Quality, technology, experience and the use of media resources in distance learning programs by two-year community colleges.

ABSTRACT

In spite of the increase in the number of Distance Learning Programs (DLP) offered by higher education institutions, not all programs have been successful. Successful programs use different types of media resources for instructional delivery. An understanding of the factors affecting decisions related to the type and number of teaching media resources used in successful DLP could provide valuable information not only to those two-year colleges currently offering DLP but also to those planning to offer them in the future.

Unfortunately, the majority of the research efforts done in the past focused on DLP in four-year colleges and universities and not on two-year community colleges. Information on the key factors affecting these decisions from the two-year college perspective could help them in budgeting and planning new or enhanced distance learning programs, make an efficient allocation of resources and also give hints on how to improve the competitiveness of the college in a rapidly growing industry.

Limited Dependent Variable models were used in this study to analyze quality, technology and experience as factors affecting these decisions made by two-year colleges. It was found that the set of statistically significant factors affecting the decision to use a specific type of media used is not the same for each type of media. It could also be noted that these factors affect differently the decision to use a given number of teaching media resources.

INTRODUCTION

In the past decades, we have experienced rapid demographic, socioeconomic and lifestyle changes. Examples include more participation of women in the labor market, additional two income families, declining birth rates, increased number of one-person households, more women in executive positions, higher life expectancy, and higher standards of living.

All these changes, in one way or another, have increased the importance of the 'nontraditional" student (full-time employed, more mature, not able to attend regular classroom classes, with family responsibilities, goal-oriented) within the college student population. This increasing number of non-traditional college students has increased the demand for "non-traditional" educational programs, among them Distance Learning Programs (DLP). Distance learning provides access to many more students than just offering higher education courses in the traditional classroom manner (Yee, 1998; Perreault et al., 2000). Distance learning encompasses many different types of teaching media like Internet-based courses, the use of satellites, interactive television (ITV), teleconferences, one-way broadcasting, electronic bulletin boards, fax machines, cable television, toll-free telephone numbers, etc. (Au and Chong, 1993; Ball and Crook, 1997; Brown and Duguid, 1998; Hall, 1990; Kubala, 1998; Luna and McKenzie, 1987; Merisotis and Phipps, 1999; Opitz, 1996; Swift et al., 1997; Teleg, 1996). In 2002, 85% of 2-year colleges and 84% of 4-year colleges offered distance education courses up from 58% and 62% respectively in 1998 (Wirt et al., 2004).

In spite of the increase in the number of DLP offered by higher education institutions, not all programs have been successful. Actually, there are more examples of failures than successes (Arkansas Department of Higher Education, 2004). Among other important characteristics, successful programs use different types of media resources for instructional delivery but concentrate in the use of just a few of them (Waits and Lewis, 2003). An understanding of the factors affecting these decisions could provide valuable information not only to those institutions currently offering DLP but also to those planning to offer them in the future.

Successful DLP are offered not only by four-year colleges but also by two-year colleges (Williams, 2003). Unfortunately, the majority of the research efforts done in the past focused on DLP in four-year, masters, and doctoral programs offered by four-year colleges and universities. Not too much research has been done for two-year community colleges (Husson and Waterman, 2002; Anderson, 2003; Jorgenson, 2003; Lorenzetti, 2003; Nair, 2003; Jorgenson, 2004). With the present literature focused upon four-year colleges and universities, it is important to have studies analyzing the factors affecting decisions related to the type and number of teaching media resources used in successful DLP from the two-year college perspective. Information on these key factors will help two-year colleges to initiate or enhance their programs (Carnevale and Olsen, 2003; Horne, 1994); improve their production strategies; make better use of resources; and become more competitive in a rapidly growing market.

The authors were unable to identify studies on the determinants of the type and number of media used by two-year colleges in successful DLP. This study attempted to correct this shortcoming. This research used Limited Dependent Variable techniques (univariate probit and ordered probit regression) to model the factors affecting the decision to use a specific type of media and the ones affecting the decision to use a specific number of media resources in DLP by two-year community colleges, paying special attention to the role of quality, experience and technology in these decisions. It could be found that the set of statistically significant factors affecting the decision to use a specific type of media used is not the same for each type of media. It could also be noted that these factors affect differently the decision to use a given number of teaching media resources.

CONCEPTUAL AND EMPIRICAL FRAMEWORK

Two-year community colleges make decisions about the adoption of a specific type of teaching media in a world of uncertainty. In deciding to adopt a specific type of media, two-year colleges compare the expected benefits and expected costs related to that decision. The two-year community college decides to use a specific type of media if the expected benefits exceed expected costs.

Formally, the difference between expected benefits and costs from using the ith teaching type of media resource is treated as an unobservable variable [y.sup.*] such that

[y.sup.*] = [alpha]x [epsilon] (1)

We do not observe the latent variable [y.sup.*], but we observe the outcome of the adoption decision, which is a dummy variable y such that y =1 (the community college uses the ith type of teaching media resource) if [y.sup.*] > 0 and y = 0 (the community college does not use the ith type of teaching media resource) otherwise. Also, x are vectors of independent variables affecting the decision to use the ith type of teaching media; [alpha] are vectors of unknown parameters; and [epsilon] are vectors of additive disturbance terms randomly and normally distributed with mean zero and variance one. Univariate probit regression analysis was used to analyze the factors determining the type of media used.

The statistical approach used to determine the factors affecting the number of teaching media used is ordered probit regression analysis. This type of analysis can be used to estimate the relationship between a dependent ordinal variable and a group of independent variables. In this case, the dependent variable is the number of media used in DLP. Formally, the model is expressed as follows

[y.sup.*] = [beta]x + [delta] (2)

[y.sup.*] defines a latent unobservable continuous variable-expected net benefit of using a given number of teaching media resources; x is a vector of independent variables affecting the decision to use a given number of teaching media resources; [beta] is a vector of unknown parameters; and [delta] is a vector of additive disturbance terms randomly and normally distributed with mean zero and variance one. Recoding the number of media used 1, 2, 3, 4, 5 and 6 for 0, 1, 2, 3, 4 and 5, what we observe is a discrete ordinal variable y such that

y = 0 if [y.sup.*] [less than or equal to] 0 y = 1 if 0 < [y.sup.*] [less than or equal to] [[mu].sub.1] y = 2 if [[mu].sub.1] < [y.sup.*] [less than or equal to] [[mu].sub.2] y = 3 if [[mu].sub.2] < [y.sup.*] [less than or equal to] [[mu].sub.3] y = 4 if [[mu].sub.3] < [y.sup.*] [less than or equal to] [[mu].sub.4] y = 5 if [[mu].sub.4] < [y.sup.*]

where the [mu]'s unknown parameters to be estimated with [beta]. The set of independent variables affecting these decisions include the quality of the program, technology, experience and an interaction term between quality and experience.

DATA SET

A survey questionnaire was prepared and sent to program administrators of DLP at two-year colleges. The survey tool asked questions on the individual's opinion as to factors, which they believed to be essential to the success of their distance learning programs (on a scale from 1 to 5). The success factors were identified in the literature. Specifically, administrators were asked the following question: "In your college's Distance Education Program, please rate the importance of each of the following criteria critical to success of the program (with 5 as extremely important, 4 very important, 3 important, 2 somewhat important, and 1 as not important).

1. Quality of course/program

2. Adequate faculty compensation

3. Quality supplemental material

4. Technology working effectively

5. Updated technology

6. Appropriate course offerings

7. Faculty training."

In addition, questions were included as to program years, whether the distance learning programs included liberal arts programs as well as business or computer programs or if only a few distance learning courses could be taken.

The questionnaires were sent to all 250 two-year colleges listed in the Peterson's Guide to Distance Learning Programs. Colleges with distance learning programs must meet certain criteria in order to be listed in the Peterson's Guide to Distance Learning Programs. They must "... have full accreditation or candidate-for-accreditation (pre-accreditation) status granted by an institutional or specialized accrediting body recognized by the U.S. Department of Education or the Council for Higher Education Accreditation" (Peterson's, 2003). This list excludes those colleges that didn't meet the criteria for publication as well as those colleges that initiated their DLP after publication of the 2002 edition of the guide.

A pilot study was done with a separate, like population to identify any problems with the instrument such as ambiguous wording or questions. This pilot study attempted to obtain a more accurate measure through the use of the survey tool. The pilot study was also used in an attempt to reduce systematic error (Van Auken and Barry, 1997), which deals with unanticipated problems that might occur with the survey questions (Kier et al., 1998).

Of the 250 questionnaires mailed, 180 were returned. However, some of these surveys were not included in the final sample due to incomplete (or not reported) responses. The final sample included 104 observations resulting in a 42 percent usable questionnaire rate.

VARIABLES

Dependent Variables

The dependent variable in the univariate probit regressions was dichotomous and indicates the use (or not) of the ith type of teaching media resource (i = correspondence, tutorials, ITV, Internet, satellite, others). The dependent variable in the ordered probit regression was ordinal and denotes the number of teaching media resources used in DLP by two-year colleges (1, 2, 3, 4, 5, 6).

Independent Variables

Crosby (1996) indicated that the quality of course programs is essential to the success of DLP. Huston (1997) noted that students' perception of the quality of the course pertained directly to adequate faculty training. If a faculty member did not receive appropriate training on the distance learning technology, the faculty member's class evaluations might be extremely low even if the faculty member knew her or his subject well and had taught that particular class numerous times before (Eddy and Spaulding, 1996; Reinig et al., 1998). More recently, Husson and Waterman (2002) developed some specific quality measures for distance learning, among them: appropriate course content and design, faculty training, and technical and academic support for students in online courses. They also noted that technology was essential when developing distance-learning programs. Lately, Jorgenson (2003) pointed out the pillars of quality that could help educators assess and improve online courses and programs. They mentioned, among others, learning effectiveness, access and faculty satisfaction.

Au and Chong, (1993) and Husson and Waterman, (2002) indicated that technology working effectively was very important in the success of DLP. Technology can often be found in the literature interconnected with the faculty training. Eddy and Spaulding (1996) noted that students' satisfaction with DLP was, among other things, due to adequate faculty training in the technology used and also that the technology was updated and working effectively. Recently, Perreault et al., (2000) identified the reliability, support for, and the use of technology as the main problems to successful delivery of distance learning courses and recommended training as the most obvious solution for them.

Considering these findings from the existing literature, we constructed a "quality" variable based on the administrators' evaluation of the following success factors: quality of course or program, adequate faculty compensation, quality supplemental materials, appropriate course offerings and faculty training. The respondent was asked to rate the importance of each of these factors on a scale from 1 to 5. The quality variable was then calculated as the sum of the respondent's evaluation rates of each of these factors. We also constructed a "technology" variable based on the respondent's evaluation (on a scale from 1 to 5) of the following factors: technology working effectively and updated technology. The technology variable was also calculated as the sum of the respondent's rates of each of these two factors. Similar approaches for constructing composed variables are commonly used in health economics (Kenkel, 1990, 1991; Nayga, 2000, 2001).

The number of years offering DLP was used as a proxy for experience. Also, it is reasonable to expect that different quality DLP respond to experience in different ways and that less and more established programs respond differently to quality changes. Then, an interaction term (quality x experience) was also included in the regression equations. The same set of independent variables was used both in the univariate and ordered probit regressions.

Definitions of the dependent and independent variables included in the models are presented in Table 1. Table 2 contains the sample statistics for the continuous and discrete variables.

RESULTS AND DISCUSSION

Univariate Probit Analysis of the Factors Determining the Type of Media used in Distance Learning Programs.

We used univariate probit regressions to analyze the major determinants of the choice of the type of media used by 2-year community colleges in DLP. Regression coefficients and their t-values are reported in table 3. The marginal effects of changes in the regressors on the probabilities of using different types of media resources are reported in Table 4. In general, the models fit the data well. The percentage of correct predictions was 60% or better. In general, the univariate probit regressions showed that quality, experience and technology were statistically significant at the 10% or 5% levels respectively for the different types of media used equations. Moreover, the set of statistically significant factors in the different equations are not the same.

Quality has a negative and significant influence on the probability of using correspondence and satellite resources. A 1% increase in quality decreases the probability of using correspondence and satellite resources by 5.4% and 4.5% respectively. Also, technology has a positive and significant influence on the use of ITV and satellite resources. A 1% increase in the use of technology increases the probability of using them by 8.0% and 8.6% respectively. It seems that two-year colleges with quality, technology based DLP tend to concentrate on the use of ITV and satellite resources (and differentiate their products) probably because of their competitive advantages and the presence of economies of scale and potential profit opportunities related to a rapidly growing monopolistic competitive market.

More established two-year colleges are less likely to use correspondence and Internet resources than are less established ones. In addition the interaction term was positive and statistically significant in the correspondence and Internet equations, indicating that the negative marginal impact of experience on the probability of using correspondence and Internet resources is smaller for more quality programs than for less quality ones (and that the negative marginal effect of quality on these probabilities is smaller for more established programs).

Ordered Probit Analysis of the Factors Determining the Number of Media used in Distance Learning Programs.

We used ordered probit regressions to analyze the major determinants of the number of teaching media resources used by two-year community colleges in DLP. Regression coefficients and their t-values are reported in Table 5. The marginal effects of changes in the regressors on the probabilities of using different numbers of media resources are reported in Table 6. (1)

The estimated coefficients of the unknown parameters [[mu].sub.1], [[mu].sub.2], [[mu].sub.3], and [[mu].sub.4] are positive, increasing and statistically different from zero. This implies that the ordered probit regression of the number of media equation is justified. The value of the likelihood ratio test statistic was statistically significant at the 1% level therefore the null hypothesis that all slopes in the regression are zero was rejected. In general, the analysis showed that most variables are statistically significant at 10% or 5% levels suggesting that these variables are important in determining the number of media used in by two-year community colleges.

The quality of the program has a negative effect on the probability of using a large number of media. The number of media used is more likely to be small the higher the quality of the program. A 10% increase in the quality of the program decreases the probability of using 3 or 4 media resources by about 2% and decreases the probability of using 5 or 6 media resources by 1.5 and 1% respectively. As mentioned before, these results may indicate that community colleges with good quality DLP tend to specialize or concentrate in the use of just a few media teaching resources (ITV, satellite and Internet) probably to take advantage of the economies of scale associated with a fast-growing market for distance education. Technology was not statistically significant.

Community colleges with more established DLP are less likely to use a large number of media than less established colleges. The empirical results demonstrated that more established colleges are 24%, 20%, 19% and 22% less likely to use 3, 4, 5 and 6 media resources than less established colleges. (2) More established programs may want to take advantage of their greater efficiency and competitive advantages to concentrate in the use of some specific types of media resources, which in turn will allow them to specialize in the production of a few specific differentiated products. The interaction term was positive and statistically significant at 5%, indicating that the negative marginal impact of quality on the number of media used is smaller for more established than for less established DLP (and also that the negative marginal effect of experience on the number of media used is smaller for higher quality DLP).

CONCLUDING REMARKS

We used limited dependent variable techniques (univariate and ordered probit analysis) to analyze the factors affecting the choice and the number of teaching media resources by two-year colleges in DLPs.

The empirical evidence demonstrated that quality, technology and experience could be important determinants of the choice and the number of media used. Moreover, the set of statistically significant factors affecting the decision to use a specific type of media will not be the same for each type of media. The results also indicated that these factors affect differently the decision to use a given number of teaching media resources. Some important findings are that experience and quality negatively affects the probability of using a large number of media resources. Also, technology can be a positive influence on the probability of using ITV and satellite resources, but it does not affect the probability of using a large number of media resources. The results of this study could help two-year colleges in budgeting and planning new or enhanced distance learning programs, make an efficient allocation of resources and also give hints on how to improve the competitiveness of the college in a rapidly growing industry. This study represents a first step at analyzing DLPs offered by two-year community colleges. Further studies may consider including other additional variables in the analysis like demographic information, financial information, factor prices and cost information. Future studies could also use demographic variables and prices of higher education to estimate the demand for DLP and study the substitutability and complementarily between different higher education products.

REFERENCES

Anderson, A. (2003). Online courses: A learner-centered approach. Community College Week, 16, 6.

Arkansas Department of Higher Education (2004). Status of distance education in Arkansas. report on the status and progress of distance learning in Arkansas. Available at www.arkansashighered.com

Au, M., and C. Y. Chong. (1993). The evaluation of the effectiveness of various distance learning methods. International Journal of Instructional Media, 20, 105-127.

Ball, J., and B. Crook, B. (1997). Managing change through distance learning. Community College Journal of Research and Practice 21, 13-22.

Brown, J. S., and P. Duguid. (1998). Universities in the digital age. Change 4, 11-20.

Carnevale, D., and F. Olsen. (2003). How to succeed in distance education. Chronicle of Higher Education, 40, A31-A34.

Crosby, P. (1996) Quality if Free: The Art of Making Quality Certain, McGraw-Hill: New York.

Eddy, J. P., and D. Spaulding. (1996). Internet, computers, distance education and people failure: Research on technology. Education 116, 391-394.

Greene, W. (1997). Econometric Analysis. New York: Prentice Hall.

Hall, W. (1990). Open Universities: Closing the distances to learning. Change 4, 44-51.

Horne, M.C. (1994) As they struggle to survive, junior colleges learn to adapt. Crisis, 101, 19-23.

Husson, W.J., and E. K. Waterman. (2002). Quality measures in distance learning. Higher Education in Europe 27, 253-261.

Huston, J. L. (1997). Factors of success for adult learners in an interactive compressed video distance learning environment. Dissertation Abstracts International, 58(04), 1A, (University Microfilms No. AAT 97-29317).

Jorgenson, H. (2003). Evaluate and improve distance programs with Sloan-C's five pillars of quality. Distance Education Report 7, 1-3.

Jorgenson, H. (2004). New quality-Enhancement plan ensures program improvement. Distance Education Report 8:5.

Kenkel, D. (1990). Consumer health information and the demand for medical care. The Review of Economics and Statistics 72, 587-95.

Kenkel, D. (1991). Health behavior, health knowledge, and schooling. The Journal of Political Economy, 99, 287-305.

Kier, F. J., J. G. Melancon and B. Thompson. (1998). Reliability and validity of scores on the personal preferences self-description questionnaire (PPSDQ). Educational and Psychological Measurement, 58, 612-623.

Kubala, T. (1998). Addressing student needs: Teaching on the Internet. T H E Journal 25, 71-74.

Lorenzetti, J.P. (2003). Critical ingredient for success: Presidential involvement. Distance Education Report 7, 22-30.

Luna, C. J., and J. McKenzie. (1997). Testing multimedia in the community college classroom." T H E Journal 24, 78-82.

Massey, W. F., and R. Zemsky. (1997). Teaching with television: One university's model. International Journal of Instructional Media 24, 221-227.

Merisotis, J. P., and R. A. Phipps. (1999). What's the difference. Change 31, 12-18.

Nayga, R. (2000). Schooling, health knowledge and obesity. Applied Economics. 32, 815-822.

Nayga, R. (2001). Effect of Schooling on obesity: Is health knowledge a moderating factor. Education Economics. 9, 129-137.

Nair, P. (2003). The changing shape of learning. New England's Journal of Higher Education and Economic Development, 17, 21-24.

Opitz, M. C. (1996). Interactive distance learning. Implications for the classroom teacher." Clearing House, 69, 325-326.

Perreault, H., L. Waldman and M. A. J. Zhao. (2000). Overcoming barriers to successful delivery of distance-learning courses. Journal of Education for Business 77, 313-18.

Peterson's Guide to Distance Learning Programs. (3rd ed.). (2003) Princeton, NJ: Peterson's.

Reinig, B. A., R. O. Briggs and J. F. Nunamaker. (1998). Flaming in the electronic classroom. Journal of Management Information Systems, 14, 45-60.

Swift, C. O., J. W. Wilson and J. W. Wayland. (1997). Interactive distance education in business: Is the new technology right for you? Journal of Education for Business 73, 85-90.

Teleg, R. W. (1996). Instructional design considerations for teaching international audiences via satellite. International Journal of Instructional Media, 23, 209-218.

Van Auken, S., and T. E. Barry. (1997). Observations: Toward the internal validation of cognitive age measures in advertising research. Journal of Advertising Research, 33, 82-85.

Waits, T., and L. Lewis. (2003). Distance Education at Degree-Granting Postsecondary Institutions: 2000-2001. U.S. Department of Education, National Center of Education Statistics (NCES 2003-17). Washington, DC.

Williams, P.E. (2003). Roles and competencies for distance education programs in higher education institutions. Journal of Distance Education, 17, 45-58.

Wirt, J., Choy, S., Rooney, P., Provasnik, S., Sen, A., and R. Tobin (2004). The condition of education 2004. U.S. Department of Education, National Center for Education Statistics (NCES 2004-077). Washington, D.C.: U.S. Government Printing Office.

Yee, J. A. (1998). Distance learning and community colleges. Community College Journal of Research and Practice, 22, 563-568.

ENDNOTES

(1) The marginal effects of the regressors x on the probabilities are not equal to the coefficients (Greene, 1997). For the six probabilities of using different numbers of media resources, the marginal effects of changes in the regressors are

[partial derivative] Prob(y = 0) / [partial derivative] x = 0 - [phi] (-[beta]'x) [beta] [partial derivative] Prob(y = 1) / [partial derivative] x = [[phi] (-[beta]'x) - [phi] ([[mu].sub.1] - [beta]'x)] [beta] [partial derivative] Prob(y = 2) / [partial derivative] x = [[phi] ([[mu.].sub.1] - [beta]'x) - [phi] ([[mu].sub.2] - [beta]'x)] [beta] [partial derivative] Prob(y = 3) / [partial derivative] x = [[phi] ([[mu].sub.2] - [beta]'x)] - [phi] ([[mu].sub.3] - [beta]'x)] [beta]) [partial derivative] Prob(y = 4) / [partial derivative] x = [[phi] ([[mu].sub.3] - [beta]'x) - [phi] ([[mu].sub.4] - [beta]'x)] [beta] [partial derivative] Prob(y = 5) / [partial derivative] x = [phi] ([[mu].sub.4] - [beta]'x) [beta] - 0 (4)

where [phi] (.) is the univariate standard normal probability density function.

(2) The approach described before to calculate the marginal effects on the probabilities is not appropriate for evaluating the effect of a dummy variable. We can analyze the effects of a dummy variable by comparing the probabilities that result when the variable takes its two different values with those that occur with the other variables held at their sample means (Greene, 1997). The marginal effects of the dummy variable moreten on the probability of using different number of media resources were calculated in the following way:

Variable      P(y=0)   P(y=1)   P(y=2)   P(y=3)   P(y=4)   P(y=5)

(moreten=0)    .004     .103     .278     .209     .188     .218
(moreten=1)    .652     .312     .034     .003     .000     .000
change         .647     .208    -.245    -.206    -.187    -.218

Justo Manrique, University of Houston-Downtown

Linda Bressler, University of Houston-Downtown

Table 1: Names and Description of Variables

Variable    Description

Dependent Variables

corresp     Binary variable; use correspondence (yes = 1, no = 0)
tutor       Binary variable; use tutorials (yes = 1, no = 0)
ITV         Binary variable; use DLP (yes = 1, no = 0)
Internet    Binary variable; use Internet (yes = 1, no = 0)
satellite   Binary variable; use Satellite (yes = 1, no = 0)
others      Binary variable; use other media resources
              (yes = 1, no = 0)
Num         Discrete ordinal variable; number of media used in DLP
            (1, 2, 3, 4, 5, 6)

Continuous Independent Variables

Quality     Quality index
Techno      Technology index
Interact    Interaction term: quality index multiplied by Number
              of years offering DLPs.

Binary      Independent Variable (yes = 1, no = 0)

Moreten     Two-year college is offering DLPs for more than 10 years.

Table 2: Descriptive Statistics

Variable    Mean    Std. Deviation

Dependent Variables

corresp       .23        .42
tutor         .24        .43
DLP           .73        .45
Internet      .89        .31
satellite     .17        .38
others        .55        .50
Num          2.82        1.20

Continuous Independent Variables

Quality     21.49        2.68
Techno       8.56        1.30
Interact     7.93       10.49

Discrete Independent Variable (yes = 1, no = 0)

Moreten       .38        .49

Table 3: Univariate Probit Estimates of type of media used in DLP

Variable          Correspondence   Tutorial     ITV

Constant              3.213         -.854       -.396
                     (1.72) **     (-.43)      (-.23)

Moreten              -4.076         -.405       -.052
                    (-1.80) **     (-.17)      (-.92)

Quality              -0.185         -.033       -.058
                    (-1.99) ***    (-.34)      (-.67)

Interact              0.203         0.06        0.025
                     (1.92) **      (.55)       (.24)

Techno                -.013          .056        .246
                     (-.10)         (.38)      (1.80) **
No. of                  104           104         104
observations

Log-Likelihood       -53.13        -52.18      -57.23

% of correct             76            76          75
predictions (a)

Variable          Internet       Satellite     Other

Constant             3.779          -.259         1.414
                    (1.46) *       (-.12)         (.85)

Moreten             -6.225          1.539         -2.45
                   (-2.04) ***      (.60)       (-1.16)

Quality              -.187          -.207          .009
                   (-1.50)        (-1.90) **      (.11)

Interact             0.299         -0.042         0.123
                    (2.09) ***     (-.35)         -1.26

Techno                .185           .394         -.179
                    (1.22)         (2.16) ***   (-1.40)
No. of                 104            104           104
observations

Log-Likelihood      -31.08         -40.48        -69.48

% of correct          89.4           81.7          60.6
predictions (a)

Asymptotic t-ratios are given in parentheses.

**, *** statistically significant at 10% and 5% levels respectively.
(a) An observations is judged to be 1 if the predicted probability
P(y=1) is 0.5 or larger otherwise the observation is judged to be zero.

Table 4: Marginal effects of changes in regressors on the
probabilities of using different types of media used in DLP

Variable   Correspondence   Tutorial    ITV

Moreten        -1.198        -.120     -.017
Quality         -.054        -.010     -.019
Interact         .060         .018      .008
Techno          -.004         .017      .080

Variable   Internet   Satellite    Other

Moreten     -.959        .334      -.970
Quality     -.029       -.045       .004
Interact     .046       -.009       .049
Techno       .028        .086      -.071

Table 5: Ordered probit regression estimates of the number of media
used in DLP

Variables              Estimate   t-ratio

Constant                  3.245     1.49
Moreten                  -3.025    -1.76 **
Quality                  -0.157    -2.20 ***
Interact                  0.173     2.19 ***
Techno                    0.163     1.43
[[mu].sub.1]              1.398     6.45 ***
[[mu].sub.2]              2.347     9.48 ***
[[mu].sub.3]              2.872    10.36 ***
[[mu].sub.4]              3.418     9.90 ***
No. observations            104
Log-Likelihood          -147.45
Likelihood ratio (a)      17.08

**, *** statistically significant at 10% and 5% respectively.
(a) The Likelihood ratio test statistic is computed as: -2 log
L = -2(log Lrestricted--log Lunrestricted). This is a valid test
statistic for the hypothesis that all slopes on the nonconstant
regressors are zero (significance level .002).

Table 6: Marginal Effects of changes in the regressors on the number
of media used in DLP Number of media used in DLP

Variables     1        2        3        4        5        6

Moreten      .6474    .2084   -.2448   -.2064   -.1871   -.2175
Quality      .0204    .0420   -.0186   -.0194   -.0145   -.0100
interact    -.0225   -.4630    .0205    .0213    .0160    .0110
techno      -.0211   -.0436    .0193    .0201    .0150    .0103