Studies on Interdisciplinary Economics and Business - Volume III

← 16 | 17 →

Mustafa Kırca & Mustafa Özer

Structural Breaks and Cointegration Analysis of the Relationship between International Tourism Demand and Sectoral Incomes in Turkey¹

1 Introduction

According to the Unbalanced Growth Theory developed by Hirschman (1958), there are forward and backward linkages among the sectors of an economy. A backward link indicates the sector’s relative importance as user input from non-primary economic activities, while forward links point out sector’s relative importance as a supplier to the other sectors in the economy. The existence of forward and backward linkages also proves the existence of an input-output relations among economic activities or sectors. Well known contributions of the tourism sector to overall economy are the creation of output and income, foreign exchange earnings, increasing government revenues, creation of employment and business opportunities, inducing of investment in infrastructure and contribution to regional development.

← 17 | 18 →

Also, it is widely argued that the contribution of the services sector, particularly tourism and hospitality, to economic revitalization by creating more labor income is considered greater than even that of agricultural and manufacturing industries. Creating more impacts on value-added, employment entails and output, the economic impact of the tourism and hospitality sector greatly outweigh those of agricultural- or manufacturing-related industry (Kim and Kim, 2015). The tourism sector has close input-output linkages with related sectors such as accommodation management, travel management, food-beverages management as well as other sectors such as agriculture and industry. Furthermore, any positive contributions of tourism incomes for a region or country’s economy also impact the industry, agriculture and services sectors. Also, an increase in tourism expenditure may lead to additional activity in related industries and causing a greater rise in spending than an initial increase in spending. Thus, as (Cernat and Gourdon, 2012) indicated, one of best ways to indicate economic benefits is to integrate tourism into the national economy by establishing strong linkages between tourism and other economic sectors such as agriculture, fisheries, manufacturing, construction and other service industries, such as culture, arts, entertainment, recreation, and sport. We know that when the tourism sector makes use of products and services produced within the local economy or sectors of the economy, it will contribute to strengthen of these other sectors and provide additional income through the multiplier effect of the tourism sector on the overall economy. In addition to having interaction with other sectors through its input-output relations, the tourism sector creates also a multiplier effect. According to this effect, an increase in the tourism sector’s income will increase the incomes of other sectors. The effects of the incomes generated by the tourism sector are generally divided into three main groups: direct, indirect and (additionally) adapted effects (Coltman, 1989; Vellas and Bécherel, 1995; Page, 2009; Dwyer, Forsyth and Dwyer, 2010).

The direct effects mentioned herein indicates the expenses tourists made by directly buying goods and services from businesses such as hotels, restaurants and transportation services. Indirect effects are effects that arise due to tourism businesses, particularly hotels, purchasing goods and services from local suppliers. Finally, adapted effects are those that are generated through the expenses made for goods and services bought by the personnel employed in the tourism businesses with their salaries. Therefore, a one-unit increase in income generated by the tourism sector has the capacity to generate more than a one-unit increase in incomes of other sectors. Thus, rise in income of the tourism sector is expected to increase the incomes of other sectors. But, this expectation should be supported by empirical studies. Also, the contribution and impacts of tourism to a country’s economy are generally by examining its effect on the economic growth of the country. According to most of these studies, international tourism has a positive effect on economic growth and these positive effects of tourism on growth are multiple. As Blake et al. (2006) concludes, the tourism plays a crucial role in rising income and human capital, and in fostering efficiency and competitiveness. But, it is hard to see rich literature particularly investigating sectors’ relationships with the incomes of other sectors, even though as the tourism industry is growing in many countries, the causal relationship between international tourism demand and incomes of various sectors is becoming important for policy makers.

← 18 | 19 →

Thus, this section aims to investigate the relationships between international tourism demand and the selected sectoral incomes in Turkey by using quarterly data between the first quarter of 1998 and the last quarter of 2013. We use Kapetanios (2005) multiple break unit root test, Maki (2012) multiple break cointegration test, and cointegration parameter estimation methods.

2 Literature Review

There are many studies that investigate the relationships between the tourism sector and other sectors. Gül (2013) conducted a study by using input-output analysis methodology for the purpose of investigating the effects of increases in tourism income in Turkey. By generating a “social accounts matrix”, intersectoral income multipliers were derived. The empirical findings of the study indicate that the demand-oriented shocks in the tourism sector may be utilized for revitalizing the economy and handling unemployment issues.

Kadiyali and Kosova (2013) conducted a study on the distribution of employment that is provided by the development of tourism for other sectors in the USA. Data pertaining to 43 states of the USA covering the period between the years 1987 and 2006 were utilized in the study, and the researchers reached the conclusion in the result of the study that 100 rooms booked daily cause around 2–5 new employment opportunities in other sectors. It was underlined that this situation is reflected most in the sectors of construction, retail, health, and vocational and technical services.

Kweka et al. (2001) investigated whether or not there is a relationship between tourism and other sectors in Tanzania, focusing on employment and tax incomes by using Input-Output Analysis. They reached the conclusion that tourism had an impact on the production of sectors due to its robust link with other sectors. However, it was mentioned that the income impact brought by tourism is found out to be not significant. The reason behind that was explained as low added-value products being produced in the tourism sector. Despite this situation, it was indicated that tourism is the sector that has the potential to make a positive impact on economic growth.

Telfer and Wall (1996) conducted a study that investigated the relationships between tourism and food production in Lombok Island in Indonesia. Authors revealed that these two sectors benefited from each other. They highlighted particularly the link between the agriculture sector and the tourism sector. It was observed that there were potentials towards increasing the backward links between tourism and local food production. They mentioned that local production would increase with the development of tourism demand, and thus local production would develop in parallel to this. However, they also pointed out that there are challenges required to be overcome in order to strengthen such links.

← 19 | 20 →

A study that explores the correlation between the tourism sector and agriculture sector was conducted by Torres (2003). Torres interviewed many groups such as tourists, tourism sector employees, farmers and migrants at important tourism destinations of Mexico. Through such interviews, the researcher tried to perceive the link between agriculture and tourism. In the result of the study, it was indicated that the intertwinement of the two sectors caused an increase in local production and thus tourism incomes would also increase. With the development of tourism, there are certain benefits that rural communities may receive from tourism. At the same time, it was pointed out that different agricultural products are produced due to tourists being willing to try different products.

Çıkın et al. (2009) conducted a study that investigated the importance of the tourism sector within Turkey’s economy and the effect of tourism on other sectors. In the light of the analyzes conducted, the authors mentioned that the development of the agricultural sector is linked with the development of the tourism sector. According to the research, it was indicated that agriculture and tourism were sectors that are in constant and close interaction and that they complemented each other. Furthermore, they expressed that most of the inputs to tourism are provided from the agriculture sector due to tourism being a sector of food and beverages. Another important finding from the study was the observation that the development of agricultural tourism in response to tourism meant also the development of both the tourism and agriculture sectors. In addition, it was mentioned in the study that Turkey possessed many natural, seasonal and sociocultural superiorities compared to many competing countries in global tourism markets. It was highlighted at the conclusion of the study that due to such superiorities it was possible for the country to receive a larger share of the tourism sector market throughout the world and for an integrated growth to be realized in the agriculture sector that entails intensive agricultural activities in rural regions. Other studies that explored the links between agriculture, food production, and tourism sector were conducted by Bélisle (1983) and Rogerson (2012). In the study conducted by Bélisle (1983) specific to the Caribbean region, it was expressed that tourism demand boosted food import and thus decreased the positive economic effect of the tourism sector. Furthermore, Bélisle expressed that the links between tourism and local food production were not defined in full. On the other hand, Rogerson (2012) in his study highlighted that the links between the tourism sector and agriculture sector are of significance for maximizing the economic effects of tourism sector particularly in developing countries.

← 20 | 21 →

Khadaroo and Seetanah (2007) conducted a study that investigated the correlations between the development level of transportation infrastructure and tourism. In this study realized specific to Mauritius island, overall tourism demand was modeled. In conclusion of the study that utilized dynamic panel data analysis, it was observed that the tourists visiting from Europe/America and Asia were sensitive against the transportation infrastructure of the island. It was also revealed through tourism demand equation that the tourism demand is dependent on tourist’s income, distance and price. These results indicate that the tourism sector is correlated with the development of transportation infrastructure. Infrastructure works affect many sectors.

Literature review shows that there is a lack of studies examining the sectoral linkages of tourism in terms of its effect on other sector incomes. Thus, we study its effects on the selected sectoral incomes in Turkey.

3 Methodology

To investigate the dynamic linkages between international tourism demand and sectoral incomes, we first examine the time-series properties of variables by using the multiple breaks unit root test developed by Kapetanios (2005) due to several structural breaks observed in the series over the study period. And then, we use the multiple breaks cointegration test developed by Maki (2012). Finally, we estimate the coefficients of cointegrating regression.

According to Perron (1989), in the presence of structural break(s) in variables, the traditional unit root tests2 produce results subject to criticism. Thus, in this case, the structural breaks unit root tests should be used to determine the degree of integration of variables. There are many tests that take into consideration of structural breaks.³ However, every test possesses both advantages and disadvantages. Kapetanios (2005) developed multiple breaks unit root test eliminating the drawbacks of these tests. To carry out the Kapetanios (2005) tests, we consider the following regressions⁴:

← 21 | 22 →

and

Where DU and DT are dummy variables that represent breaks for the constant term and on the trend respectively; μ₀ indicates the constant, μ₁ indicates the coefficient of the trend, φ indicates the differential constant term, ψ indicates the differential coefficient of the trend, and ɛ_t indicates the error term (disturbance). Similar to many structural breaks time series analyzes, DU and DT are defined as follows; Where Tb_,i indicates break date and “i” takes on values as i=1,2,…m.

As Kapetanios (2005:127) also expressed clearly, the method developed by Bai and Perron (1998) is used when determining break dates in this test. Furthermore, the number of breaks is also decided through this method. In order to carry out the test, following null and alternative hypotheses are tested: H₀: Variable has unit root (α=1, μ₁= φ₁=…= φ_m= ψ₁=…= ψ_m=0) and H₁: m is stationary with the breaking (α<1, φ_i+1=ψi+₁=…= ψ_m=0, (i= 1,…,m)).

Kapetanios (2005:129) calculated critical values referring to 5 breaks by using Monte Carlo simulation method. In order to determine whether or not variables are stationary with the structural break, the calculated values of the test statistic and the table values are compared: when the sample value of the test statistic (t statistic value) is greater than the table value, H₀ is rejected and it is concluded that the variable Y is stationary with “m” number of breaks.

After determining the degree of integration of the variables, we determine the existence of cointegration by using multiple breaks cointegration test developed by Maki (2012), since in the presence of structural breaks, traditional cointegration tests5 are most likely produce biased results. Even though there are other cointegration tests take into account of structural break, such as Gregory and Hansen (1996a, 1996b) and Hatemi-J (2008), we preferred Maki test due to the fact that better results will be generated in cases where the number of breaks is not known and that it allows for four different model estimations. These models are as shown Maki’s (2012:2011–2012) article. In the article Model 0, 1, 2, and 3 represents Equation 1, 2, 3, and 4, respectively.

← 22 | 23 →

Maki (2012:2012) determines both the number of breaks and break dates by using Bai and Perron (1998) test. The only difference is that both the number of breaks and break dates of the residual terms (ǔ_t) are determined based on the following equation:

The residual terms in Equation (3) obtained from Model 0,..,3. For the first breakpoint, “t” statistics belonging to the hypothesis of H₀: ρ=0 (H₁: ρ<0) for all possible break dates are obtained and the minimum t statistic is recorded as τ¹ and break date is determined based on minimum RSS. The hypotheses belonging to this test are as follows; H₀: ρ=0, there is no cointegration between variables with structural breaks and H₁: ρ<0, there is cointegration between variables with structural breaks.

Maki (2012:2013) calculated the critical values for each model we explain above with five breaks. When the calculated value of test statistic is greater than the critical value, H₀ is rejected and it is concluded that there is a cointegration between variables with structural breaks.

After determining the existence of cointegration, we use three fully efficient estimation methods of cointegrating regression of Fully Modified Least Squares (FMOLS) estimator developed by Phillips and Hansen (1990), Canonical Cointegrating Regression (CCR) estimator developed by Park (1992), and finally Dynamic Least Square (DOLS) estimator developed by Stock and Watson (1993). These three estimators are known as fully efficient estimators and allow us to estimate the coefficients of cointegrating regression and provides unbiased estimates of the coefficients.

4 Data and Empirical Results

← 23 | 24 →

To investigate the impacts of international tourism demand on sectoral incomes by using the quarterly time series between 1998 and 2013. We extract the sectoral income series from the Central Bank of Republic of Turkey (CBTR) (2014) Electronic Data Delivery System (EVDS) system and number of foreign tourist as a measure of international tourism demand (LNYZ) from Turkish Statistics Institute (2014) (TurkSTAT). All income series are expressed in constant prices of 1998 and logarithmic values of variables are used in the analysis. Sectors that included in the study are Culture, arts, entertainment, recreation and sports activities (LNKSEDS), Agriculture, forestry and aquaculture (LNTOB), Real estate activities (LNGF), Accommodation and Food Service Activities (LNKHY) and Activities of households as employers (LNHIV). Figure 1 represents the time series plots of variables.

The time series plots of variables show that all variables exhibit strong seasonality and structural breaks. Thus, to focus on other components of time series, we remove the seasonal components of variables by using Census X12 method and use seasonally adjusted variables in our empirical analysis.

In this section of the study, first, we present the results of unit root and cointegration tests and then coefficient estimates. Table 1 presents the results of the Kapetanios (2005) unit root tests for all variables.

Figure 1. Time Series Plots of Variables

← 24 | 25 →

← 25 | 26 →

According to results in Table 1, each variable is first difference stationary with structural breaks. The first break date for LNKYH 1999Q3 caused by the earthquakes of 1999. Second break date is 2003Q4 caused by terror attacks in Istanbul (15 November 2003, a bomb attack at synagogue; 20 November 2003 attack at HSBC Bank). Third break date is 2008Q1 caused by Global Crisis initiated in late 2007 in the USA. The last break date is 2010Q1 caused by announcement İstanbul as “European Culture Capital” in the year 2010 (Istanbul Metropolitan Municipality, 2009). The only significant break date for LNYZ is 2004Q4 influenced by European Union (EU) deciding on December 2004 to commence negotiations with Turkey on 3 October 2005 (Ministry of European Union Affairs, 2016). The first break date of LNTOB is 2001Q1 caused by unexpected drought throughout Turkey in the year 2001. The second break date is 2004Q4 initiated by commencing negotiations between the EU and Turkey. The third break date is 2007Q1 influenced by the “Communiqué on Making Direct Income Support Payment” published in the year 2007. The significant structural break dates for LNKSEDS, based on a model with constant and trend, are 2001Q4 caused by the economic crisis in 2001 and 2005Q1 period may have occurred due to the membership negotiations commenced with the EU. Also, the test results based on a model with a constant term indicate the significant breaks in 2005Q4. The significant structural break dates LNGF are 1999Q4, most probably caused by 1999 earthquakes, 2002Q1 following the economic crisis of 2000 and 2001, 2003Q2, influenced by the Law no. 4916 adopted on the date 3 July 2003 that allows sales of real estate to foreigners, 2006Q2 and 2010Q3 caused by the changes made to the Real Estate Tax Law no. 1319. The significant structural break dates of LNHIV are 2001Q4 followed by the crisis in 2001, 2008Q4 caused by the global crisis of 2008. Also, only considering the model with constant, the break dates are 1999Q4 caused by earthquakes in 1999. The results in Table 1 indicate that LNYZ has the break date of 2004Q4 influenced by European Union (EU) deciding on December 2004 to commence negotiations with Turkey on 3 October 2005. Table 2 presents the results of Maki tests.