Factors Influencing the Purchasing Decision of Crystal Guava

Factors influencing purchasing decision is vital for the product development and market line expansion. This study aimed to analyze consumer’s characteristics and factors influencing the purchasing decision of crystal guava in agrotourism sites owned by the Agricultural Technical Implementing Service Unit in Semarang City. This was a survey study conducted in Cepoko Agrotourism and Purwosari Agrotourism. These locations selected according to several considerations: managed by the Agricultural Technical Implementing Service Unit, Agricultural Board of Semarang City, distributed crystal guava with lower price, and consumer’s high interest in crystal guava that offered in the agrotourism site. A quota sampling technique was employed to select the number of participant required in the study. There were 100 consumers, 67 consumers from Cepoko Agrotourism and 33 consumers from Purwosari Agrotourism. Accidental sampling tecnique then applied to choose the eligible study participants. Descriptive and multiple linear regression analysis enrolled to analyze the study data. Results revealed that the majority of the participant were women, aged between 39 to 45 years old, graduated from senior high school and university (bachelor/diploma), housewives and entrepreneurs, had an average income of IDR 4,500,000/month, and had four family members. Product quality, promotion activity, attitude,

lifestyle, and the number of family members did not significantly affect the purchasing decision of organic fruits and vegetables. Surprisingly, the physical appearance and flavor of the product had delivered a significant impact on the purchasing decision (Kuhar & Juvancic, 2010). Despite of the evidence provided, price could influence the total purchase. Mensah et al. (2012) found that the variable of price and advertisement influenced the total of the purchase.
Purchasing decisions could be affected by another factor, such as attitude. The attitude reveals the preferences of a product that might define their loyalty to that product. Consumer loyalty, marketing strategy, and country of origin also could affect the purchasing decision (Momani, 2015). Convenient access to the product also contributed a significant role in the purchasing decision. Astuti et al. (2015) investigated purchasing decisions among green malang apple's consumers found that the variable of product, price, location, promotion activity, and price affected their purchasing decision.
One purpose of an agrotourism site was to attain beneficial profits. Consumer's characteristic provides essential information for comprehensive understandings of consumer's needs and desires. This data would become a notable source of information in contriving the marketing strategies.
The effect of product quality, price, location, promotion activity, attitude, motivation, and lifestyle on the purchasing decision was examined in this study. Factors influencing the purchasing decision of crystal guava were vital to be identified as an accurate guide in improving those beneficial factors to increase consumer's loyalty to the product.
Study associated with purchasing decisions is essential. This type of study provides a significant contribution for policymakers and stakeholders. (Gracia & De Magistris, 2007). Furthermore, Massaglia et al. (2019) elaborated that factors influencing purchasing decisions were vital to map the product's characteristics that probably contributed to higher profit.
The purpose of this study was 1) to analyze the characteristic of the crystal guava's consumer and 2) to examine factors influencing the purchasing decision of the crystal guava in Cepoko Agrotourism and Purwosari Agrotourism. The outcome of this study was to provide scientific evidence of factors influencing the purchasing decision of crystal guava in agrotourism sites. Hence adequate improvement and product innovation could be developed according to the findings to increase the selling profit.

RESEARCH METHOD
This study was conducted in Cepoko Agrotourism and Purwosari Agrotourism, Semarang City. A two-month study was carried from December 8th, 2019, to February 8th, 2020, to collect the purchasing behavior among the crystal guava consumer. These locations chosen due to several considerations: managed by the Agricultural Technical Implementing Service Unit, Agricultural Board of Semarang City, distributed crystal guava with lower price, and consumer's high interest in crystal guava offered in the agrotourism sites.
A survey method with an accidental sampling technique applied in this study. A quota sampling from the non-probability sampling technique employed to select 100 participants from two study locations, 67 participants from Cepoko Agrotourism and 33 participants from Purwosari Agrotourism.
Quantitative type of data enrolled in this study. The sources of data employed were primary and secondary data. Primary data employed in this study were collected by interview sessions with the crystal guava's consumer in Cepoko and Purwosari Agrotourism. A questionnaire with close-ended questions and a Likert scale was applied to collect the study data. Secondary data were obtained from Agricultural Technical Implementing Service Unit, Agricultural Board of Semarang City, book, and other works of literature. Interview and observation were applied to collect the study data.
Data analyzed by descriptive analysis and multiple linear regression. Descriptive analysis analyzed the consumer characteristic that consisted of age, gender, educational background, occupation, income, and the number of family members. This analysis aimed to explain the data without making a global conclusion (Purnomo, 2017). Multiple linear regression analysis conducted to analyze the factors influencing the purchasing decision of crystal guava in Cepoko and Purwosari Agrotourism. Kurniawan & Yuniarto (2016) stated that multiple linear regression analysis is an essential analysis tool to examine the effect of more than one independent variable on the dependent variables. Prior to these analysis, the collected data were tested by the validity and reliability test to ensure the integrity and quality of the study instruments.
There were seven independent variables in this study: product quality, price, location, promotion activity, attitude, motivation, and lifestyle. Hence, the mathematical model of the multiple linear regression is as follow: Y = a + b1X1 + b2X2 + b3X3 + b4X4 + b5X5 + b6X6+ b7X7 + e Where: Y : purchasing decision (score) a : regression formulation constanta b1-b7 : regression coefficient of the variable of 1-7 X1 : product quality (score) X2 : price (score) X3 : location (score) X4 : promotion activity (score) X5 : attitude (score) X6 : motivation (score) X7 : lifestyle (score) e : standard error This regression equation examined by the F-test, t-test, and determination coefficient (R2). F-test conducted to know the feasibility of the regression model (Gani & Amalia, 2015). The effect of each independent variable on the dependent variable in this study was identified by the t-test (Kurniawan & Yuniarto, 2016). The strength of this effect demonstrated by the determination coefficient, commonly written in the percentage unit (Nawari, 2010).
Classical assumption test applied to explore the significance of the multiple linear regression equation in predicting the effect of the independent variable on the dependent variable. Classical assumption tests consisted of the normality test, multicollinearity test, autocorrelation test, and heteroscedasticity test. Normality test aimed to know the distribution of data in a study (Santoso, 2018). The correlation between variables in a multiple linear regression model recognized by a multicollinearity test (Ansofino et al., 2016). The data correlation in a period of t with the previous period identified by the autocorrelation test. Sutopo & Slamet (2017) had explained that a good regression analysis found no correlation between the recent data and the previous data. A heteroscedasticity test conducted to assure no occurrence of similarity on the variant from studied residual on the regression model in the recent study. A regression model would be claimed as a good model if no particular pattern was found on the heteroscedasticity test graphic.

Participant's Characteristic
The consumer's characteristic as the study participant was classified according to age, gender, educational background, occupation, income, and the number of the family member. Table 1 shows the characteristic of the crystal guava's consumer. Source: primary processed data (2020) According to Table 1., the majority of crystal guava's consumers in Cepoko and Purwosari's Agrotourism were women (63%). This finding indicated that women had driven a vital role in making a purchase decision. Women are usually responsible for deciding the family food menu. In line with this finding, Eliza et al. (2011) stated that women play a significant role in choosing and providing food for their family, including fruit purchasing. Massaglia et al. (2019) also explained that the majority of fruit and vegetable consumer were women.
The majority of the crystal guava's consumers were aged between 39-45 years old, with a total percentage of 29%. It possibly occurred because the population of those ages begins to raise their awareness to live a healthy lifestyle, especially by consuming an adequate intake of fruit.
The majority of crystal guava consumers in Cepoko and Purwosari Agrotourism graduated from senior high school and university (bachelor/diploma) (46%). The level of educational background could affect the level of knowledge in consuming fruits. Gracia & De Magistris (2007) had elaborated that individuals with higher levels of education was more likely to understand the benefit of consuming organic products. Alfauzan et al. (2015) confirmed that the level of educational background had provided a significant effect on the level of knowledge.
The majority of the consumer were classified into other occupation section (32%). These occupations were listed specifically into housewives (15 participants), university students (11 participants), had retired from their job (3 participants), State-Owned Enterprises' workers (2 participants), dan laborers (1 participant). If these data were ignored, the entrepreneur must have placed as the majority of consumers' occupation. This finding also yielded the role of a mother in making a purchase decisions for the daily food needs. The type of occupation is highly correlated to the level of education and income. Isen et al. (2016) explained that working is a stage after the studying process in life and would determine one's income.
Twenty-six percent of the participants had a relatively high income (IDR 4,500,000). The total income obtained would alter the purchasing power, such as purchase frequency and the total purchase. A study by Kuhar & Juvancic (2010) also supported this finding. They found that income would influence the frequency of fruit and vegetable purchasing. Alfauzan et al. (2015) also stated that the total income linked to the type of purchases.
The majority of the participants had four family members (40%). The level of purchase, especially the total purchase, is associated with the number of family members. A higher number of family members led to a higher total purchase in meeting the food need. Parallel with this finding, Jannah et al. (2018) stated that the number of family members affected the total of food required in a family.

Factors Influencing the Purchasing Decision
Factors influencing the purchasing decision of crystal guava in Cepoko dan Purwosari Agrotourism were analyzed through several steps. The first step was the validity and reliability test for the study instruments applied. The second step was the classic assumption test. The third step was a multiple linear regression test composed of F-test (simultaneous), t-test (partial), and determination coefficient test (R 2 ).
The quality of study instruments were identified through a validity test. This test signified the integrity and accuracy of the study instruments. According to the validity test, all study instruments had a significant value of <0.0. These results indicated that all study instruments applied were valid. If the significant value of the questionnaire's items was <0.05, a study instrument would be declared as valid data collection tool (Purnomo, 2017).
Reliability test conducted to test the reliability of the study instruments. Table  2 shows the result of this reliability test. shows that the value of Cronbach's alpha in the study instruments was more than 0.6. This result indicated that the study instruments applied were reliable.
If the value of Cronbach's alpha ranged between 0.6 -0.79, the data reliability would be accepted (Herlina, 2019).
Multicollinearity test conducted to examine the correlation between the independent variables. Table 3. shows the result of this multicollinearity test.  (2020) The tolerance value in all variables was more than 0.1, or in other words, the value of VIF on all variables was less than ten. This result indicated that no multicollinearity occurred between the dependent variables. Ghozali (2016) explained that multicollinearity would not happen if the VIF value < 10 or the tolerance value > 0.1.
Autocorrelation test conducted to know the correlation between a certain period and the previous period. Durbin Watson Method commonly applied in the autocorrelation test. The value of Durbin Watson (DW) in a regression model was 1,900. The value of Du and dL for seven independent variables and one dependent variable with a total of 100 was 1.8498 dan 1.5060, respectively. The test criteria revealed that the value of dW was greater than dU and less than k-dU, 1.8498 < 1.900 < 2.1502. This finding indicated that no autocorrelation appeared on the data. In line with this finding, Ansofino et al. (2016) stated that if dU<DW<4-dU, no autocorrelation would be found on the data.
Heteroscedasticity conducted to ensure no occurrence of similarity on the variant from studied residual on the regression model in the recent study. A good regression model happened if no heteroscedasticity found on the graphic. The graphic from the heteroscedasticity analysis is revealed in Figure 1.

Figure 1. Result of Heteroskedasticity Test
Source: Primary processed data (2020) Figure 1. shows points or plots that randomly distributed between the number of zero. This distribution described a regression model with no occurence of heteroscedasticity. Sutopo & Slamet (2017) also stated that a good regression model had no specific pattern on the graphic produced by the heteroscedasticity test.
Multiple linear regression analysis was applied to know the effect of the independent variables (product quality, price, location, promotion activity, attitude, motivation, and lifestyle) on the dependent variable (purchasing decision). Table 4 reveals the result from the multiple linear regression analysis. Based on Table 4, the regression equation is as follow: Y = 0.778 + 0.154X1 + 0.021X2 + 0.055X3 + 0.196X4 + 0.247X5+ 0.100X6 + 0.139X7 + e This regression equation is elaborated as follow: 1. The value Constanta (a) was 0.778. This result indicated that if the value of the product quality, price, location, promotion activity, attitude, motivation, and lifestyle variable was assumed as 0, the value of the purchasing decision of the crystal guava was 0.778. 2. The regression coefficient value on the product quality variable was 0.154. Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.154, assuming the value of the other independent variables was constant. 3. The regression coefficient value on the price variable was 0.021. Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.021, assuming the value of the other independent variables was constant. 4. The regression coefficient value on the location variable was 0.055. Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.055, assuming the value of the other independent variables was constant. 5. N Regression coefficient value on the promotion activity variable was 0.196.
Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.196, assuming the value of the other independent variables was constant. 6. The regression coefficient value on the attitude variable was 0.247. Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.247, assuming the value of the other independent variables was constant. 7. The regression coefficient value on the motivation variable was 0.100. Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.100, assuming the value of the other independent variables was constant. 8. The regression coefficient value on the lifestyle variable was 0.139. Hence if this variable increases by one score, then the purchasing decision of crystal guava fruit will increase by 0.139, assuming the value of the other independent variables was constant.
The feasibility of a model was calculated by the F-test. The value of Ftable for 100 participants with seven independent variables was 2.10. The value of Fcount was 19.592. Hence the value of the Fcount > Ftable. This result signified that the variable of product quality, price, location, promotion activity, attitude, motivation, and lifestyle simultaneously affected the purchasing decision. Widiyanto (2013) also emphasized that the value of the F count that greater than the value of the F table showed the refused Ho. T-test was conducted on 100 participants and seven independent variables and one dependent variable with a degree of freedom (df) of 92. This result produced the value of the ttable of 1.986. Based on Table 4, variables that partially affected the purchasing decision were product quality, promotion activity, attitude, and lifestyle. Price, location, and motivation had no partial effect on the purchasing decision.
Product quality was highly related to the purchasing decision. This correlation might be happened due to the high quality of crystal guava offered in the agrotourism sites. Its high quality showed by the freshness, color, shape, and size. In line with this finding, Kuhar & Juvancic (2010) had stated that the physical appearance of the fruit had contributed to the consumer's purchasing decision. The variable of price did not significantly influence the purchasing decision. Contrary to this finding, Mensah et al. (2012) found that price and advertisement were significantly associated with purchasing decisions. But in some situations, the price will not affect the purchasing decision. The high price of crystal guava perceived as a decent price for high-quality crystal guava commodities. The majority of consumers stated that the price of crystal guava in the agrotourism sites was relatively affordable. They believed that the quality of the crystal guava was appropriate with the price offered. Moreover, when the price goes up, the majority of them remain want to re-purchase the crystal guava. Parallel with this finding, Utami & Saputra (2017) also stated that price might do not affect the purchasing decision if the product offered had a high quality. Location was also found to be not correlated with the purchasing decision. We observed that consumers were still purchasing the crystal guava despite the remote location of the agrotourism sites. Anastasia (2017) also stated that consumers tended to settle in their decision in purchasing a product despite the remote location of the business. Attractive product promotion activity was also influenced the purchasing decision of crystal guava. This finding was in line with Astuti et al. (2015) that yielded promotion activity as a factor that affected the purchasing decision.
Attitude was also affected the purchasing decision of the crystal guava. Consumers stated that the unique flavor of the crystal guava in the agrotourism sites had affected their purchasing decision. Attitude showed consumer's preference for a product that would affect their purchasing decision (Firmansyah, 2018). The variable of motivation identified to be not associated with the purchasing decision. This might be happened due to the low level of motivation. Anastasia (2017) also found that motivation did not affect the purchasing decision because of consumer's low motivation. Lifestyle also influenced the purchasing decision of the crystal guava. A healthy lifestyle drove the consumer to purchase fruit to maintain their health. The value of adjusted R Square was 0.568. This value indicated that the variable of product quality, price, location, promotion activity, attitude, motivation, and lifestyle described 56.8% of the purchasing decision. Other unstudied variables explained 43.2% of this purchasing decision.

CONCLUSION
Based on the findings, we concluded that the majority of crystal guava's consumer in Cepoko and Purwosari Agrotourism was women, aged betwen 34to 45 years old, graduated from senior high school and university (bachelor/diploma), classified into other work section (with the majority of housewives and entrepreneur), had an average income of IDR 4,500,000 income/month, and had four family members. The variable of product quality, price, location, promotion activity, attitude, motivation, and lifestyle were simultaneously influencing the purchasing decision of crystal guava in Cepoko and Purwosari Agrotourism. Partially, the purchasing decision affected by the variable of product quality, promotion activity, attitude, and lifestyle. Price, location, and motivation delivered no partial effect on the purchasing decision.

RECOMMENDATION
We suggest the agrotourism management parties improving the variable of attitude by maintaining the quality of the crystal guava commodity. Direct parties involved in the cultivation process need to adhere the standard operational procedure in cultivating crystal guava. A benchmarking study to another site of agrotourism is also vital in upgrading their skill and knowledge. Further studies required to add more variables, such as perception, personality, culture, situation, and referred group from the previous studies.