2019-3061-AJT

1

International Tourism Demand and Determinant 1

Factor Analysis in Ethiopia 2

3 In Ethiopia, to position tourism as a key economic sector, recently it is identified as 4 one of the major strategies of using tourism for poverty alleviation. The country 5 receives tourist from various source markets and tourism in Ethiopia has grown over 6 the years. As the country gears up its efforts to make itself a top five tourism premier 7 destination in Africa, it becomes more important to understand the factors that 8 influence international tourism demand among others things to attain economic 9 growth. International tourism demand for Ethiopia lags behind from other African 10 countries like Egypt, Morocco, Tunisia, Kenya and South Africa. Furthermore, the 11 tourism product offered by Ethiopia is becoming increasingly noncompetitive. Hence, 12 there is need to offer demand driven tourism products that ensure visitors to come to 13 Ethiopia and stays longer. Motivated by this need, the study sought to investigate the 14 determinants of international tourism demand. Specifically, on the effect of economic 15 factors, tourist socio-demographic characteristics, political factors and destination 16 characteristics on international tourism demand in Ethiopia. The study used both 17 longitudinal and cross sectional research designs and panel data for economic 18 variables from eleven countries for the period of only one month time as of (10

th 19

December 2018 and 10th January 2019). Data were collected from the World Bank 20

Database, United Nations Database, International Monetary Fund Database and 21 Ministry of Culture and Tourism Statistics. Survey data were collected from individual 22 tourists leaving the country using questionnaires. The study used a dynamic panel 23 regression model to determine the effect of economic factors on international tourism 24 demand and a count data regression model to determine the effect of socio-25 demographic characteristics, political factors and destination characteristics on 26 international tourism demand. The study results indicated that tourism price, 27 travelling cost, trade openness and word of mouth effect were the main economic 28 factors influencing international tourism demand in Ethiopia. The tourist’s socio-29 demographic characteristics such as annual household income, age and occupational 30 status were found to significantly influence international tourism demand. The 31 political factors composite index and destination characteristics composite index were 32 also important determinants of international tourism demand. Taking into 33 consideration of all these factors affecting tourism demand, the government and all 34 the tourism stakeholders should work towards making Ethiopia’s tourism product 35 competitive; prices remain reasonable, development of tourism infrastructure and 36 offering quality services, with diversification of tourism product. 37 38 Keywords: International Tourism Demand, Tourism Products, Destination 39 Characteristics and Political Factors. 40 41

42

Introduction 43 44

Background of the Study 45 46

The World Tourism Organization (WTO) has recognized tourism as one of 47 the largest and fastest growing industries in the world. The growth of tourism 48 industry is demonstrated by the ever increasing number of destinations opening 49

2019-3061-AJT

2

up and investing in tourism development, turning modern tourism into a key 1 driver for socio-economic progress through the creation of jobs and enterprises, 2 infrastructure development and the export revenues earned (WTO, 2017). 3

According to World travel and tourism council (WTTC, 2018) travel and 4

tourism in 2017 globally employed about 8.7 percent of total employment, 5 generated 9.1 percent of total gross domestic product and visitor exports 6 generated US$1,170.6 billion (5.3 percent of total exports). 7

International tourism plays an important role of promoting world peace, 8 both by providing an incentive for peace keeping and by building a bridge 9

between cultures (Eilat & Einav, 2004). In developing countries, tourism plays 10 an important role in stimulating investments in new infrastructure, as well as 11 generating government revenues through various taxes and fees. 12

In Africa, tourism has been identified as a key sector for the achievement 13 of shared economic growth and poverty alleviation (Mitchell & Ashley, 2006; 14 World Bank, 2006). International tourist arrivals have shown increased growth 15 rising from 25 million in 1950, to 278 million in 1980, 528 million in 1995, 16 and 2,035 million in 2018. According to UNWTO Tourism Towards 2030, the 17

number of international tourist arrivals worldwide is expected to reach 1.8 18 billion by the year 2030 (WTO, 2018). 19

However, world tourism has experienced continued expansion and 20 diversification over the past six decades, Africa’s tourism market share remains 21

small compared to other world regions. This is despite the fact that, Africa has 22 a lot to offer that cannot be found elsewhere as it holds a rich history as the 23

continent of the explorers and as a place for adventures (Christie & Crompton, 24

2001). 25

In Africa there are unique tourist destinations, where some of the greatest 26 views in the world and natural attractions find. This is true not only for the 27

natural resources, but also for its culture, traditions and customs. Likewise, 28 Ethiopia offers a variety of travel experience as it is bestowed with diverse 29 tourist attractions to foreign visitors. 30

Tourist attractions in Ethiopia comprise abundant wildlife, natural habitats, 31 world heritage sites and a number of cultural attractions from the various ethnic 32 groups. Similarly, Addis Ababa the capital city with good hotels and 33 conference facilities is well positioned to attract the business markets such as 34

meetings, incentives, conferences and exhibitions. 35 Ethiopia receives tourism from various source markets and tourism in 36

Ethiopia has grown over the years. As Ethiopia gears up its efforts to make 37 itself a top five tourism premier destination in Africa, it becomes more 38 important to understand the factors that influence international tourism demand 39 for Ethiopia among others things. 40

41

Statement of the Problem 42 43

International tourism demand for Ethiopia lags behind other African 44 countries like Egypt, Morocco, Tunisia, Kenya and South Africa (WTO, 2017). 45

2019-3061-AJT

3

This is despite the fact that, Ethiopia is bestowed with diverse tourist 1 attractions and thus has great tourism potential. 2

Furthermore, the number of tourist’s arrivals to Ethiopia from different 3 corner of the world does not increase constantly but have experienced cyclical 4

fluctuations over the years. Likewise, the Ethiopia tourism product offered to 5 the market is becoming increasingly noncompetitive. Ethiopia is experiencing 6 problems of competition due to degradation and reduction of the quality of 7 Ethiopia’s tourism product as more tourists are switching to other countries in 8 the region such as Kenya, Uganda, Tanzania, Egypt and Zimbabwe which offer 9

similar tourist attractions (World Bank, 2016). 10 There is a need therefore, for Ethiopia to offer demand driven tourism 11

products that ensure visitors to come and stays longer. The government, 12

tourism planners and marketers therefore need to clearly understand which 13 important factors influence international tourist’s decision to visit Ethiopia as 14 chosen destination. 15

Song and Li (2008) noted that identifying the determinants of tourism 16 demand and estimating magnitudes of their influence on tourism demand are of 17

great interest to decision makers in tourist destinations. Hence, an 18 understanding of tourism demand is a starting point for the analysis of why 19 tourism develops, who patronizes specific destinations and what appeals to the 20 client market. 21

Empirical studies on international tourism demand have focused on 22 tourism demand in developed countries while Africa has received very little 23

attention (Rogerson, 2007; Xiao & Smith, 2016). The studies have mainly 24

considered the economic factors affecting tourism demand even though it 25

ignores the non-economic factors due to unavailability of data. 26 Eilat and Einav (2004) and Naude and Saayman (2005) noted that when 27

modeling international tourism demand in Africa, both economic and non-28 economic factors such as the destination characteristics should be taken into 29 consideration since tourism demand may also be significantly affected by non-30

economic factors. 31 Empirical studies explaining international tourism demand for Ethiopia are 32

limited. Previous studies on tourism in Ethiopia mainly focused on tourism 33 development (Ondicho, 2016), choice of attractions, expenditure, destination 34

competitiveness and satisfaction of international tourists (Odunga, 2016) and 35 the economic contribution of tourism and carried out a study on estimation of 36

tourism demand (Valle & Yobesia, 2009). Also, the study did not consider 37 non-economic factors which can also influence international tourism demand 38 for Ethiopia. Hence, this study, therefore, attempted to fill this gap by 39 investigating the economic and non-economic factors influencing international 40 tourism demand. 41

42 Objectives of the Study 43 44

The general objective of the study is to investigate the determinant factors of 45 international tourism demand in Ethiopia. 46

47

2019-3061-AJT

4

Specific Objectives 1

2 a) To find out the effect of economic factors on international tourism 3

demand in Ethiopia. 4

b) To find out the effect of political factors on international tourism 5 demand in Ethiopia. 6

c) To investigate the effect of tourist socio-demographic characteristics on 7 international tourism demand in Ethiopia. 8

d) To discover the effect of destination characteristics on international 9

tourism demand in Ethiopia 10 11

Research Hypotheses 12 13

The study sought to test the following null hypotheses: 14

H01: There is no significant relationship between economic factors and 15 international tourism demand in Ethiopia. 16

H02: There is no significant relationship between political factors and 17 international tourism demand in Ethiopia. 18

H03: There is no significant relationship between tourist’s socio-19 demographic characteristics and international tourism demand in 20 Ethiopia 21

H04: There is no significant relationship between destination 22 characteristics and international tourism demand in Ethiopia 23

24

25 Literature Review 26

27 International Tourism Demand 28

29 Pearce (1995) referred to tourism demand as the relationship between 30

individual’s motivation to travel and their ability to do so. Song and Witt 31 (2000) considered tourism demand as the amount of a set of tourist products 32

that the consumers are willing to acquire during a specific period of time and 33 under certain conditions which are controlled by the explanatory factors used 34

in the demand equation. 35

The total demand for tourism is considered to consist of three basic 36

components which include actual demand, suppressed demand and no demand 37 (Cooper, Fletcher, Fyall, Gilbert & Wanhill, 2008; Page & Connell, 2006). The 38 no demand component constitutes the category of those who do not wish to 39 travel or are unable to travel. Suppressed demand refer to the section of the 40 population who do not travel for some reason while actual demand refer to the 41

aggregate number of tourists recorded in a given location or at a particular 42 point in time. 43

Page and Connell further noted that actual demand depended on the 44 specific features and characteristics of those product and service alternatives 45

2019-3061-AJT

5

that the customer evaluates to make the final purchase decision. This includes 1 choosing the destination, the time and duration of travel, the activities 2 undertaken at the destination, and the amount of money spent for the holiday. 3 Demand for tourism is segmented and is distinguished through a number of 4

different markets. 5 Tourism demand can be analyzed for groups of countries, individual 6

countries, regions or local areas. Demand can also be disaggregated by 7 categories as types of visits and types of tourists. Tourism visits can take place 8 for various reasons including holidays, business trips, visits to friends and 9

relatives (VFR), conferences, and religious purposes among others (WTO, 10 2017). 11

International tourism demand is usually measured in terms of the number 12

of tourist visits from an origin country to a destination country, in terms of 13 tourist nights spent in the destination country or in terms of tourist expenditures 14 by visitors from an origin country in the destination country. The number of 15 tourist arrivals is most frequently used as the measure of demand, followed by 16 tourist expenditure or tourist receipts (Crouch, 1994, 1995; Lim, 1999; Song & 17

Li 2008; Witt & Witt, 1995). 18

19 Determinants of International Tourism Demand 20

21 Tourism demand modeling and forecasting studies rely heavily on 22

secondary data in terms of model construction and estimation. To determine 23

the explanatory variables influencing tourism demand, the empirical models 24

borrow heavily from the consumer theory (Song & Witt, 2008). 25

The traditional theory of demand considers the price of a commodity, 26 prices of other commodities, consumer’s income and tastes as the most 27

important determinants of market demand (Koutsoyians, 1991). Koutsoyians 28 further explains that, demand is also affected by other factors such as the 29 distribution of income, total population and its composition, wealth, credit 30

availability, government policy, past levels of demand and past levels of 31 income. 32

Cooper et al., (2008) noted that at the individual level the factors that 33 influence demand for tourism are closely linked to models of consumer 34

behavior and that no two individuals are alike and differences in attitudes, 35 perceptions, images and motivation have an important influence on travel 36

decisions. Goeldner and Ritchie (2003) also stated that, the demand for travel 37 to a particular destination is a function of the person’s propensity to travel and 38 the reciprocal of the resistance of the link between origin and destination areas. 39

A person’s propensity to travel is determined largely by his/her 40 psychographic profile, socio economic status and marketing effectiveness. 41

Resistance relates to the relative attractiveness of various destinations and 42 depends on the economic distance, cultural distance, cost of tourist services, 43 quality of service and seasonality. 44

Though tourism demand could be affected by a wide range of factors, such 45 as economic, attitudinal and political factors, the most commonly considered 46

2019-3061-AJT

6

factors influencing tourism demand are origin income, the own price of a 1 destination, substitute prices of alternative destinations and dummy variables to 2 capture the effects of one-off events (Song, Li, Witt & Fei, 2010). Overviews 3 of previous empirical studies on international tourism demand (Li, Song & 4

Witt, 2005; Lim, 1999; Song & Li, 2008) revealed that tourists income, tourism 5 prices in a destination relative to those in the origin country, tourist prices in 6 the competing destinations and exchange rates were the most significant 7 determinants of international tourism demand. 8

The accuracy of different modeling methods vary for data from country to 9

country over different time periods, and therefore, neither a single model nor a 10 single method will necessarily be appropriate for all origin-destination pairs 11 (Witt & Witt, 1995). 12

13 International Tourism in Ethiopia 14

15 Ethiopia offers foreign visitors a variety of travel experience as it is 16

bestowed with diverse tourist attractions. Addis Ababa the capital city with 17

good hotels and conference facilities is well positioned to attract the business 18 markets such as meetings, incentives, conferences and exhibitions. Tourism in 19 Ethiopia for many years played a vital role in the economy of the country. The 20 total contribution of travel and tourism to GDP in Ethiopia (including wider 21

effects from investment, the supply chain and induced income impacts) was 22 ETB 91, 898.4 million in 2014 and grow by 1.6% to ETB 93, 330.3 million in 23

2015. It is also forecast to rise by 4.9% pa to ETB 150, 738.0 Million by 2025 24

(WTTC, 2016). 25

Furthermore, tourism in Ethiopia has led to growth of industries such as 26 transport sector, accommodation, entertainment, travel agencies and related 27

services, administration, finance and health, among others, which are directly 28 linked to it. Through its multiplier effect, tourism has stimulated growth in 29 several other industries like construction, agriculture, hand crafts, 30

manufacturing and processing. Tourism development has led to economic 31 growth of the country through creation of indirect and direct job opportunities. 32 33

34

Research Methodology 35 36 Research Design 37 38

The study used both longitudinal and cross sectional research design 39 besides a descriptive research design. In a longitudinal study, measurements 40 are taken on each variable over two or more distinct time periods. This allows 41 the researcher to measure change in variables over time. 42

In longitudinal studies of the panel variety, the research may study the 43

same subjects over time and the use of panel data in this study allowed 44 controlling for individual heterogeneity among the different countries in order 45 to avoid biased results. 46

2019-3061-AJT

7

The use of panel data also enabled to have more degrees of freedom than 1 with time-series or cross-sectional data, to control for omitted variable bias and 2 to reduce the problem of multicollinearity, hence improving the accuracy of 3 parameter estimates. In addition, panel data are better able to study the 4

dynamics of adjustment thus are better suited to study the dynamics of change 5 of tourism demand. 6

The panel data is constructed from observations of the several study 7 variables made in 11 countries, that is, United Kingdom, Germany, Italy, 8 France, United States of America, Canada, India, Japan, Israel and Uganda for 9

only one month of the time period as of (10th

December 2018 and 10th

January 10 2019). 11

A cross sectional design involves the collection of data from a number of 12

cases at a single point in time and is also referred to as a survey design. The 13 choice of the survey research design was made based on the fact that it allowed 14 the researcher to collect data from respondents at a particular point in time and 15 to measure the relationship of the study variables at the specified time in order 16 to describe how variables were related. 17

In this study, data from tourists visiting Ethiopia is collected to determine 18 the current status of demand for international tourism in Ethiopia with respect 19 to specified variables. Hence, the survey study enabled the researcher to learn 20 about the factors influencing tourists to visit Ethiopia as well as their personal 21

characteristics. 22 23

Target Population 24

25

The study has two target populations, that is, target population at national 26 level and at individual level. The target population at the national level is the 27

tourist departures from Ethiopia for tourists coming from eleven countries, 28 United Kingdom, Germany, Italy, France, United States of America, Canada, 29 India, Japan, Israel and Uganda for only one month of time period as of (10

th 30

December 2018 and 10th

January 2019). Accordingly, the researcher totally 31 takes 323 representing sample data from a total of 2000 target populations 32 (statistical abstract, 2019), that is, target population at national level and at 33 individual level. The target population at the national level is the tourist 34

departures from Ethiopia for tourists coming from eleven countries, which 35 accounts 2000. 36

37 Sampling Design 38 39

At the national level, the number of tourist departures to Ethiopia for only 40 one month of the time period as of (10

th December 2018 and 10

th January 41

2019) from the 11 countries is considered. The target population at the national 42 level is the tourist departures from Ethiopia for tourists coming from eleven 43 countries. This is calculated at a 95 percent level of confidence and an error 44 margin of 5 percent, using sample size determination formula. Based on the 45 following sample size determination formula adapted from Cochran (1963:75) 46

2019-3061-AJT

8

which provides a simplified method to calculate the sample sizes and the 1 following equation helps to yield a representative sample for proportions. 2

3 Where 4 5

=n0 is the sample size, 6 = is the abscissa of the normal curve that cuts off an area α the tails (1 – α 7 equals the desired confidence level, e.g., 95%) 8 =e is the desired level of precision, 9 =p is the estimated proportion of an attribute that is present in the population, 10 and q is 1-p. The value for Z is found in statistical tables which contain the area 11 under the normal curve. 12 p=.5 (maximum variability). Furthermore, suppose the researcher desire a 95% 13

confidence level and ±5% precision. 14 15 The resulting sample size is demonstrated as: 16

= (1.96) ² (.5) (.5) = 385

(.05)²

17 The sample size (n0) adjusted using the following equation as of: 18

19 20 21

22 23

24 Where: 25

26 =n is sample size 27 =N the population size 28

29

= 385

(385-1) = 323

2000

30 It is also possible to calculate with the use of sample size determination 31

formula adapted from Yamane (1967:886) which can also help to yield a 32 representative sample for proportions. 33

2019-3061-AJT

9

1 Where; N = the total population 2 n = the required sample size 3 e = the precision level which is = (±5%) (Maximum variability) 4

5 6

n = 20000

1+2000(.05)² =333

7 Accordingly, researcher will take 323 sample sizes and from this number 8

and the distributions of the sample size were distributed to those tourists that 9 came from 11 countries (United Kingdom, Germany, Italy, France, United 10 States of America, Canada, India, Japan, Israel and Uganda). 11

The representative section from the identified number of sample was 12

selected using systematic random sampling technique. 13 Six manager’s representatives from FDRE ministry of culture and tourism 14

were taken and it was selected purposively for interview through their 15 knowledge and experience plus the selection of the number will be conducted 16 by using convenience sampling technique (non-probability sampling) that is 17

based on their working experience. 18

19 Data Collection Instruments and Procedure 20 21

Data Collection Instruments 22 23

Primary and secondary data were used by employing both qualitative 24 (discussion, and in depth interview) and quantitative (questioner survey). The 25 primary data were collected through questionnaire and key informant 26 interview. Structured questionnaire will be applied to the sample individuals or 27

respondents. 28 Most questions in the questionnaire were close-ended questions and 29

contain and organized on the basis of the research specific objectives to ensure 30 relevance to the research problem. However, opportunities were given to the 31 respondents to say more through open-ended questions. 32

The questionnaires were broken into five sections representing general 33 information of the tourists and the number of nights spent. Tourist’s socio-34

demographic characteristics section which sought the demographical and social 35 characteristic of the tourist, international tourism demand section where the 36 numbers of nights spent by the tourist were determined. 37

Political factors section dealing with political factors which influenced the 38 tourist decision to visit. Destination characteristics section dealing with 39

destination attractions and facilities considered to have an influence on tourist’s 40

2019-3061-AJT

10

decision to come to and the final section on other information dealing with 1 finding out the satisfaction of the tourists with their holiday in Ethiopia. 2 Moreover, structured, unstructured and matrix questions were used in the 3 questionnaires. Accordingly, structured questions are easier to administer while 4

unstructured questions permit a greater depth of response. Matrix questions are 5 easier to complete hence the respondent is unlikely to be put off. Use of these 6 three types of questions enhanced collection of relevant data. 7

Key informant interview were also conduct to collect primary data. With 8 regard to this primary data collection instrument, structured interview were 9

undertaken with the selected institute. 10 Secondary data which are relevant to the study were included and obtained 11

from various sources such as Statistical Abstracts from FDRE, MoCT, World 12

Bank (WB), United Nations (UN) database and International Monetary Fund 13 (IMF) database. 14

15 Data Collection Procedure 16

17 Secondary data was collected through analysis of statistical abstracts, 18

international Monetary Fund financial statistics, and World Bank database and 19 statistical abstracts and entered in an Excel spreadsheet to easily organize the 20 data in panel form. 21

Primary data collection took approximately one month. The data was 22 collected from tourists leaving the country during the period 10

th December 23

2018 to 10th

January 2019 at the Bole international departure lounges. The data 24

was collected on three different days selected randomly every week for a 25

period of three weeks. 26 27

Empirical Model 28 29

The study followed the standard theory of demand which formed the 30

theoretical foundations for modeling international tourism demand. The study 31 used the single estimation framework and derived the international tourism 32 demand function for Ethiopia from all 11 countries rather than from a 33 particular country of origin. In order to explain the determinants of 34

international tourism demand for Ethiopia, the study used two regression 35 models. 36

A dynamic panel regression model was used to analyze the effect of 37 economic factors on international tourism demand for Ethiopia while a count 38 data regression model was used to determine the effect of socio-demographics 39 factors, political factors and destination characteristics on international tourism 40 demand for Ethiopia. 41

42 Dynamic Panel Regression Model 43 44

Panel regression models are based on panel data and the two commonly 45 used techniques in panel data estimation are the Fixed Effects Model (FEM) 46

2019-3061-AJT

11

and the Random Effects Model (REM). When the data contains a lagged 1 dependent variable in the regression equation the fixed effects and the random 2 effects estimators produces biased results. 3

In this case, dynamic panel regression models should be used. The 4

dynamic panel regression models are estimated using generalized method of 5 moments (GMM), a method of estimation that provides consistent estimates 6 (Baum, 2006; Roodman, 2009). 7

A dynamic log-linear model was specified as given in equation: 8 InTDit = βo + β1 InTDit-1 + β2 InGDPit + β3 InTPit + β4 InSPitSA + β5 9

InTCit + β6 InTOit + β7 D2018 + αi + Uit 10

Where: 11 12

=i denotes the cross section units (i = 1,2,...,11) and t denotes the time 13 period (t =1,2,...,21). 14 =InTDit and InTDit-1 = Number of tourist departures from Ethiopia back 15 to the ith tourists origin for year t and t-1 respectively. 16 =GDPit = Real GDP per capita for tourists origin country i at year t. 17

=TPit =Price of tourism services (destination country) at year t. 18 =TCit= Travel cost from origin country i to Ethiopia at year t. 19 =SPitSA= Price of tourism services in (competing destination) at year t. 20 =TOit = Trade openness between tourists origin country at year t. 21

=D2018= Dummy variable for political instability taking a value of one 22 in 2017 and zero otherwise. 23

=αi = Unobservable country effect and represents all factors affecting 24

international tourism demand that do not change over time. 25

= Uit = stochastic disturbance term. 26 =βo = Constant term 27

28 β1 (i = 1, 2, 3,..., 7) are the parameters to be estimated and refer to the 29 contemporaneous short- run effect of the independent variables on the 30

dependent variable at time t. However, an effect on time t will have an impact 31 on t + 1, t + 2, and so on through the lag of the dependent variable. The long 32 run effect can be calculated as = 33

β1 34

1- β1 35 The double logarithmic dynamic regression panel model was used in the 36

study to allow the researcher interpret the model parameters directly in terms of 37 demand elasticities. An elasticity greater than one, that is, where the demand is 38 elastic, implies that the demand for tourism services responds proportionately 39 more than the change in the independent variable. On the other hand, an 40 elasticity of less than one, that is, demand is inelastic, implies that the demand 41

for tourism services responds proportionately less than the change in the 42 explanatory variables. Income elasticity of demand is expected to be positive 43 for most goods and services. 44

2019-3061-AJT

12

Demand for basic goods and services should be income inelastic whilst 1 that for luxury items should be elastic. Negative income elasticity indicates 2 inferior goods, in this case an inferior tourism destination (Koutsoyians, 1991). 3

4

Count Data Regression Model 5 6

Since the study dependent variable is measured using count data, a count 7 data regression model is used. A count data regression model is used for 8 modeling the count of things as a function of covariates and the counts are non-9

negative integers. 10 The most commonly used count data regression models are Poisson and 11

negative Binomial models. The Poisson distribution is often used to model 12

count data when the events being counted are somewhat rare. As the expected 13 number of events increases, a normal distribution can be used as an 14 approximation to the distribution of counts of events. 15

The Poisson distribution assumes that the mean and variance of the data 16 are equal. The binomial regression model is used instead of Poisson if the data 17

is over dispersed, that is when the variance of the data is not equal to the mean 18 of the data (Heeringa, West & Berglund, 2010). Since the study data is found 19 to be over dispersed, the negative binomial model is used. 20

The negative binomial regression model in log-linear form is used to 21

address objectives two, three and four of the study. The model is given in 22 equation. 23

24

In E(y i /x i) =InTDit = βo + β1 Expenditure + β2 Income i + β3 Age i 25 + β4 DEi + β5 DOi + β6 DGi + β7 DMI + + β8 Di+ β9 DTi + β10 DRi ++ 26 β11 Poli+ εi 27

Var (y i /x i) = E(y i /x i) (1 + α) 28 29

Where: 30 31

= Y i Number of nights spent by a tourist in Ethiopia 32

= Expenditure i= Total amount of money a tourist spent in Ethiopia in US 33 dollars Expenditure 34 =income i= Annual household income of the tourist in US dollars 35

=Age i = Age of the tourist in years. The mid-point of grouped data issued 36

instead of age categories. 37

= DEi = Level of education dummy of the tourist with no college education 38 = 0 and at least college level education = 1 39

= DOi = Occupation status dummy of the tourist with unemployed = 0 and 40 employed = 1 41 = DGi = Gender dummy of the tourist where male = 0 and female = 1 42

= DMi = Marital status dummy of the tourist with single = 0 and married = 43 1 44

= DCi = Children dummy of the tourist with have no child = 0 and with 45 children = 1 46

2019-3061-AJT

13

= DTi = Trip companion dummy of the tourist with travelled alone = 0 and 1 travelled with companions = 1 2 = DRi = repeat visit dummy of the tourist with first time = 0 and repeat 3 visit = 1 4

= Poli = the composite index representing the political factors influencing 5 the ith tourist 6 = Desti = the composite index representing the destination characteristics 7 influencing the ith tourist 8 = εi= the random error term 9

= E (y i /x i) = Conditional mean of the dependent variable given the value 10 of the independent variables of the ith individual. 11 = α= Dispersion parameter. 12

Since count data regression techniques model the log of incident counts, 13 the coefficients can be interpreted as follows, for a one unit change in the 14 independent variable, the log of dependent variable is expected to change by 15 the value of the regression coefficient. 16

Rather than reporting Poisson or negative binomial results as a regression 17

coefficient, there is the option of measuring the effect of the independent 18 variable on the dependent variable through the Incidence Rate Ratio (IRR). The 19 IRR represents the change in the dependent variable in terms of a percentage 20 increase or decrease, with the precise percentage determined by the amount 21

IRR is either above or below 1. 22 23

Operationalization and Measurement of Variables 24

25

The variable of the study is operationalized as explained in table 1 below: 26

27 Table 1. Operationalization and Measurement of Variables 28 Category Variable Operationalization Measurement Hypothesized

direction of the

variable

Dependent

variable

International

tourism

demand

-Tourists

departures

-Number of nights

spent

Total number of

tourists leaving

the country back

to their country of

residence at

month t.

None

Independent Variables

Economic

factors

Word of

mouth effect

Dependent

variable

Number of tourist

who left the

country back to

their country of

residence in the

previous year.

Positive

Income

Real GDP per-

capita

Real GDP per

capita of the

tourist’s origin

country measured

at constant 2000

Positive/Negative

2019-3061-AJT

14

US$.

Tourism

price

Relative tourism

price

- Negative

Travelling

cost

The total expenses

travellers incur for

their

transportation

from their country

of origin to

destination

price in US dollars

at year t

multiplied by the

tourists country of

origin exchange

rates at year t.

Negative

Political factors

Political

factors

Travel restrictions Yes/ No Negative

Political

instability

Measured using a

scale of 1-4

Negative

Travel warnings

by origin country

government

Measured using a

scale of 1-4

Negative

Visa formalities Measured using a

scale of 1-4

Positive/Negative

Good relations

between tourist

origin country

and Ethiopia

Measured using a

scale of 1-4

Positive

Socio-

demographic

characteristics

Socio-

demographic

characteristic

Sex Male/ Female Unknown

Occupational

status

The occupational

status of the

tourist measured

using five

categories.

Positive

Occupation The profession of

the tourist

measured using

four categories.

Unknown

Annual household

income

Measured using

six categories.

Positive

Marital status Measured using

four categories.

Unknown

Family size of

tourist

Measured using

five categories.

Unknown

Party composition

of the tourist

Measured using

seven categories.

Unknown

Tourist spouse

Measured using a

scale of 1-3

Unknown

Tourist children

Measured using a

scale of 1-3

Unknown

Tourist friends &

relatives

Measured using a

scale of 1-3

Measured using a

scale of 1-3

Unknown

Tourist available

free time

Measured using a

scale of 1-3

Positive

2019-3061-AJT

15

Tourist paid

holiday

entitlement

Measured using a

scale of 1-3

Positive

Destination

attractiveness

Attractions

Natural (weather,

wildlife, scenery),

culture and sports

Measured using a

1-3 scale

Positive

Facilities

Availability and

quality of

accommodation,

food, drinks,

transport,

communication,

hospitality of

local people,

entertainment,

direct flights,

health facilities.

Measured using a

1-3 scale

Positive

Package

Refers to all

inclusive travel

package to

independent travel

Measured using a

1-3 scale

Unknown

Distance

travelled

Refers to distance

between AA and

tourist country of

residence capital

city.

Measured using a

1-3 scale

Negative

Security

Crime, terrorism,

ethnic, accidents

Measured using a

1-3 scale

Negative

1 2 Data Analyses and Presentation 3 4

The data collected was analyzed in accordance of the study objectives and 5 hypotheses. To address objective one, a dynamic panel data regression analysis 6

was carried out. Descriptive statistics (mean and standard deviation) were first 7 computed. 8

Before carrying out the regression analysis, the data was first tested. 9 Regression analysis requires that data be stationary in order to be able to make 10 meaningful inferences. The use of non-stationary variables in regressions leads 11

to spurious results. The study used the Fisher-type (Choi, 2001) panel 12 regression unit root tests to investigate whether there were any variables in the 13 model that were non-stationary. 14

Fisher-type tests conduct unit-root tests for each panel individually, and 15

then combine the p-values from these tests to produce an overall test. The 16 SGMM estimator was used to estimate the regression parameters. The 17 regression model diagnostic tests for autocorrelation and instruments over 18 identification were carried out using the Arellano- Bond tests and Hansen 19 statistic respectively. 20

2019-3061-AJT

16

The survey data was both qualitative and quantitative; the qualitative data 1 was first reduced through coding and categorization into common themes. The 2 reduced data was then displayed using frequency tables and charts and 3 conclusions drawn. 4

Quantitative data was analyzed using descriptive statistics such as the 5 frequency distribution, bar charts, pie charts, percentages and mean ranks. 6 Inferential statistics were also used for hypotheses testing. Before hypothesis 7 tests and regression were carried out, the data was first tested for normality 8 using the Shapiro-Wilk's test of normality. 9

To address objectives two, three and four of the study, a count data 10 regression model was used. For regression analysis purposes, the study used 11 summated scores of individual item scores on a scale of 1-4 for political factors 12

and summated scores of individual item scores on a scale of 1-3 for destination 13 attractiveness factors to come up with a composite score for each variable. 14 15 Reliability and Validity Tests 16

17 Triangulation were the approach to take to ensure that both validity and 18

reliability of the research findings to ascertain. Through triangulation different 19 sources of information and concepts/theories were adopted. 20

The use of multiple methods or diverse sources of information allowed the 21

study to address the research questions and cross-check information 22 exhaustively. Moreover, efforts were made to ensure that both validity and 23

reliability of the empirical data into consideration. 24

Cronbach's Alpha coefficient was used as a measure of internal 25

consistency as it provides a unique quantitative estimate of the internal 26 consistency of a scale (Mugenda, 2008). Cronchbach's alpha is a reliability 27

coefficient that indicates how well the items in a set are positively correlated to 28 one another. The study used the benchmark defined by Sekaran and Bougie 29 (2011) which states that reliabilities less than 0.60 are considered to be poor, 30

those in the 0.70 range, acceptable and those over 0.80 good (Sekaran & 31 Bougie, 2011). Thus, reliability of 0.70 and above was acceptable for the study. 32

33 34

Research Findings and Discussions 35 36 Descriptive Statistics 37

38 The first section reports the descriptive statistics of the study variables, the 39

second section reports and discusses regression results for economic variables 40 while the last section reports and discusses regression results for non-economic 41

variables. 42 A panel data regression model was used to examine the effects of 43

economic factors on international tourism demand at the national level 44 (specific objective one). A count data regression model using the survey data to 45 determine the effects of socio-demographic characteristics, political factors and 46

2019-3061-AJT

17

destination characteristics on international tourism demand at the individual 1 level (specific objective two, three and four). 2

3 Economic Factors 4

5 A total of 323 observations were used for analysis. The economic variables 6

considered were income measured by real GDP per capita, relative tourism 7 price, substitute price, travel cost and trade openness. 8

9 Table 2. Descriptive Statistics for Economic Variables 10

Variable Observations Mean Standard

Deviation

Minimum Maximum

Travel Cost 323 17,234.84 40,056.59 7.678017 280,669.7

Total Tourist

Departure

323 63,658.01 68,925.11 4,300 313,600

Holiday Tourists

Departures

323 49,971.86 57,302.68 3,800 258,100

Real GDP per capita 323 18,482.54 12,597.93 189.791 40,837.27

Relative tourism

price

323 5.280243 9.098876 0.004877 35.41114

Trade openness 323 0.0107475 0.0120338 0.0000397 0.0529732

Source: Study Data (2019) 11 12

Majority of the tourists who visited Ethiopia were holiday tourists (an 13 average of 49,972 tourists per 3 month). The average real GDP per capita was 14

18,482.54 US dollars, with a minimum of 189.791 US dollars and a maximum 15 of 40,837.27 US dollars. 16

This high variability (standard deviation of 12,597.93) was because the 17 data was from countries which were at different levels of development 18 (developed and developing countries). 19

There was also high variability for travel cost, which was caused by the 20 different exchange rates between the tourist origin country’s local currency and 21

the US dollar. The other variables did not show much variability. 22 23

Non-Economic Factors 24 25

The non-economic factors considered in the study were the tourist socio-26

demographic characteristics, political factors and destination characteristics. 27

The survey data used to analyze the influence of socio-demographic factors, 28 political factors and destination attractiveness factors on tourism demand was 29 collected from tourists leaving the country during the data collection period by 30 the use of questionnaires. 31

From the total questionnaires response rate the descriptive statistics for 32

tourism demand measured by the number of nights spent by tourist and it 33 shows that most of the tourists stayed in Ethiopia for only two weeks. There 34 was great variability in the number of tourist nights spent as reflected in the 35 minimum (1 night) and maximum (12 nights) values of the data. 36

2019-3061-AJT

18

Socio-Demographic Factors 1 2

In terms of sex, marital status, county of origin, age of respondents, annual 3 household income, level of education and occupation status data were 4

collected. Table 3, presents the descriptive statistics for the tourist’s socio-5 demographic characteristics. 6

7 Table 3. Tourists’ Socio-demographic Characteristics 8

Variable Description Percentage

Sex

Male 59

Female 41

Marital Status

Single 26

Married 54

Windowed 11

Divorced 9

County of Origin Europe 61

America 13

Asia 6

Africa 11

Australia 9

Family Size

No children 71

one child 16

Two children 7

Three children 3

More than three children 3

Age

18-24 78

25-34 12

35-46 4

47-66 2

Over 66 4

Source: Study Data (2019) 9 10 Table 3 shows that, tourists from Europe constituted the biggest percentage (61 11

percent), followed by America at 13 percent, Asia with 6 percent and Africa 12 with 11 percent and lastly Australia with 9 percent. These results implied that 13 the majority of the tourists who came to Ethiopia were males, married, aged 14

between 24 to 46 years and did not have children. 15

The tourist’s trip characteristics were given in terms of purpose of visit, 16 number of travel companions and whether it was a repeat visit. The study 17 results are presented in following Table 4. 18

19

20

2019-3061-AJT

19

Table 4. Trip Characteristics 1

Variable

Description Percentage

Purpose of visit

Holiday 71

Business 16

Visiting friends/ relatives 13

Travel companions

Alone 28

Spouse 12

Spouse & children 25

Colleagues 20

Friends 10

Children -

Relatives 5

Repeat visits

1st 76

2nd

19

3rd

5

4th

-

Source: Study Data (2019) 2 3 Political Factors 4

5 Five political factors were considered in the study namely political 6

stability, fear of terrorist’s attacks, travel warnings origin of travel government, 7 visa formalities, and good relations between tourist’s countries. Table 5 shows 8

their descriptive statistics. 9

10 Table 5. Political Factors 11

Factors Observations Mean Standard

Deviation

Political stability 323 2.67 1.02

Fear of terrorists attacks 323 2.70 .885

Travel warnings origin of travel

government

323 3.45 .918

Visa formalities 323 3.53 0.852

Good relations between tourist’s

countries

323 3.09 1.068

Source: Study Data (2019) 12 13

Table 5 shows the tourists were highly influenced by visa formalities 14 (mean =3.53) followed by government travel warning (mean = 3.45), good 15

relations between their countries (mean = 3.09), fear of terrorists attacks (2.70) 16 and lastly political stability (2.67) when making decision on whether to visit or 17

not. This shows that political factors influence tourist’s decision to visit 18 Ethiopia. 19

2019-3061-AJT

20

Destination Characteristics 1 2 Table 6 shows the descriptive statistics for destination characteristics 3

considered to influence tourist’s decision to visit Ethiopia. 4 5

Table 6: Destination Characteristics 6

Factors Observations Mean Standard

deviation

Information availability 323 2.14 0.612

Advertisement and

promotion

323 2.66 0.616

Availability of tour

packages

323 2.52 0.697

Inexpensive tourism product 323 2.42 0.570

Availability of direct flights 323 2.43 0.686

Nearness of destination 323 2.56 0.718

Scenery/wildlife/ diverse

culture

323 1.62 0.832

Museums 323 1.67 0.837

Fear of insecurity/crime 323 2.01 0.944

Availability of shopping

facilities

323 1.90 0.868

Good transport/

communication

323 2.55 0.712

Source: Study Data (2019) 7 8

Table 6 shows that each of the considered destination factors considered 9

influenced tourist decision to visit Ethiopia to some extent. 10 Availability of shopping facilities, advertisement and promotion of 11

Ethiopian tourism, fear of attack by disease/crime, museums, nearness of 12

destination, insecurity and crime level and availability of safe and good 13 transport/ communication facilities were the most important factors (had a 14

mean of 2.50 and above) the tourists considered when making a decision to 15 come to Ethiopia. 16

The other factors such as availability of direct flights, inexpensive tourism 17 product, and information availability, diverse culture, scenery and wildlife 18 were also found to influence the tourist decisions to visit to some extent. 19

20 Effect of Economic Factors on International Tourism Demand 21

22 The first objective of the study was to establish the effect of economic 23

factors on international tourism demand. The study employed a dynamic panel 24 regression model with the model parameters estimated by the SGMM 25 estimator. The study variables were first tested for stationary before regression 26

analysis was carried out and the Fisher-type panel regression unit root tests 27

2019-3061-AJT

21

(Choi, 2001) was used to investigate whether there were any variables in the 1 model that were non-stationary. 2

The null hypothesis being tested by Fisher- type tests were that all panels 3 contain a unit root against the alternative that at least one panel was stationary. 4

The Augmented Dickey-Fuller (ADF) option and the inverse normal Z statistic 5 were used to test for unit roots as recommended by Choi's (2001). 6

Further, the drift option was considered as the mean for all the study 7 variables for all countries was nonzero. A lag of one was used in the ADF 8 regressions performed to compute the test statistic since the study used annual 9

data. 10 The demean option was used to mitigate the impact of cross-sectional 11

dependence as suggested by Levin, Lin, and Chu (2002). A summary of the 12

Fisher panel unit root test results are presented in Table 7, the p-values are 13 given in parentheses. 14

15 Table 7. Panel Unit Roots Tests Results 16

Note: ** denotes rejection of null hypothesis at 5% significance level 17 Source: Study Data (2019) 18 19

From Table 7, considering the drift option, it can be observed that at 5 20 percent level the null hypothesis that all panels contain unit roots was rejected 21

for the variables total tourists departures; holiday tourists departures, relative 22 tourism price, tourism price, travel cost and trade openness, all measured in 23 natural logarithm form. 24

This means that these variables were stationary at levels. The implication 25 is that these variables were integrated of order zero. The natural logarithm of 26

the income variable (real GDP per capita) at level was not stationary but 27

Variable (in natural logarithms) Drift option Trend

option

Total tourist departures

-5.6872**

(0.0000)

0.4672

(0.6798)

Holiday departures -5.7103**

(0.0000)

0.4072

(0.6581)

Income -0.9867

(0.1619)

3.9479

(1.0000)

First difference of income -6.6045**

(0.0000)

-3.2165**

(0.0006)

Relative tourism price -5.1427**

(0.0000)

-1.0683

(0.1427)

Tourism price -5.1427**

(0.0000)

1.0683

(0.1427)

Travel cost -4.0449**

(0.0000)

-0.9923

(0.1605)

Trade openness -7.8890**

(0.0000)

-4.3816**

(0.0000)

2019-3061-AJT

22

became stationary after the first differencing implying it was integrated of 1 order one. 2

These results reflects those of a study carried out by Naude and Saayman 3 (2005) but contradicts results of a study by Chaiboonsri et al. (2010) who 4

found the study variables to be co-integrated of order one. 5 A panel dynamic regression model was used to determine the effect of 6

economic factors on international tourism demand (objective one). Two 7 regression analyses were carried out separately for the two dependent variables 8 considered (total tourist departures and holiday tourist departures). The total 9

tourist’s departures consisted of all those people who visit for holiday, business 10 or are on transit. 11

The analysis for holiday tourist was necessary since most of the 12

international tourists visiting for a holiday experience. The study used the 13 SGMM estimator to estimate the regression coefficients. The validity of the 14 results obtained in SGMM depends on the model statistical diagnostics; hence 15 the model diagnostics first were first carried out. 16

Compared to OLS model, SGMM does not assume normality and it allows 17

for heteroskedasticity in the data through use of robust standard errors. The 18 SGMM estimator assumes that the twice-lagged residuals are not auto 19 correlated and that the instruments used are exogenous. Specification testing in 20 the SGMM model thus involved testing for instruments exogenity and for 21

residual serial correlation. 22

23 Hypotheses Tests on Tourists Socio-Demographic Characteristics 24

25 A normal distribution is assumed in many statistical procedures such as 26

correlation, least squares regression and hypothesis testing (Sekaran & Bougie, 27

2011). When the response variable is the counted number of occurrences of an 28 event, this may not be true; hence there was need to test for normality first 29 before any further analysis was carried out on the survey data. 30

The Shapiro-Wilk‟s normality test was used to test whether the response 31 variable was normally distributed. The results of the test are presented in Table 32 8. 33

34

Table 8. Test of Normality for Survey Data 35

Dependent variable

Test statistic Degree of freedom P-value

Tourist- nights spent 0.854 304 0.000

Source: Study Data (2019) 36 37

The results from Table 8 show that the p-value of 0.000 is less than the 38

level of significance of 0.5; hence the null hypothesis that the data is normally 39 distributed was rejected. This implies that statistical procedures which assume 40 normality cannot be used as this would bias the results. 41

The test results agree with those for the study by Menezes et al. (2008) and 42 Chaiboonsri and Chaitip (2012) who found that the response variable was not 43

normally distributed. The number of tourist nights spent can differ by the 44

2019-3061-AJT

23

individual socio-demographic characteristics such as sex, marital status, annual 1 household income, level of education and age. It was therefore in order, to first 2 establish whether tourists night differ by the tourist socio-demographic 3 characteristics since all the various categories of the socio-demographic 4

variables could not be considered in the regression equation. 5 Since the data was not normally distributed the analysis of variance 6

(ANOVA) test could not be used. Instead alternative nonparametric tests, that 7 is, the Kruskal-Wallis and the Mann-Whitney U tests were used to examine 8 whether tourist nights differ significantly by the tourist socio-demographic 9

characteristics. 10 The null hypothesis being tested by both test statistics was that the mean 11

ranks of number of nights spent by respondents in each group of the variable 12

under consideration were equal against the alternative hypothesis that not all 13 the mean ranks were equal. The null hypothesis was rejected if the p-value was 14 less than 0.05. 15

In order to evaluate whether the number of nights spent differ by sex on 16 the average a Mann-Whitney U test was conducted and the results are 17

presented in Table 9. 18

19 Table 9: Test of Normality for Survey Data 20

Factor

Categories Mean Rank Test Statistic P-value

Gender Male 147.34 U = 10384.5

Z=-1.194

0.233

Female 159.50

Source: Study Data (2019) 21 22

Since the p-value of 0.233 is greater than 0.05 the null hypothesis that the 23

average number of nights spent was the same for both groups was not rejected. 24 This means that there was no significant difference between the number of 25

nights spent by male and female respondents. Hence, the gender of the tourist 26 does not influence the number of nights they spend in Ethiopia. To test whether 27

the number of nights spent by a tourist in Ethiopia differed by their marital 28 status, the Kruskal-Wallis test was used and the results are presented in Table 29 10. 30

31 Table 10. Test Results on Tourist Nights by Marital Status 32

Factor

Categories Mean Rank Test Statistic P-value

Marital

status

Single 143.56 X² 5.764 Degrees of

freedom =3

0.124

Married 154.67

Widowed 190.29

Divorced 77.50 Source: Study Data (2019) 33 34 35

2019-3061-AJT

24

The results in Table 10 indicate that the p-value of 0.124 was greater than 1 the level of significance of 0.05. Hence, the null hypothesis that the number of 2 tourist nights spent was the same for all groups was not rejected. 3

This implied that there was no significant difference in the number of 4

nights spent by tourists in whether single, married, divorced or widowed. 5 6

Regression Results on Non-Economic Factors 7 8

Count data regression analysis was carried out to establish the effect of 9

tourist socio-demographic characteristics, political factors and destination 10 characteristics on international tourism demand. Variance Inflation Factor 11 (VIF) measures how much the variance of the regression coefficients is inflated 12

by multicollinearity problems. 13 If VIF equals 0, there is no correlation between the independent measures. 14

A VIF measure of 1 is an indication of association between predictor variables, 15 but generally not enough to cause problems. Tolerance is the amount of 16 variance in an independent variable that is not explained by the other 17

independent variables. A tolerance value of 0.10 corresponding to a VIF of 10 18 is acceptable (Sekaran and Bougie, 2011). The collinearity statistics are given 19 in Table 11. 20

21 Table 11. Results for Collinearity Statistics 22

Explanatory Variables Collinearity Statistics

Tolerance VIF

Expenditure 0.680 1.472

Income 0.743 1.346

Age in years 0.613 1.632

Education dummy 0.890 1.123

Occupation dummy 0.591 1.691

Female dummy 0.893 1.120

Married dummy 0.506 1.975

Children dummy 0.540 1.852

Trip companion dummy 0.771 1.296

Repeat visit dummy 0.605 1.653

Political factor index 0.723 1.384

Destination characteristics index 0.565 1.768

Source. Study Data (2019) 23

24 Table 11 shows that the explanatory variables are not highly correlated 25

since none of them has a VIF of more than 10 and torelance value of less than 26

0.2 (Sekaran & Bougie, 2011). This implies that there is no sizeable 27 muliticollinearity that can affect regression. Poisson and negative binomial 28

regression models are designed to analyze count data. 29 Count models are estimated using maximum likelihood method. Choosing 30

between Poisson and negative binomial models depends on the nature of the 31 distribution of the dependent variable. Therefore, one should measure the 32

2019-3061-AJT

25

distribution of the study data before choosing between Poisson and negative 1 binomial regression. The researcher used the Pearson Chi‐Square 2 goodness‐of‐fit test to test the hypothesis that the dependent variable followed 3 a Poisson distribution. The regression model diagnostics results are presented 4

in table 12. 5 6

Table 12. Regression Model Diagnostic Results 7

Source: Study Data (2019) 8 9 The Pearson test statistic is significant at 5 percent level, thus the null 10

hypothesis that the data fits the Poisson distribution is rejected. In addition, 11 Poisson regression model assumes that the conditional mean and variance 12

should be equal, which is not the case (mean of 13.5 and variance of 101.25). 13 The model diagnostic results are in line with those diagnoses by Chaiboonsri 14 and Chaitip (2012). The negative binomial regression model was used to 15

determine the effect of noneconomic factors on international tourism demand 16

at the individual level. Table 13 presents the results of the negative binomial 17

regression model. 18 19

Table 13. Negative Binomial Regression Results 20 Variable Coefficient

IRR

Standard Error

Z-Statistic

P-value

Expenditure 0.0001094* 1.000109 0.0000293 3.74 0.000

Income -0.0000115* 0.999985 0.00270184 -10.52 0.000

Age 0.0071775** 1.007203 0.0031095 2.31 0.021

Education -0.2166887 0.8051806 0.249181 -0.870 0.387

Occupation 0.4847904* 1.623835 0.1095877 4.42 0.000

Female 0.0965974 1.101417 0.0808405 1.19 0.232

Married -0.146501 0.8637248 0.098315 -1.49 0.136

Children 0.1335843 1.142918 0.91677 1.46 0.145

Trip attribute -0.2196762*** 0.8027787 0.1199412 -1.83 0.067

Repeat visit -0.2073308** 0.8127507 0.093681 -2.21 0.027

Political factors -0.1459322** 0.86421163 0.0660987 -2.21 0.027

Destination

attractiveness

0.4389932*

1.551145 0.1378551 3.18 0.001

Constant 1.9282* - 0.512828 3.76 0.000

Waldi-chi square = 180.14

Degrees of freedom = 12

P- value = 0.0000

Number of observations 184

Note: * significant at 1% level; ** significant at 5% level and *** significant at 10% level. 21 Source: Study Data (2019) 22

Test Statistic Degrees of

Freedom

P-value

Pearson Chi-square (X² ) = 621.3348 172

0.0000

Likelihood-ratio test (of alpha = 0) = 202.37 1 0.0000

Alpha 0.1735217

Mean 13.51645

Variance 101.25391

2019-3061-AJT

26

1 The table shows that, the number of observations used in the count data 2

regression analysis were 184. Wald chi-square test statistic had a value of 3 180.14. The null hypothesis that all the estimated regression coefficients were 4

equal to zero was rejected at 5 percent level of significance since the p-value of 5 0.000 was less than 0.05. 6

This means that some of the regression coefficients were different from 7 zero. Positive coefficients indicate higher rate whereas negative coefficients 8 indicate lower rate. Rather than reporting the negative binomial regression 9

results using regression coefficients, the effect of the independent variable on 10 the dependent variable was reported in terms of the IRR. 11

12

Effect of Tourists Socio-Demographic Characteristics on International 13 Tourism Demand 14

15 The tourist socio-demographic characteristics included in the regression 16

model were expenditure, annual household income in US dollars, age in years, 17

education dummy with a value of 0 for tourists without college education and a 18 value of 1 for tourist with college education and above, occupation dummy 19 with a value of 0 for unemployed tourists and a value of 1 for those employed. 20

Female dummy with a value of 0 for a male and a value of 1 for a female 21

tourists, married dummy taking 0 value if the tourist was not married and value 22 1 if the tourist was married, children dummy taking value of 0 for tourists 23

without children and a value of 1 for those with children. Trip companion 24

dummy with a value of 0 for those travelling alone and value of 1 for those 25

travelling together with others and repeat visit dummy with a value of 0 for 26 those visiting the first time and a value of 1 for those visiting again. 27

The coefficient of the expenditure variable had the expected positive sign 28 and was significant at 5 percent level (p-value of 0.000 was less than 0.05). 29 The IRR value of 1.000109 for the expenditure variable suggested that the 30

number of nights spent increased by approximately 0.011 percent with every 31 one US dollar increase in expenditure. The annual household income variable 32 had a negative coefficient which is contrary to what was expected. This 33 coefficient was statistically significant at 5 percent level since the p-value of 34

0.000 was below 0.05. The IRR value of 0.99999 implied that if the annual 35 household income increased by one US dollar, the number of nights spent 36

would decrease by 0.001 percent. 37 Further investigation into the data to find out the cause of the unexpected 38

negative coefficient revealed that the respondents who had higher annual 39 household income spent fewer nights compared to those with lower annual 40 household income. The implication of this is that tourist with high annual 41

income may not be willing to stay for a longer period in Ethiopia but will 42 probably prefer to visit other destinations as well. 43

Age coefficient had a positive sign and was significant at 5 percent level 44 (p-value of 0.021 which is less than 0.05). The implication was that when the 45

2019-3061-AJT

27

tourist’s age increased by one year, the number of nights spent increased by 1 0.72 percent. 2

This means that older tourists spent more nights in Ethiopia compared to 3 the younger ones. Occupation dummy coefficient had a positive sign and was 4

significant at 5 percent level since the p-value of 0.000 was less than 0.05. This 5 meant that the respondents who were employed spent 1.62 times the incident 6 rate for those not employed. This implies that the respondents who were 7 employed spent more nights in Ethiopia compared to those who were not 8 working. The coefficient of the trip companion dummy was negative and not 9

significant at 5 percent but was significant at 10 percent level (p-value of 0.067 10 was less than 0.1 but greater than 0.5) with an IRR value of 0.803. 11

This implies that those respondents who travelled accompanied by others 12

spent 0.803 times the number of nights less than those who travelled alone. 13 This means that those who travelled alone spent more nights in Ethiopia than 14 those travelling with others or those travelling in groups. 15

The coefficient of repeat visit dummy had a negative sign and was 16 significant at 5 percent level of significance (p-value of 0.27 was less than 0.5). 17

The IRR value for this variable was 0.813 implying that those who had 18 visited Ethiopia before spent 0.813 times the number of nights less compared 19 to those who had not visited before. This means that the tourists who were on 20 their first visit to Ethiopia spent more nights than those who were on a repeat 21

visit. 22 The coefficient for the level of education dummy was negative but was not 23

significant at 5 percent level. The insignificance could have arisen due to the 24

reduction of categories in the regression equation. 25

The reduction of categories was necessary in order to reduce the number of 26 variables in the regression equation to avoid losing many degrees of freedom 27

because of too many dummy variables being included. The coefficient of the 28 female dummy was negative but not significant at 5 percent level implying that 29 gender is not a major determinant of international tourism demand. The 30

coefficient for the married dummy variable was negative but was not 31 significant at 5 percent level. This means that marital status was not a 32 significant factor influencing tourism demand. The coefficient for the children 33 dummy was also not significant at 5 percent level of significance and had a 34

positive sign. The coefficients for education, gender and marital status 35 dummies had the same signs and still in this study they were not significant. 36

37 Effect of Political Factors on International Tourism Demand 38 39

For regression analysis purposes, the political factors composite index was 40 constructed by calculating the summed scores per respondent and then dividing 41

it by the number of items. The political composite index had the expected 42 negative coefficient and was significant at 5 percent level since the p-value of 43 0.027 was less than 0.05, with IRR value was 0.864. 44

This means that if the political factors index decreased by one unit, the 45 number of nights spent in Ethiopia will decrease by 13.6 percent, that is, if the 46

2019-3061-AJT

28

political situation in the country deteriorates tourism demand will decrease as 1 not many people will be willing to come to Ethiopia. 2

This is in line with the political instability to be an important determinant 3 of international tourism demand affecting demand negatively. 4

5 Effect of Destination Characteristics on International Tourism Demand 6

7 For regression analysis purposes, the destination characteristics composite 8

index was constructed by calculating the summed scores per respondent and 9

then dividing it by the number of items. The coefficient for destination 10 characteristics index was significant at 5 percent level (p-value of 0.01 was 11 lower than 0.05. 12

The coefficient was positive as expected and the IRR value was 1.55 13 implying that if the destination characteristics index increased by one unit, the 14 number of nights spent would increase by 55 percent. 15

16 17

Conclusion and Recommendations 18 19 Conclusion 20 21

The aim of the study was to estimate the international tourism demand in 22 Ethiopia with respect to 11 main tourism suppliers. The study applied the 23

single equation methodology in estimating the international tourism demand 24

model for tourists from different origins. It can be concluded from the study 25

that international tourism demand is price inelastic in the short run but price 26 elastic in the long run. 27

Elasticity of cost of travel was found to be both inelastic in the short run 28 and long run. If the price of tourism and the cost of travelling increase, the 29 international tourism demand is expected to decrease. The lagged dependent 30

variable (word of mouth effect) was significant implying that international 31 tourism demand is influenced by tourists report to others about their holiday 32 experience. Tourists are sensitive to political instability; hence the last two 33 years ethnic tension affected international tourism demand in Ethiopia 34

negatively. Bilateral trade was an important determinant of international 35 tourism demand, hence good relations between Ethiopia and tourist generating 36

countries is expected to enhance tourism demand. 37 The tourist’s socio-demographic characteristics (that is, expenditure, 38

annual household income, age, occupation, travelling companions and repeat 39 visits) were found to have an effect on international tourism demand. 40

The study finding suggested that those with higher annual household 41

income spent fewer nights than those with lower annual household income. It 42 implying that those with high income prefer short visits and then probably 43 move on to other destinations. The findings further showed that the tourists 44 who were employed (either part time or full time) spent more nights compared 45

2019-3061-AJT

29

to those who were not working. The older tourists were found to spend more 1 nights compared to the younger ones. 2

Those visiting the country for the first visit time spent more nights than 3 those who were on a repeat visit. The tourists travelling alone were found to 4

spend more nights than those travelling with others or those travelling in 5 groups. Political factors were found to affect demand negatively while 6 destination factors were found to influence demand positively. 7

8 Recommendations 9

10 Based on the study findings, the following are some of policy implications 11

and strategies that can be proposed for the development of tourism industry in 12

terms of sustaining the growth of tourist’s arrivals and attracting more tourists 13 into the country. 14

Since the study found that the word of mouth effect was significant in 15 explaining international tourism demand, the tourism industry should embark 16 on offering high quality services to enhance the country’s image in order to 17

attract new and repeat tourists. 18

The government together with the ministry of culture and tourism should 19 engage in sustainable tourism development to avoid tourism products 20

degradation. 21

Tourism price was also an important determinant of international tourism 22 demand, thus the government and tourism key actors/players in the tourism 23 sector (government, the tour operators, travel agents and car rental 24

companies, hoteliers among others) should aim at making the tourism price 25 competitive compared to other countries in Africa to avoid most of the 26 competition from neighboring countries. 27

The Government in combination with Ethiopian Airline should invest in 28 reasonable cost of air transportation operating within and outside the 29

country. 30

The government should come with favorable policies that will encourage 31 investors to invest in the tourism sector. Travel agents should ensure the 32

transportation costs within the country are competitive and should not 33 exploit visitors. 34

International tourism demand was found also to be affected by tourist socio-35 demographic characteristics. Therefore, the government should facilitate in 36 the design of new products catering for different age groups, professionals 37

and people of different socio-economic status. 38

In addition, the existing tourism products, facilities and services should be 39 improved in order to encourage repeat visits as well as attract new tourists. 40

The government should continue to engage in bilateral trade with more 41

countries as it is currently doing. 42

Furthermore, since political factors and destination characteristics were also 43 found to be significant determinants of demand, the government should 44

enhance political stability in the country. 45

2019-3061-AJT

30

All the tourism sectors stakeholders and other related industries as well as 1 every citizen, should work towards creating a positive image through 2

favorable friendly and peaceful environment. 3

Future research could build on the results of this study to enrich the existing 4 knowledge of determinants on international tourism demand. 5

6 7 Reference 8 9 Akama, J. S. (2000). The efficacy of tourism as a tool for economic development. In 10

A. Bujra (Ed), Development policy management in Sub-Saharan Africa: Tourism 11 and African Development - Trends and Critical Issues DPMN Bulletin, VII (1). 12 Addis Ababa: Commercial Printing Enterprise. 13

Baltagi, B. H. (2005). Econometric analysis of panel data (3rd ed.) Sussex: John 14 Wiley & Sons, ltd. 15

Baum, F. B. (2006). An Introduction to modern econometrics using stata. Texas: Stata 16 Press. 17

Blundell, R., & Bond. S. (1998). Initial conditions and moment restrictions in dynamic 18 panel data models. Journal of Econometrics, 87, 115–143. 19 Bond, S. R. (2002). Dynamic panel data models: A guide to micro data methods and 20

practice. Centre for microdata methods and practice working paper CWP09/02, 21 London. Institute for Fiscal Studies. Available at http:// cemmap. ifs. org. uk/ 22 wps/cwp0209. 23

Burger, C. J. S. C., Dhnal, M., Kathrada, M., & Law, R. (2001). A practitioners guide 24 to time series methods for tourism demand forecasting-A case study of Durban, 25 South Africa. Tourism Management, 22, 403-409. 26

Burkart, A. & Medlik, S. (2nd Eds). (1981). Tourism: Past, Present and Future. 27 London: Heinemann. 28

Cakici, A. C. & Harman, S. (2007). Importance of destination attributes affecting 29 destination choice of Turkish birdwatchers. Journal of Commerce and Tourism 30 Education Faculty, 1. 31

Cespedes, D. V., Gomez, E. G. & Becerra, A. T. (2012). International demand for 32 nature-based tourism in Costa Rica: Socio-demographic and travel indicators. 33 Tourismos: An international multidisciplinary journal of tourism, 7 (1), 269-287. 34

Chaiboonsri, C. & Chaitip, P. (2012). The Modelling International Tourism Demand 35 for Length of Stay in India: Social Development, Economics Development. 36 Journal of Knowledge Management, Economics and Information Technology, II 37 (2). 38

Chaiboonsri, C., Sriboonjit, J., Sriwichailamphan, T., Chaitip, P., & Scriboonchitta, S. 39 (2010). A panel cointegration analysis: An application to international tourism 40 demand of Thailand. Annals of the University of Petrosani, Economics, 10 (3), 41 69-86. 42

Choi, I. (2001). Unit root tests for panel data. Journal of International Money and 43 Finance, 20, 249–272. 44

Collis, J., & Hussey, R. (2009). Business Research (3rd Ed.). Palgrave, UK: 45 Macmillan 46

Cooper, C., Fletcher, J., Fyall, A., Gilbert, D. & Wanhill, S. (4th Ed.). (2008). 47 Tourism: Principles and Practice. United Kingdom: FT Prentice Hall. 48

Cochran (1963:75) ). The study of research: A review of findings. Journal of Travel 49 Research, 33, 12-23. 50

2019-3061-AJT

31

Christie, I. T., & Crompton, D. E. (2001). Tourism in Africa. Africa Region Working 1 Paper Series, 12, World Bank, Washington, DC. 2

Crouch, G. (1994a). The study of international tourism demand: A survey of practice. 3 Journal of Travel Research, 32, 41-55. 4

Crouch, G. (1994b). The study of international tourism demand: A review of findings. 5 Journal of Travel Research, 33, 12-23. 6

Crouch, G. (1995). A meta-analysis of tourism demand. Annals of Tourism Research, 7 22, 103-118. 8

Dann, G. (1977). Anomie, ego-enhancement and tourism. Annals of Tourism 9 Research, 4, 184-194. 10

Dann, G. M. S. (1981). Tourism motivation: An Appraisal. Annals of Tourism 11 Research, 8, 187-219. 12

Divisekera, S. (2003). A model of demand for international tourism. Annals of 13 Tourism Research, 30 (1), 31-49. 14

Eilat, Y. & Einav, L. (2004). Determinants of international tourism: a three 15 dimensional panel data analysis. Applied Economics, 36, 1315-1327. 16

Gallup, J. L., Sachs, J. D., & Mellinger, A. (1999). Geography and economic 17 development, CID Working paper No. 1, Harvard University, Cambridge, MA. 18

Gujarati, D. N., & Porter, D.C.(5th Ed.). (2009). Basic Econometrics: McGraw-Hill 19 Habibi, F. & Abbasinejad, H. (2011). Dynamic panel dataa of European tourism 20

demand in Malaysia. Iranian Economic Review, 15 (29). 21 Hair, F., Anderson, E. & Tatham, L. (5th Ed.) (1998). Multivariant data analysis. New 22

Jersey: Prentice Hall. 23 Hansen, L. P. (1982). Large sample properties of generalized method of moments 24

estimators. Econometrica, 50, 1029-1054. 25 Holloway, J. C., Humphreys, C. & Davidson, R. (8th Ed.). (2009). The Business of 26

tourism. United Kingdom: FT Prentice Hall. 27 Horner and Swarbrooke. (2001). Business travel and tourism. MA: Butterworth-28

Heinemann 29 Ibrahim, M. A. (2011). The determinants of international tourism demand for Egypt: 30

Panel data evidence. European Journal of Economies, Finance and 31 Administration Sciences, (30). 32

Idowu, K. O. & Bello, A.K. (2010). What are the factors determining tourists 33 destinations in Africa? Parkistan Journal of Social Sciences, 7, 301-303. 34

Jonsson, C. & Devonish, D. (2008). Does nationality, gender, and age affect travel 35 motivation? A case of visitors to the Caribbean island of Barbados. Journal of 36 Travel & Tourism Marketing, 25, 3-4. 37

Kaufmann, D., Kraay, A., & Zoido-Lobaton, P. (1999). Aggregating governance 38 indicators, policy research department working paper No. 2195. World Bank, 39 Washington, DC. 40

Koutsoyians, A. (1991). Modern Microeconomics (2nd Ed.). London: Macmillan 41 Education Ltd. 42

Levin, D. M. (1988). The opening of vision: Nihilism and the postmodern situation. 43 London: Routledge. 44

Lim, C. (1997). Review of international tourism demand models. Annals of Tourism 45 Research, 24 (4), 835-849. 46

Lim, C. (1999). A meta-analytic review of international tourism demand. Journal of 47 Travel Research, 37, 407-428. 48

Lim, C. (2004). The major determinants of Korean outbound travel to Australia. 49 Mathematics and Computers in Simulation, 64, 477-485. 50

2019-3061-AJT

32

Li, G., Song, H. & Witt, S. F. (2005). Recent developments in econometric modelling 1 and forecasting. Journal of Travel Research, 44, 82-89. 2

Lumsdon, L. (1997). Tourism marketing. London: International Thomson Business 3 Press. 4

McMillan, J. H., & Schumacher, S. (2nd Ed.). (1989). Research in Education: A 5 conceptual introduction. London: Scott, Foresman & company. 6

Maslow, A. H. (1943). A theory of human behaviour. Psychological Review, 50. 370-7 396. 8

Mathieson, A. & Wall, G. (1982). Tourism: Economic, Physical and Social impacts. 9 London: Longman. 10

May, T. (1997). Social research: Issues, Methods and Process (2nd Ed.). Trowbridge: 11 Redwood Books. 12

Mello, D. M., Pack, A., & Sinclair, M.T. (2002). A system of equations model of UK 13 tourism demand in neighbouring countries. Applied Economic, 34, 509-521. 14

Menezes, A. G; Moniz, A. & Vieira, J.C. (2008). The determinants of length of stay of 15 tourists in the Azores. Tourism Economics, 14 (1), 205–222. 16

Mitchell, J., & Ashley, C. (2006). Can tourism help reduce poverty in Africa. 17 Overseas Development Institute, London. 18

Morley, C. L. (1991). Modelling tourism demand: model specification and structure. 19 Journal of Travel Research, 30, 40-44. 20

Moscardo, G. M., & P. L. Pearce (1986). Historical Theme Parks: An Australian 21 experience in authenticity. Annals of Tourism Research, 13 (3), 467‐79.4 22

Muchapondwa, E., & Pimhidzai, O. (2011). Modeling international tourism demand 23 for Zimbabwe. International Journal of Business and Social Science, 2, 71-81. 24

Mugenda, A. G. (2008). Social Science Research. Nairobi: Applied research and 25 training services. 26

Mugenda, O. M. & Mugenda A. G. (2003). Research Methods: Quantitative and 27 Qualitative Approaches. Nairobi: African Centre for Technology Studies. 28

Munoz, T. G. (2006). Inbound international tourism to Canary Islands: A dynamic 29 panel data model. Tourism Management, 27, 281-191. 30

Munoz, T. G. (2007). German demand for tourism in Spain. Tourism Management, 31 28, 12-22. 32

Moutinho, L., 1987. Consumer behaviour in tourism. European Journal of Marketing, 33 21 (10), 5-44. 34

Nagy,S., & Biber, H. (2010). Mixed Methods Research: Merging theory with practice. 35 New York: The Guilford Press. 36

Naude, W., & Saayman, A. (2005). Determinants of tourist arrivals in Africa: A panel 37 data regression analysis. Tourism Economics, 11 (3), 365-391. 38

Nordstrom, J. (2002). Dynamic and stochastic structures in tourism demand 39 modelling. Umea Economic Studies, (596). 40

Odunga, P. O. (2005). Choices of Attractions, Expenditure and Satisfaction of 41 International Tourists. Unpublished doctoral thesis, Wageningen University. 42

Ondicho, T. G. (2016). International tourism: Development, problems and challenges. 43 Eastern Africa Social Research Review, 16(2), 49–70. 44

Ondicho, T. G. (2003). Tourism Development: Constraints and Opportunities. The 45 African anthropologist, 10, (1). 46

Page, S. J and Connell, J. (2006). Tourism: A Modern Synthesis (2nd Ed.). London: 47 Thomson Learning. 48

Pearce, D. (1995). Tourism Today: A Geographical Analysis (2nd Ed.). Harlow: 49 Longman. 50

2019-3061-AJT

33

Pearce, P. L. (1988). The Ulysses factor: evaluating visitors in tourist settings. 1 NewYork: Springer-Verlag. 2

Plog, S.C. (1974). Why destination areas rise and fall in popularity. Cornell hotel and 3 restraint quarterly. 14(4), 55-58. 4

Research, A. (2006). Sample size table. http://data.research.org/sample/. 5 Roodman, D.(2009). How to do xtabond2: An introduction to difference and system 6

GMM in Stata. The Stata Journal, 9 (1), 86-136 7 Rogerson, C. M. (2007). Reviewing Africa in the global tourism economy. 8

Development Southern Africa, 24 (3), 361-379. 9 Sargan, J. D. (1958). The estimation of economic relationships using instrumental 10

variables. Econometrica, 26, 393-415. 11 Salleh, N. H. M., Hook, L. S., Ramachandran, S., Shuib, A., & Noor, Z. M. (2008). 12

Asian tourism demand for Malaysia: A bound test approach. Contemporary 13 Management Research, 4 (4), 351-368. 14

Saunders, M., Lewis, P. & Thornhill, A. (2007). Research methods for business 15 students (4th Ed.). London: Prentice Hall. 16

Sekaran, U. & Bougie, R.. (2011). Research methods for business:A skill building 17 approach (5th Ed.). India: Aggarwal printing press. 18

Schiffman, L. G. & Kanuk, L. L. (1997). Consumer behaviour. London: Prentice Hall. 19 Schmoll, G.A. (1977). Tourism promotion. London: Tourism International Press. 20 Sindiga, I. (1999). Tourism and African development: Change and challenge of 21

tourism. Nairobi: Ashgate publishers. 22 Sindiga, I. (1996). Domestic Tourism. Annals of Tourism Research, 23 (1), 19-31. 23 Smeral, E. (1988). Tourism demand, economic theory and econometrics: An 24

integrated approach. Journal of Travel Research, 26, 40. 25 Song, H., & Li, G. (2008). Tourism demand modeling and forecasting: A review of 26

recent research. Tourism Management, 29, 203-220. 27 Song, H. & Witt, S. F. (2000). Tourism demand modelling and forecasting: Modern 28

econometric approaches. Cambridge: Pergamon. 29 Summary, R. (1987). Estimation of tourism demand by multivariable regression 30

analysis. Tourism Management, 8, 317-322. 31 Uysal, M. & Crompton, J.L. (1985). An Overview of the approaches used to forecast 32

tourism demand. Journal of Travel Research, 23, 7-15. 33 Valle, E. & Yobesia, M. N. (2009). Economic contribution of tourism. Tourism 34

Analysis, 14, 401-414. 35 Witt, S. F. & Witt, C. A. (1995). Forecasting tourism demand: A review of empirical 36

research. International Journal of Forecasting, 11 (3), 447-475. 37 World Bank, (2016). http://data.worldbank.org/indicator/. 38 World Bank. (2006). Ethiopia: Towards a strategy for pro-poor tourism. Africa Private 39

Sector Development Note 24. Washington DC: World Bank. 40 World Tourism Organization (2017). International recommendations for tourism 41

statistics 2008. New York: UNWTO. 42 World Tourism Organization (2010). International recommendations for tourism 43

statistics 2008. New York: UNWTO. 44 World Tourism Organization(2016). Tourism highlights. Madrid: UNWTO. 45 World Tourism Organization (2012). Methodological notes to the tourism statistics 46

database. Madrid, Spain: UNWTO. 47 World Travel &Tourism Council (2018). Travel and tourism economic impact. 48

London: World Travel and Tourism council. 49 World Travel & Tourism Council (2012). Travel and tourism economic impact, 50

World. London: World travel and Tourism council. 51

2019-3061-AJT

34

Xiao, H., & Smith, S. (2016). Case studies in tourism research: A state of-the-art 1 analysis, Tourism Management, 27, 738–749. 2

Yamane (1967:886) Sample size table. http://data.research.org/sample/. 3 Zikmund, G. W. & Babin, J. B. (9th Ed.). (2007). Exploring market research. South-4

Western, USA: Mason, Thomson. 5 6