 2012, respectively. Korea implemented a complete digital broadcasting service by 2012.

There are several reasons why countries considered the transition to digital broadcasting as an important national policy. First of all, digital signaling offers much better picture definition and sound quality. The digital screen and sound quality are known to be five to seven times better than those offered by analog transmission. Digital broadcasting also makes it possible to experience various multimedia services with communication features that are transmitted through a digital format. With digital transmission, subscribers can experience an interactive two-way communication. Subscribers can provide input for their needs through television (TV), and can enjoy Internet surfing, shopping and gaming, all while watching TV.

From the perspective of the national economy, a digital broadcasting service stimulates investment and development of the next generation broadcasting technology, leading to economic growth. Moreover, the government would be able to manage the limited bandwidth more efficiently, freeing up space for future broadcasting and communication services.

Figure 1 shows the trend in the number of digital cable TV subscribers and the proportion of digital TV subscribers out of total cable TV subscribers in Korea. Both of them have dramatically increased. By the end of 2015, the number of digital cable TV subscribers reached 7.6 million households and the proportion of digital TV subscribers out of total cable TV subscribers in Korea had reached 52.8%.

Figure 1The number of digital cable television subscribers

A number of studies on broadcasting services have been conducted in Korea. Joo (2004) investigated users’ attitude on the adoption factors of digital TV with regard to new media technologies. The results show no correlation between users’ demographic characteristics and adoption factors, and respondents considered price as an important adoption factor. Kang et al. (2005) analyzed the economic value of next-generation broadcasting, and showed that the economic values of Wireless Broadband (WiBro), Internet-based Wideband Code Division Multiple Access (WCDMA), Voice over Internet Protocol (VoIP), and Digital Multimedia Broadcasting (DMB) with WiBro were similar.

Koh et al. (2006) examined, through a conjoint survey, the types of attributes that critically affect competition among new convergence broadcasting services and how competitive they will be. The results of the study indicate that price is the most important factor for consumers, and it is expected that a considerable consumer valuation exists for the unique advantages of new convergence broadcasting services like mobility and two-way interactivity, which will add competitiveness of those services in the future. Choi et al. (2008) applied a Bayesian mixed logit model to conjoint survey data collected prior to the launch of Korea’s two current types of mobile TV service (the satellite digital multimedia broadcasting and the terrestrial digital multimedia broadcasting) to investigate consumer preferences regarding the core attributes of mobile TV. The results imply that consumers regard subscription rates and media quality as the two most important attributes of mobile TV services.

This study applies a choice experiment (CE) to investigate consumer preferences for digital cable broadcasting services in Korea. The remainder of this paper is organized as follows. Section 2 reviews CE and its methodological issues. Section 3 presents statistical models to derive the willingness to pay (WTP) for digital cable broadcasting service. Section 4 discusses the results. The final section concludes.

 

II. Choice Experiment

The CE approach offers a promising opportunity to measure the various economic values of a digital cable broadcasting service, since it is concerned with modeling choices varying over a range of characteristics, rather than with the estimation of WTP for a single option. We identified six attributes: channels, definition, video-on-demand (VOD) service, pay-per-view (PPV) service, commerce based on TV (T-commerce), and online game. Table 1 describes these attributes, including price attribute, and shows how each level of attributes was defined.

Table 1Descriptions and levels of chosen attributes

CEs involve the use of the statistical design theory to construct choice sets which can yield coefficient estimates that are not confounded by other factors. In this study, we employed the orthogonal main effects design1 by using the SPSS 12.0 package. In the CE questions, there were three alternatives. Two represented a digital cable broadcasting service featuring combinations of attribute levels and specific price levels. The third alternative represented the status quo. Subsequently, there were 32 × 210 × 42 possible combinations of attributes and levels to form choice sets. Since it was impractical to ask respondents to choose from all the combinations, we drew a subset of all choice sets for estimating coefficients, and we drew eight choice sets. They were then randomly divided into two sets of four choices each. Figure 2 presents the example of the choice set that was actually used.

Figure 2A sample choice set used in this study

To analyze the data, this study employs a multinomial probit (MNP) model. The previous discrete choice models have mostly used the multinomial logit (MNL) model, because MNL model has an advantage in that choice probability has a simple closed form. However, it has a serious drawback, known as the independence from irrelevant alternatives (IIA) property. To rectify this problem, other advanced approaches can be applied, such as the generalized extreme value model (Train, 1986), the heteroskedastic extreme value model (Bhat, 1995), and the MNP model. Among these, the MNP model is most appealing because it allows the greatest flexible pattern of error correlation structure. This approach allowed the authors to relax the assumption that all respondents have the same preference for attributes being valued, which is usually required in empirical CE studies. The ability to model unobservable heterogeneous preferences in a population is a clear advantage of this technique (Layton, 2000).

 

III. Model

1. Multinomial Probit Model

In this study, the random utility model was used to explain individual choices by specifying functions for the utility derived from the available alternatives. The indirect utility function for each respondent i who chooses alternative j in the choice set can be expressed as:

The indirect utility function Uij can be decomposed into the deterministic part Vij, which is typically specified as a function of the attributes Xij in alternative j chosen by respondent i, and the stochastic part eij, which represents the unobservable influence on individual choices. The MNP model assumes that the errors eij’s are distributed as multi-variate normal with mean vector zero and covariance matrix defined as follows:

Consider the probability that the first alternative will be chosen.

where

η21 = ei2 − ei1, η31 = ei3 − ei1, V12 = Vi1 − Vi2, and V13 = Vi1 − Vi3.

f(η21, η31) has a bi-variate normal distribution with covariance matrix Ω1, which is transformed from Ω. We can identify the first diagonal element of the covariance matrix to one to add a property that a scale shift would not change observed choices. Ω1 is given by:

where and .

The probability that second and third alternatives will be chosen can be similarly derived (Maddala, 1983).

2. Utility Function and Marginal WTP

The utility function of the model, with the exception of the error term, can be expressed as a linear function of an attribute vector (Z1, Z2, Z3, Z4, Z5, Z6, Z7) = (Channels, Definition, VOD, PPV, T-commerce, Online game, and Price). We include two alternative-specific constants (ASCs), which represent dummies for respondent’s choice of alternatives A and B in the choice set. ASC captures the utility of an alternative that attributes fail to capture (Adamowicz et al., 1994).

where β’s are parameters to be estimated for each attribute that influences respondents’ utility. If we are calculating a marginal WTP (MWTP) from the status quo level of each attribute and assume that all other variables remain constant, we can obtain the following MWTP estimates by totally differentiating equation (5) and omitting i for brevity.

The MWTPs of each attribute represent the marginal rate of substitution between the price and each attribute.

 

IV. The Results

A total of 800 face-to-face interviews were conducted over a month beginning December 2008. The survey was carried out in seven large cities: Busan, Incheon, Daegu, Daejeon, Gwangju, Ulsan, and Seoul. Since sampling and fieldwork were done by interview experts from a professional polling firm, Dongseo Research, Inc., the response rate is almost 100%. After missing and inconsistent answers to valuation questions were deleted, 731 (91.0%) individuals were found to be suitable for further examinations for this study, resulting in a total of 2,924 observations.

The sample demographics are similar to those of the target areas. Male respondents account for half the total sample. The average age and education level of respondents were 43.18 and 13.09 years, respectively. The sample was, thus, representative.

The estimation results of the MNP model are presented in Table 2. The model was estimated by the GHK simulation2 (Geweke, 1989; Keane, 1994; Hajivassiliou and McFadden, 1998). Using the Wald statistic, estimated equations were found to be statistically different from zero at the 1% level.

Table 2Notes: a The variables are defined in Table 1. b ASCA and ASCB mean alternative-specific constants, which represent dummies for a respondent’s choice of the alternatives A and B, respectively. c * and ** indicate statistical significance at the 5% and 1% levels, respectively.

Judging from the statistical results, the coefficients of all attributes, except online game, were significant, and their signs were consistent with our expectation. For instance, the coefficients of channels, definition, VOD, PPV, and T-commerce were positive. As the level of these attributes increases, the utility to respondents also increases. On the contrary, the coefficient of price attribute was negative and statistically significant. The higher the price, the lower the respondents’ utility. However, the coefficient of online game was not statistically significant (i.e., this does not indicate that respondents place a significant value on this attribute).

To test whether the IIA property holds, a likelihood ratio test was carried out. The likelihood ratio test statistic follows asymptotically a chi-square distribution with one degree of freedom under the null hypothesis that errors are uncorrelated. The result of the IIA test rejected the null hypothesis that the errors are uncorrelated at the 1% significance level. Therefore, we cannot apply the MNL model which assumes that choices satisfy the IIA assumption. Since the MNP model does not require the IIA assumption, it is clear that the MNP model is a better model to estimate the data of this study.

The results of MWTP estimates are shown in Table 3. The respondents are willing to pay KRW 112 (USD 0.09) per one channel, and KRW 12,734 (USD 10.19) to change standard definition (SD) to full high definition (full HD). The results suggest that, if VOD, PPV, and T-commerce are available, respondents would be willing to pay KRW 11,301 (USD 9.04), KRW 6,753 (USD 5.40) and KRW 4,165 (USD 3.33), respectively.

Table 3Notes: a USD 1 equals approximately 1,250 Korean won. b The t-values were computed using the delta method. * and ** indicate statistical significance at the 5% and 1% levels, respectively.

To this end, the Monte Carlo simulation technique of Krinsky and Robb (1986) was used to generate 95% confidence intervals for MWTP. The 95% confidence intervals are presented in Table 3.

One of the strengths of CE is that the estimated coefficients of attributes may be used to estimate WTP for various scenarios in which the attributes can be combined. In this paper, three different digital cable broadcasting services were considered. Table 4 presents three hypothetical scenarios and the results of households’ WTPs.

Table 4Scenarios of digital cable broadcasting service (Korean Won)

With other conditions held constant, households’ monthly WTPs for a digital cable broadcasting service such as scenario A, B, and C were approximately KRW 32,873 (USD 26.30), KRW 44,932 (USD 35.95) and KRW 57,263 (USD 45.81), respectively. Furthermore, WTPs for all households can be calculated by multiplying households’ WTPs by the number of households. Since the survey of this study was conducted on households whose ages are between 20 and 65 in Busan, Incheon, Daegu, Daejeon, Gwangju, Ulsan, and Seoul, the authors used the number of households in those age brackets in the seven large cities. According to the Korea National Statistical Office, there were 6,471,247 households where ages ranged from 20 to 65 in 2005. The last two rows of Table 4 show the total WTP for households under each scenario. For example, households in the seven cities are willing to pay for KRW 4,447 billion (USD 3,557 million) for digital cable broadcasting service, which had the highest levels presented in this survey: 200 channels, full HD, VOD, PPV, and T-commerce. In similar fashion, we can make other scenarios using the attributes of digital cable broadcasting service.

 

V.Concluding Remarks

In this study, CEs were used to measure the potential value of each attribute in the Korean digital cable broadcasting service, and the MNP model was employed to relax the assumption of IIA. Overall, the survey was successful in eliciting MWTP values for a digital cable broadcasting service. Respondents’ choices for selecting preferred alternatives were found to be within their own abilities, and the MWTP estimates were statistically different from zero, except for online game attribute. The results suggest that respondents value channels, definition, VOD, PPV, and T-commerce. On the other hand, online game may be less important as an attribute for the digital cable broadcasting service in Korea.

The monthly MWTP for one channel was calculated as KRW 112 (USD 0.09). The monthly MWTP for definition was calculated as KRW 12,734 (USD 10.19), and the household monthly WTP for VOD and PPV were KRW11,301 (USD 9.04) and KRW 6,753 (USD 5.40), respectively. In the case of T-commerce, MWTP was computed as KRW4,165 (USD 3.33). Households from seven cities in Korea are willing to pay approximately KRW 4,447 billion (USD 3,557 million) for a digital cable broadcasting service, which had the highest levels presented in this survey.

This study demonstrated the feasibility of extending the use of CE to digital cable broadcasting services. The results of this study provide a useful framework for incorporating such quantitative information in the evaluation of various policies with regard to digital cable broadcasting services.