patrick paper by patmaz

VIEWS: 9 PAGES: 20

									DETERMINANTS OF THE MULTINOMIAL CHOICE FOR THE ENVIRONMENTAL
ATTRIBUTE IMPROVEMENT IN THE NATIONAL PARKS IN MALAWI. APPLICATION
OF MULTINOMIAL LOGIT CHOICE MODEL TO THEORETICAL DATA



                PATRICK MAZICK



                ENVIRONMENTAL ECONOMICS SERIES




                  JANUARY, 2012




UNIVERSITY OF MALAWI, BUNDA COLLEGE, DEPARTMENT OF AGRICULTURAL AND APPLIED ECONOMICS, LILONGWE, MALAWI



Page 1 of 20
THE AUTHOR

Patrick Mazick is a Resource economist, who is currently undertaking his research for the Collaborative Masters in
Agricultural and Applied Economics (CMAAE) and expected to graduate in September, 2012.

A part from school, Mr Mazick is business person; he produces potatoes, beans, supplies ground nuts, rice and runs salon
and cosmetics shop but enjoys writing papers. He may be contacted at:

University of Malawi

Bunda College

Department of Agricultural and Applied Economics

Box 219

Lilongwe

Malawi

Mobile: (00265) 999 365 952

pat_mazick@rocketmail.com or mazickpatrick@gmail.com




Page 2 of 20
TABLE OF CONTENTS



TABLE OF CONTENTS ................................................................................................................ 3
ABSTRACT .................................................................................................................................... 4
1.0 INTRODUCTION .................................................................................................................... 5
2.0 LITERATURE REVIEW ......................................................................................................... 7
3.0 METHODOLOGY ................................................................................................................... 8
3.1 THE EMPIRICAL MODEL ..................................................................................................... 8
4.0 RESULTS AND DISCUSSION ............................................................................................. 10
   Table 4.1 BASE CATEGORY 3 ............................................................................................... 10
   Table 4.2 base category 1 .......................................................................................................... 11
   Table 4.3 base category 2 .......................................................................................................... 12
   Table 4.4: base category 4 ......................................................................................................... 12
      Table 4.5 marginal effects base category 1 ........................................................................... 13
      Table 4.6 1V3 ........................................................................................................................ 14
      Table 4.7 1V4 ........................................................................................................................ 14
      Table 4.8 1V2 ........................................................................................................................ 15
      Table 4.9 2V3 ........................................................................................................................ 15
      Table 5.0 2V4 ........................................................................................................................ 16
      Table 5.1 3V4 ........................................................................................................................ 16
   Table 5.2 relative risk ratio ....................................................................................................... 17
5.0 CONCLUSIONS AND RECOMMENDATION ................................................................... 18
6.0 REFERENCES ....................................................................................................................... 19




Page 3 of 20
ABSTRACT
In this paper, we are demonstrating how individuals would choose their preference of the
environmental attribute given their socioeconomic characteristics and we can conclude that level
of awareness of an individual has positive contribution to somebody‟s choice of an
environmental attribute improvement. This is so not only In theoretical applications of the model
but also in real life situations, if an individual has knowledge or information about something,
the decisions that are made are well informed unlike somebody who is not informed of some
attributes about the park. No wonder, it is only information/knowledge about the park that
apparently influences the probability of opting for an environmental attribute improvement not
others because, even if you are educated, well advanced in years and have a lot of income but
you are not informed about a particular situation, your choice can not be rational. So, it makes a
lot of sense to see that these other three regressors (income, education and age) are insignificant
statistically even though to extent may help to make decisions. Due to this finding, this paper
recommends that through awareness campaigns, trainings, outreaches inform the community
members be about the park( what it offers, how to participate, charges) and ensure that all the
categories of individuals poor, rich, educated or not be informed to belong to the park




Page 4 of 20
1.0 INTRODUCTION
For several decades, social scientists have undertaken research on the motivations of individuals
who engage in pro-environmental behavior. In other words, when we see an individual trying or
wanting to promote environmental improvement, what factors are behind that person‟s
motivation? Gaining a detailed understanding of why individuals undertake in environmental
improvement options is important for policy makers and researchers seeking solutions to
environmental problems that require behavioral change. Many research efforts thus far, however,
tend to polarize around predominant themes in specific disciplines, (Moore et al, 2002)

Economists, for example, examine the influence of external circumstances, such as income,
price, and socio-economic characteristics, on human behavior. This implies that when as
economists, we observe a consumer in a market, how she/he behaves in that particular market is
hypothesized to be influenced by several factors of which income, prices and so forth are some
of them. Our approach is based on the assumptions of neoclassical economic theory, which
presupposes that individual decisions are based on a specific definition of rational self-interest.
Solutions to environmental problems that reward, penalize, or regulate behavior result from this
mode of analysis. Psychologists, on the other hand, concentrate on linking internal or
psychological variables to behavior. Their literature suggests that pro-environmental behavior
originates from values, beliefs, and attitudes that orient individuals toward particular actions.
Consequently, psychologists recognize awareness, education, guilt, and persuasion as tools for
invoking behavioral change.

In trying to bring conformity between economists and psychologists, we need to identify the link
between the economists‟ external influences and the psychologists‟ internal influences. We know
that psychological variables like values, beliefs, morals, obedience, are based on individuals‟
background of religion, culture and exposure. The two disciplines are just interdependent in the
sense that the increase in knowledge is done through education (informal or formal) which
affects our religious knowledge, cultural values and so on. But to have education, we need
income for us to access better education. Therefore, the understanding of these factors between
economists and psychologists are just the same, the difference comes when it comes to
measurements. Since economists like working with figures, we just focus on those factors that
are observable and can be objectively measured to make scientific conclusions. To reconcile with
psychological variables, the explanations of those figures are linked to some immeasurable
issues like those ones suggested by psychologists.

Despite the dominant role psychologists attribute to internal factors for motivating pro-
environmental behavior, a handful of researchers identify the need to formulate an

Page 5 of 20
interdisciplinary perspective. Van Liere and Dunlap (1980) argue that researchers should pay
equal attention to cognitive variables and demographic determinants that underlie environmental
concern. They assert that ‘‘the most powerful analyses of the social bases of environmental
concern will likely be those which consider both its demographic and cognitive determinants’’.

Considering pro-environmental behavior in particular, Guagnano, Stern, and Dietz (1995) argue
that „„science and policy require a socioeconomic theory of behavior that incorporates both
external conditions and internal processes‟‟. They note that reluctance of applied researchers to
merge insights from economics and psychology has led to narrowly defined policies that often
fall short of objectives. In this paper, we are going to consider both sides of the coin; economic
and psychological considerations, so that we can have concrete understanding of why do people
choose the environmental improvement options they choose. Hypothesized Internal variables
according to psychology consist of a newly developed scale for altruistic attitudes and external
variables according to economic theory consist of household income and standard socio-
demographic characteristics.

In this paper, we will focus on only participants in the park and analyze their specific
socioeconomic attributes for choosing the environmental attribute improvement. From theory we
know that individuals go for the environmental attribute improvement for different motives.
These include motives relating to several concerns: ecosystem health, personal health,
environmental quality for residents and even to upgrade the recreational value of the park.

The main hallmark of this paper is examine the determinants of the peoples‟ choice for the
environmental attribute improvement (water quality in the lake in the park, preservation of
elephants, introduction of the white rhinos, building of a guest house near the park). After asking
different respondents to the questionnaires, individuals indicated their desires in wanting these
four different environmental improvement attributes done.




Page 6 of 20
2.0 LITERATURE REVIEW
Despite research attempts devoted to theory building with regards to the factors influencing
firms‟ environmental behaviors (e.g. Alvarez Gil et al., 2001; Andersson and Bateman, 2000;
Henriques and Sadorsky, 1996, 1999; Hoffman, 1999; Moon and de Leon, 2007; Rivera, 2002;
2004; Sharma, 2000) to date theories are contested and empirical findings are inconclusive. This
conclusion comes after considering the fact that the framework within which firms or individuals
make decisions they make about environmental management is difficult to be easily disclosed
without their honest openness. The choice of firms on their contribution to better environmental
management depends in part on the knowledge of the environmental benefits or costs of our
actions towards the environment and at the same time looking at how important the environment
is to the overall performance of the firm.

Traditionally, one theoretical approach or a single level of analysis has been used to explain
firms‟ adoption of environmental management attributes, providing an incomplete picture.
Interaction between multiple theoretical perspectives and various levels of analyses is, however,
argued to be essential to provide a better explanation of such a complex phenomenon (Bansal
and Roth, 2000). Although multiple studies had addressed this issue in manufacturing, few
examine tourism businesses (Bohdanowicz, 2006, Chan and Wong, 2006).

The analysis starts at the individual-level arguing that environmental management attributes are
not only driven by organizational-level determinants but also they may be outcomes of
managers‟ environmental paradigms or belief systems. This is consistent with theories that
emphasize the importance of organizational actors holding eco-centric values to be able to help
their companies in the move towards sustainability (Gladwin et al., 1995; Shrivastava, 1995a,
Stead and Stead, 1992; Starik and Rands, 1995). Empirical research has shown also that eco-
oriented managers may play a role in corporate greening, although more empirical analyses are
still needed in this area. Andersson and Bateman (2000), for example, have demonstrated the
critical role that a “strong environmental paradigm” plays in a firm‟s decision to adopt
environmental management attributes. Applying the Ajzen theory of planned behavior, Cordano
and Frieze (2000) and Flannery and May (2000) have also identified managers‟ attitudes as an
important antecedent to preferences for source reduction activity. In this light, hotel or tourism
businesses in general are expected to vary in terms of their level of environmental commitment
according to how strongly their executives embrace eco-centric values inherent in their beliefs
systems. Besides, the understanding of the community members surrounding the tourism
business about the environment may also help to have an influence on how much of the

Page 7 of 20
environmental attributes would they want to change to improve the recreational value and
increase the welfare of the community from the national park.


3.0 METHODOLOGY
 In this paper, we are using the data given about the socioeconomic characteristics of the
respondents on various questions concerning their preference on the environmental attribute they
wanted to see it improved. The environmental attributes were water quality in the lake in the
park, preservation of elephants, introduction of the white rhinos, and building of a guest house
near the park which were coded 1, 2, 3 and 4 respectively. The socioeconomic attributes of
respondents were income, education level, age and awareness measuring the level of knowledge.

To estimate how each of the independent variables affect the probability of opting for a
particular environmental attribute, the multinomial logit was employed as an analytical
technique.


3.1 THE EMPIRICAL MODEL
The Multinomial Logit model used in this paper is basically represented as follows;



Where;

Environattribute is the dependent variable (where 1, 2, 3, 4 represent water quality in the lake in
the park, preservation of elephants, introduction of white rhinos and building of a guest house
near the park respectively)

The betas are the parameters to be estimated.

  is the stochastic disturbance error term assumed iid

Age, income, awareness (where 1 implies an individual knows about the park and 0 means
otherwise) and education (where 1, 2, 3 and 4 mean no school, primary, secondary and tertiary
education respectively) are hypothesized to affect the individual preference of the environmental
attribute improvement.

In the multinomial logit model, the numbers 1 up to 4 do not indicate order of importance. In
other words 1 is not any better to all other attributes.

In computing the probabilities of each of the options, the formula for option one for instance is as
follows;




Page 8 of 20
In all the four probabilities, the base category is option four.




Page 9 of 20
4.0 RESULTS AND DISCUSSION
The table 4.1 below provides the summary for the Multinomial Logit output. In this table base
category is 3. The overall model when the base is 3, is statistically adequate and reliable as
indicated by probability of the chi-square value. The p-value of 0.0222 shows that the model is
reliable at 5% level of significance. However, except awareness under outcome 2, all other
coefficients are statistically insignificant, meaning that relative to base category 3, most factors
do not affect the individual‟s probability of opting for a particular environmental attribute
improvement.

Table 4.1 BASE CATEGORY 3
Multinomial logistic regression                                Number of obs       =           250
                                                               LR chi2(12)         =         23.71
                                                               Prob > chi2         =        0.0222
Log likelihood = -304.50686                                    Pseudo R2           =        0.0375


           env           Coef.     Std. Err.          z      P>|z|       [95% Conf. Interval]

1
         age        -.0454051      .0555393        -0.82     0.414      -.1542602        .0634501
levelofedu~n        -.0209394       .210021        -0.10     0.921      -.4325731        .3906942
   awareness          .011212      .3603275         0.03     0.975      -.6950169        .7174409
         inc         .0004403      .0008119         0.54     0.588      -.0011509        .0020315
       _cons         .6533387      1.307279         0.50     0.617       -1.90888        3.215558

2
         age         .1106172      .0870123         1.27     0.204      -.0599237        .2811581
levelofedu~n         .0261653      .2908959         0.09     0.928      -.5439802        .5963109
   awareness         20.48798      2.160454         9.48     0.000       16.25357        24.72239
         inc         .0007475      .0011985         0.62     0.533      -.0016015        .0030965
       _cons        -24.53341             .            .         .              .               .

4
         age         .0313629       .067926         0.46     0.644      -.1017697        .1644954
levelofedu~n         .3842928      .2806613         1.37     0.171      -.1657932        .9343789
   awareness        -.4411668      .4423297        -1.00     0.319      -1.308117        .4257835
         inc         .0001335      .0010565         0.13     0.899      -.0019372        .0022042
       _cons        -2.619674      1.671624        -1.57     0.117      -5.895998        .6566491

(env==3 is the base outcome)



In all outcomes relative to the same base as in table 4.1 above, age, level of education in outcome
1 and awareness in outcome 4 have negative direction which means that any changes in the such
regressors, may reduce the probability while the rest of the regressors in all outcomes have
positive direction implying increase in probability if the regressors change.

Table 4.2 below provides a summary of the multinomial logit when the base category is changed
from 3 as in above case to 1. As it is the case in table 4.1 above, the chi-square p-value (0.0222)
is also statistically significant at 5% level of significance. This means that the overall model is
reliable for policy implications. In other words, the model results are reliable enough to make
decisions on. By changing the base category to 1, it is only awareness variable in outcome 2
relative to base 1 which is statistically significant while all other variables in all outcomes are
statistically insignificant. At the same time, all coefficients have positive direction in outcome 2,
and coefficients for income and awareness have negative direction in outcomes 3 and 4.


Page 10 of 20
Table 4.2 base category 1
Multinomial logistic regression                                 Number of obs       =           250
                                                                LR chi2(12)         =         23.71
                                                                Prob > chi2         =        0.0222
Log likelihood = -304.50686                                     Pseudo R2           =        0.0375


           env           Coef.      Std. Err.          z     P>|z|        [95% Conf. Interval]

2
         age         .1560223       .0894742        1.74      0.081      -.0193439        .3313885
levelofedu~n         .0471048       .3004365        0.16      0.875        -.54174        .6359495
   awareness         20.47677       2.207063        9.28      0.000         16.151        24.80253
         inc         .0003072       .0012115        0.25      0.800      -.0020673        .0026818
       _cons        -25.18675              .           .          .              .               .

3
         age         .0454051       .0555393        0.82      0.414      -.0634501        .1542602
levelofedu~n         .0209394        .210021        0.10      0.921      -.3906942        .4325731
   awareness         -.011212       .3603275       -0.03      0.975      -.7174409        .6950169
         inc        -.0004403       .0008119       -0.54      0.588      -.0020315        .0011509
       _cons        -.6533387       1.307279       -0.50      0.617      -3.215558         1.90888

4
         age         .0767679       .0715182        1.07      0.283      -.0634052         .216941
levelofedu~n         .4052323       .2904159        1.40      0.163      -.1639724        .9744369
   awareness        -.4523788       .4598547       -0.98      0.325      -1.353677        .4489198
         inc        -.0003068       .0010825       -0.28      0.777      -.0024284        .0018148
       _cons        -3.273013        1.73302       -1.89      0.059      -6.669669        .1236434

(env==1 is the base outcome)



Regardless of the signs of the coefficients, except for awareness coefficient in outcome 2, all
other coefficients are statistically insignificant, meaning that the changes in the magnitudes of
the regressors do not have any influence of the probability.

Table 4.3 below provides the summary of the results for the multinomial logit output with base
category of 2. In the table below, we can see gain that the overall model is statistically significant
(p=0.0222) implying that it is a reliable model for policy implications.




Page 11 of 20
Table 4.3 base category 2
Multinomial      logistic    regression                         Number of obs       =           250
                                                                LR chi2(12)         =         23.71
                                                                Prob > chi2         =        0.0222
Log   likelihood =    -304.50686                                Pseudo R2           =        0.0375


           env              Coef.      Std.   Err.     z     P>|z|        [95%   Conf.   Interval]

1
         age         -.1560223         .0894742      -1.74    0.081      -.3313885        .0193439
levelofedu~n         -.0471048         .3004365      -0.16    0.875      -.6359495          .54174
   awareness         -20.47677         2.450439      -8.36    0.000      -25.27954         -15.674
         inc         -.0003072         .0012115      -0.25    0.800      -.0026818        .0020673
       _cons          25.18675          1.73302      14.53    0.000        21.7901        28.58341

3
         age         -.1106172         .0870123      -1.27    0.204      -.2811582        .0599237
levelofedu~n         -.0261653         .2908959      -0.09    0.928      -.5963109        .5439802
   awareness         -20.48798         2.450977      -8.36    0.000      -25.29181       -15.68415
         inc         -.0007475         .0011985      -0.62    0.533      -.0030965        .0016015
       _cons          24.53341         1.671624      14.68    0.000       21.25709        27.80974

4
         age         -.0792544         .0985223      -0.80    0.421      -.2723545        .1138458
levelofedu~n          .3581275         .3557628       1.01    0.314      -.3391548         1.05541
   awareness         -20.92915          2.47944      -8.44    0.000      -25.78876       -16.06953
         inc          -.000614         .0014059      -0.44    0.662      -.0033694        .0021414
       _cons          21.91374                .          .        .              .               .

(env==2   is    the base    outcome)

.




In the table 4.3 above, it is only coefficient for awareness that is statistically significant
(p=0.0000) in all outcomes. It is also worth noting that all the coefficients have negative
directions in all outcomes apart from level of education in outcome 4 which has positive
direction.

Table 4.4 below gives a summary of the results for multinomial logit with the base category of 4.
The overall model is statistically significant (p=0.0222) and we can use the model results to
make policy decisions.

Table 4.4: base category 4
Multinomial      logistic    regression                         Number of obs       =           250
                                                                LR chi2(12)         =         23.71
                                                                Prob > chi2         =        0.0222
Log   likelihood =    -304.50686                                Pseudo R2           =        0.0375


           env              Coef.      Std.   Err.     z     P>|z|        [95%   Conf.   Interval]

1
         age         -.0767679         .0715182      -1.07    0.283       -.216941        .0634052
levelofedu~n         -.4052323         .2904159      -1.40    0.163      -.9744369        .1639724
   awareness          .4523788         .4598547       0.98    0.325      -.4489198        1.353677
         inc          .0003068         .0010825       0.28    0.777      -.0018148        .0024284
       _cons          3.273013          1.73302       1.89    0.059      -.1236433        6.669669

2
         age          .0792544         .0985223       0.80    0.421      -.1138458        .2723545
levelofedu~n         -.3581275         .3557628      -1.01    0.314       -1.05541        .3391548
   awareness          20.92915          2.47944       8.44    0.000       16.06953        25.78876
         inc           .000614         .0014059       0.44    0.662      -.0021414        .0033694
       _cons         -21.91374                .          .        .              .               .

3
         age         -.0313629          .067926      -0.46    0.644      -.1644954        .1017697
levelofedu~n         -.3842928         .2806613      -1.37    0.171      -.9343789        .1657932
   awareness          .4411668         .4423297       1.00    0.319      -.4257835        1.308117
         inc         -.0001335         .0010565      -0.13    0.899      -.0022042        .0019372
       _cons          2.619674         1.671624       1.57    0.117       -.656649        5.895998

(env==4   is    the base    outcome)

.



In the table 4.4 above, the coefficient of awareness under outcome 2 is the only coefficient which
is statistically significant while all other coefficients are statistically insignificant. At the same

Page 12 of 20
time, level of education coefficient under all outcomes has negative direction whereas age
coefficient under outcomes 1 and 3 has negative direction and income coefficient is negative
only in outcome 3 while the coefficient for awareness has positive direction in all outcomes.

Now, since we have an understanding about the significance of the coefficients of the various
factors and their direction of influence, it is important to interpret the coefficients corresponding
to changes in the independent variables. Interpreting the coefficients for multinomial logit
requires the marginal effects so that we may know by how much will the probability change
when there is a unit change in the magnitude of the independent variable (except the dummy
variable).

Table 4.5 below provides the summary of the marginal effects when the base category is 1. In
this table, it is only awareness marginal effect that is significant while all others are insignificant,
the probability when the environmental attribute is preservation of elephants relative to water
quality is computed to be 0.00101113(1v2).

Table 4.5 marginal effects base category 1
Marginal effects after mlogit
      y = Pr(env==2) (predict, outcome(2))
         = .00101113

variable             dy/dx      Std. Err.          z     P>|z|    [      95% C.I.      ]        X

     age         .0001224          .00014       0.90     0.368     -.000144     .000389        22.76
levelo~n        -.0000368          .00027      -0.14     0.891     -.000564     .000491        3.272
awaren~s*        .1325244            .026       5.10     0.000      .081573     .183476         .756
     inc         5.71e-07          .00000       0.51     0.613     -1.6e-06     2.8e-06       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1
The interpretation for the marginal effect for awareness for instance, if the individual knows or
as information about the park, the probability to opt for the preservation of elephants will
increase by about 13.3%. In other words, if the individual has knowledge about the park, there is
13.3% chance that that individual will choose preservation of elephants relative to water quality
in the lake in the park.

Table 4.6 below gives a summary of the marginal effects of base category 1 but in relation to
outcome 3. In the table, all the marginal effects for all the regressors are insignificant meaning
that changes in the regressors do not have an influence on the probability of opting for a
particular environmental attribute. However, we still need o give an interpretation. The
probability of opting for introduction of white rhinos (envi=3) is computed as 0.45916495.




Page 13 of 20
Table 4.6 1V3
Marginal effects after mlogit
      y = Pr(env==3) (predict, outcome(3))
         = .45916495

variable            dy/dx      Std. Err.         z     P>|z|    [     95% C.I.      ]        X

     age         .0047807         .01227      0.39     0.697    -.019259     .028821        22.76
levelo~n        -.0287236         .04748     -0.61     0.545    -.121774     .064326        3.272
awaren~s*       -.0240158         .07777     -0.31     0.757    -.176435     .128404         .756
     inc        -.0000838         .00018     -0.46     0.649    -.000445     .000277       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1


If we take age for example, it means that if somebody‟s age increases from 23 years to 24 years,
the probability of that individual opting for introduction of white rhinos relative to water quality
will increase by about 0.04%. But the probabilities for level of education, income will go down.
For awareness, If the person has knowledge about the park, the probability of opting for
introduction of the white rhinos will decline by almost 2.4%.

Table 4.7 below provides a summary of the marginal effects of the multinomial logit when the
base category is one but with respect to outcome 4. The predicted probability that an individual
will opt for building of a guest house relative to water quality in the park is p(env==4)
=0.18220019. Again, all the marginal effects are insignificant.

Table 4.7 1V4
Marginal effects after mlogit
      y = Pr(env==4) (predict, outcome(4))
         = .18220019

variable            dy/dx      Std. Err.         z     P>|z|    [     95% C.I.      ]        X

     age         .0076113         .00949      0.80     0.422    -.010985     .026208        22.76
levelo~n         .0586205         .03884      1.51     0.131    -.017498     .134738        3.272
awaren~s*       -.0934957         .06786     -1.38     0.168    -.226503     .039511         .756
     inc        -8.93e-06         .00015     -0.06     0.952    -.000298      .00028       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1

Taking level of education as example, it means that if the education of an individual increases
from secondary to tertiary, the probability that that individual will opt for building of a guest
house relative to water quality will increase by about 5.86%. if the person has knowledge about
the park, the probability of opting building the guest house will decline by 9% where as income
increases the probability of opting for the building guest house will decline with negligible
component.

Let us see what can happen to the direction and significance of the marginal effects if we change
the base from1 to 2.

Page 14 of 20
Table 4.8 gives a summary of the marginal effects when the base category is 2 with respect to
outcome 1 (water quality). The probability p (env==1) that an individual will opt for water
quality relative to preservation of elephants is 0.35762372. All the marginal effects are
insignificant.

Table 4.8 1V2
Marginal effects after mlogit
      y = Pr(env==1) (predict, outcome(1))
         = .35762372

variable            dy/dx      Std. Err.         z      P>|z|   [      95% C.I.      ]        X

     age        -.0125144         .01195      -1.05     0.295    -.035938     .010909        22.76
levelo~n          -.02986          .0455      -0.66     0.512    -.119039     .059319        3.272
awaren~s*       -.0150128          .0734      -0.20     0.838    -.158866      .12884         .756
     inc         .0000922         .00017       0.53     0.597     -.00025     .000434       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1

The marginal effects mean that for instance, if the income increases by a unit, the probability that
an individual will opt for water quality relative to preservation of elephants will increase by
0.009%. For age, the probability will decrease by 1.25%, level of education will decrease the
probability by about 3% and if an individual has knowledge about the park will decrease the
probability by 1.5%.

Table 4.9 gives a summary of the marginal effects when the base category is 2 with respect to
outcome 3 (introduction of rhinos). The probability p (env==3) that an individual will opt for
introduction of rhinos relative to preservation of elephants is 0.45916495. All the marginal
effects are statistically insignificant.

Table 4.9 2V3
Marginal effects after mlogit
      y = Pr(env==3) (predict, outcome(3))
         = .45916495

variable            dy/dx      Std. Err.         z      P>|z|   [      95% C.I.      ]        X

     age         .0047807         .01227       0.39     0.697     -.01926     .028821        22.76
levelo~n        -.0287236         .04748      -0.61     0.545    -.121775     .064328        3.272
awaren~s*       -.0240158         .07777      -0.31     0.757    -.176435     .128404         .756
     inc        -.0000838         .00018      -0.46     0.649    -.000445     .000277       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1

These marginal effects mean that for example if age increases from 23 to 24 the probability that
an individual will opt for introduction of rhinos relative to preservation of elephants will increase
0.47% and decrease by 0.008% if income increases by unit, decrease by 2.9% if education goes
up to tertiary and decrease by 2.4% if an individual has knowledge about the park.

Table 5.0 below provides a summary of the marginal effects when the base category is 2 with
respect to outcome 4 (building guest house). The probability p (env==4) that an individual will

Page 15 of 20
opt for building guest house relative to preservation of elephants is 0.18220019. All the marginal
effects are statistically insignificant.

Table 5.0 2V4
Marginal effects after mlogit
      y = Pr(env==4) (predict, outcome(4))
         = .18220019

variable            dy/dx      Std. Err.        z     P>|z|    [     95% C.I.      ]        X

     age         .0076113         .0095       0.80     0.423   -.011003     .026226        22.76
levelo~n         .0586205        .03885       1.51     0.131   -.017517     .134758        3.272
awaren~s*       -.0934957        .06786      -1.38     0.168   -.226503     .039511         .756
     inc        -8.93e-06        .00015      -0.06     0.952   -.000298      .00028       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1

In the table 5.0 above, if age increases by a unit, the probability of opting building guest house
relative to preservation of elephants will increase by 0.7%, level of education will decrease the
probability by 5.9%, knowledge about the park will decrease the probability by 9% and increase
in income will decrease the probability by a negligible component of 0.000893%.

Table 5.1 below gives a summary of the marginal effects of multinomial logit when the base
category is 3 (introduction of white rhinos) relative to option 4 (building guest house). The
probability of an individual opting for building guest house relative to introduction of white
rhinos [p (envi==4)] is. The marginal effects of all the regressors are statistically insignificant
meaning that their marginal effect to the probability is equal to or close to zero.

Table 5.1 3V4
Marginal effects after mlogit
      y = Pr(env==4) (predict, outcome(4))
         = .18220019

variable            dy/dx      Std. Err.        z     P>|z|    [     95% C.I.      ]        X

     age         .0076113        .00949       0.80     0.422   -.010985     .026208        22.76
levelo~n         .0586205        .03884       1.51     0.131   -.017497     .134738        3.272
awaren~s*       -.0934957        .06786      -1.38     0.168   -.226503     .039511         .756
     inc        -8.93e-06        .00015      -0.06     0.952   -.000298      .00028       431.96

(*) dy/dx is for discrete change of dummy variable from 0 to 1

If for example income increases by a unit from 432 to 433, the probability of an individual
opting for building guest house relative to introduction of white rhinos will decrease by
0.00000893%, if an individual has knowledge about the park, the probability will decrease by
9.3%, education level will increase the probability by 5.86% and age will increase it by 0.078%.

The table 5.2 below provides the summary of the relative odds ratio which gives the relative risk
ratio of the change in the regressors.



Page 16 of 20
Table 5.2 relative risk ratio
Multinomial logistic regression                                Number of obs        =          250
                                                               LR chi2(12)          =        23.71
                                                               Prob > chi2          =       0.0222
Log likelihood = -304.50686                                    Pseudo R2            =       0.0375


           env              RRR    Std. Err.          z      P>|z|       [95% Conf. Interval]

1
         age         .9556103       .053074        -0.82     0.414         .857049       1.065506
levelofedu~n         .9792783       .205669        -0.10     0.921        .6488374       1.478006
   awareness         1.011275      .3643902         0.03     0.975         .499066       2.049183
         inc          1.00044      .0008122         0.54     0.588        .9988497       1.002034

2
         age         1.116967      .0971899         1.27     0.204        .9418364       1.324663
levelofedu~n         1.026511      .2986078         0.09     0.928        .5804334       1.815409
   awareness         7.90e+08      1.71e+09         9.48     0.000        1.15e+07       5.46e+10
         inc         1.000748      .0011994         0.62     0.533        .9983998       1.003101

4
         age          1.03186      .0700901         0.46     0.644        .9032376       1.178798
levelofedu~n         1.468575      .4121723         1.37     0.171        .8472214       2.545632
   awareness         .6432854      .2845442        -1.00     0.319        .2703286       1.530789
         inc         1.000134      .0010567         0.13     0.899        .9980646       1.002207

(env==3 is the base outcome)

These relative risk ratios may change subject to the base category. As seen from the table 5.2, it
is only awareness relative risk ratio in outcome 2 that is statistically significant and it actually
means that the relative risk ratio of a unit change in awareness (if the person has knowledge
about the park) relative to the introduction of the white rhinos is 790000000 units. In other
words, the risk of outcome 2 (preservation of elephants) relative to base category 3 (introduction
of rhinos) is 790000000.

The interpretation of all other outcomes is the same. For example, if the income level in outcome
2 (preservation of elephants) changes by a unit, the risk ratio relative to introduction of rhinos is
1.000748, for outcome 4 risk ratio is 1.000134 and so forth.




Page 17 of 20
5.0 CONCLUSIONS AND RECOMMENDATION
After going through the nitty gritty of the multinomial logit model on the choice of the
preference of the environmental attribute improvement, we can conclude that level of awareness
of an individual has positive contribution to somebody‟s choice of an environmental attribute
improvement. This is so not only In theoretical applications of the model but also in real life
situations, if an individual has knowledge or information about something, the decisions that are
made are well informed unlike somebody who is not informed of some attributes about the park.
No wonder, it is only information/knowledge about the park that apparently influences the
probability of opting for an environmental attribute improvement not others because, even if you
are educated, well advanced in years and have a lot of income but you are not informed about a
particular situation, your choice can not be rational. So, it makes a lot of sense to see that these
other three regressors (income, education and age) are insignificant statistically even though to
extent may help to make decisions. Due to this finding, this paper makes the following
recommendations;

         Through awareness campaigns, trainings, outreaches inform the community members
         be about the park( what it offers, how to participate, charges)
         Ensure that all the categories of individuals poor, rich, educated or not be informed to
         belong to the park




Page 18 of 20
6.0 REFERENCES
Alvarez Gil, M. J., J. Burgos Jimenez, and J.J. Céspedes Lorente. (2001). An analysis of
environmental management, organizational context and performance of Spanish hotels. Omega
29 (6), 457-471.

Andersson, L. M. and T.S. Bateman. (2000). Individual Environmental Initiative, Championing
Natural Environmental Issues in U.S. Business Organizations. Academy of Management Journal
43 (4), 548-570.

Bansal, P. and K. Roth. (2000). Why Companies Go Green, A Model of Ecological
Responsiveness. Academy of Management Journal 43 (4), 717-737.

Bohdanowicz, P. (2006). Environmental awareness and initiatives in the Swedish and Polish
hotel industries-survey results. International Journal of Hospitality Management 25 (4), 662-
682.

Chan, E. and Wong, S. (2006) Motivations for ISO14001 in the hotel industry. Tourism
Management 27, 481-492

Christopher F. Clarka, Matthew J. Kotchenb,c, Michael R. Moore (2002). Internal and external
influences on pro-environmental behavior: Participation in a green electricity program. Journal
of Environmental Psychology 23 (2003) 237–246

Cordano. M, & Frieze, I.H. (2000). Pollution reduction preferences of US environmental
managers, applying Ajzen‟s theory of planned behavior. Academy of Management Journal 43
(4), 627–641.

Flannery, B. L. & May D. R. (2000). Environmental ethical decision-making in the U.S. metal-
finishing industry Academy of Management Journal 43(4), 642–663.

Gladwin, T. N., Kennely, J.J. & Krause, T.S. (1995). Shifting paradigms for sustainable
development, implications for management theory and research Academy of Management
Review 20 (4), 874-907.

Guagnano, G. A., Stern, P. C., & Dietz, T. (1995). Influences on attitude–behavior relationships:
A natural experiment with curbside recycling. Environment and Behavior, 27(5), 699–718.

Henriques, I. Y. & Sadorsky, P. (1996). The Determinants of an Environmentally Responsible
Firms, An Empirical Approach. Journal of Environmental Economics and Management 30, 381-
395.
Henriques, I. Y. & Sadorsky, P. (1999). The Relationship between Environmental Commitment
and Managerial Perceptions of Stakeholder Importance. Academy of Management Journal 42
(1), 87-99.



Page 19 of 20
Hoffman, A. J. (1999). Institutional Evolution and Change, Environmentalism and the U.S.
Chemical Industry. Academy of Management Journal 42 (4); 351-371.

Moon, S. G. and P. de Leon. (2007). Contexts and Corporate Voluntary Environmental
Behaviors, Examining the EPA‟s Green Lights Voluntary Program. Organization and
Environment 20 (4), 480-496.

Rivera, J. (2002). Assessing a voluntary environmental initiative in the developing world, The
Costa Rican Certification for Sustainable Tourism. Policy Sciences 35 (4), 333-360.

Rivera, J. (2004). Institutional pressures and voluntary environmental behavior in developing
countries, evidence from the Costa Rican hotel industry. Society and Natural Resources 17, 779-
797.

Sharma, S. (2000). Managerial Interpretations and Organizational Context as Predictors of
Corporate Choice of Environmental Strategy. Academy of Management Journal 43 (4), 681-697.

Shrivastava, P. (1995a). The Role of Corporations in Achieving Ecological Sustainability.
Academy of Management Review 20 (4), 936-960.

Starik, M. & Rands, G. (1995). Weaving an integrated web, Multilevel and multisystem
perspectives of ecologically sustainable organizations. Academy of Management Review 20 (4),
908-935.

Stead, W. E. & Stead, J.G. (1992). Management for a Small Planet, Strategic Decision Making
and the Environment. Newbury Park, CA, Sage.

Van Liere, K., & Dunlap, R. (1980). The social bases of environmental concern: A review of
hypotheses, explanations, and empirical evidence. Public Opinion Quarterly, 44, 181–197.




Page 20 of 20

								
To top