WHEN IS THE RIGHT TIME TO FLY? THE CASE OF SOUTHEAST ASIAN LOW- COST AIRLINES Chun Meng Tang, Abhishek Bhati, Tjong Budisantoso, Derrick Lee James Cook University Australia, Singapore Campus ABSTRACT This paper examines whether airfares of three Southeast Asian low-cost airlines, i.e. AirAsia, Jetstar and Tigerair, vary according to flight departure time, flight departure day, and flight origin and destination. Data was gathered by visiting the flight booking pages of the three low-cost airlines between September 28, 2013 and October 27, 2013. The data set consisted of flights originating in Singapore (SIN) to Jakarta (JKT), Kuala Lumpur (KUL), Bangkok (BKK) and vice versa. In total, there were 3,450 cases in the data set. The cases were categorized by airfare, flight departure time, and flight departure day. An ordinal logistic regression analysis showed that airline, flight departure day, and flight origin and destination resulted in statistically significant differences in airfare. However, flight departure time did not. JEL Classifications: C21, L93, M20 Keywords: Airfare, Low-cost Airlines, Southeast Asian, Logistic Regression Corresponding Author s Email Address: abhishek.bhati@jcu.edu.au INTRODUCTION Air transport is essential to the development and growth of tourist destinations. The phenomenal expansion of low-cost airlines has allowed more people to travel further at an affordable cost, particularly in the Southeast Asian region as the number of tourists continues to grow. However, existing low-cost airlines face the challenge of operating in an intensely competitive industry, with the emergence of such new players as Malaysia s Malindo and Singapore s Scoot. The business environment is dynamic and complex. New digital technologies have made airfare comparison easy and convenient. Airline ticket prices are more transparent to price-sensitive travellers, many of whom are techsavvy. One of the key success factors of low-cost airlines is pricing their tickets at the right price. Having the right ticket pricing strategy helps low-cost airlines attract price-sensitive travellers and yet maximize profits. Puller and Taylor (2012) reckon that there are a number of factors that can affect ticket prices, e.g. airfare refund eligibility, number of days before departure, and number of vacant seats. However, setting the right ticket price is challenging amid heavy industry competition and profitability. To understand pricing practices of low-cost airlines, in this paper we examine if airfares vary significantly by airline, flight departure time, flight departure day, and flight origin and destination. The following sections provide a literature review, explain research methodology, present research findings, and conclude the paper. LITERATURE REVIEW Airfare In a study about airline ticket pricing decisions, Hsu and Lee (2012) examined how four US airlines priced their tickets on the first day when the tickets went on sale as well as two days before departure. They explained that airlines could price their tickets at full cost or marginal cost and that a better understanding of pricing could help airlines derive an optimum ticket price to promote ticket sales and revenue. Having analyzed airfares of both low-cost and conventional airlines in the London-Paris market, Pels and Rietveld (2004) found that low-cost airlines reacted to each other s airfare movement. Some airlines followed closely that of others while some lowered their airfares when competitors increased theirs. Flight Origin and Destination Narangajavana et al. (2014) suggested that destination airport was one of the factors that affected three types of pricing behaviour: long-term prices or strategic prices; short-term prices with a dynamic demand perspective; and short-term prices with a production perspective. The other factors to influence pricing behaviour were environmental circumstances such as seasonality and airline characteristics. Flight Departure Time and Day
Mantin and Koo (2010) analyzed the airfare history of US domestic airline routes and concluded that there was a price difference between weekdays and weekends. The weekend s price dispersion was about 15% higher than weekdays. They contended that airlines gave ticket discounts on some weekends so as to attract pricesensitive leisure travellers. In another study, Puller and Taylor (2012) found that airline tickets were priced differently based on the day the tickets were sold. On those routes that saw a mix of business and leisure travellers, the price difference was greater. They reckoned that airline tickets were about 5% cheaper on weekends as compared to weekdays, suggesting that Saturdays and Sundays were the best days to buy a ticket. Generally, airlines increase airfares when the departure date gets closer (Pels and Rietveld, 2004). Martin (2012) reported that airline tickets were about 6% lower than the average price six weeks before the departure day. Ticket prices started to go up a week before the departure day and could be about 40% above the average price on the departure day. Zay (2011) also reported that on a weekly basis airlines priced their tickets lower on late Monday. The tickets stayed cheaper till Wednesday or Thursday when they became more expensive again. Ticket sales were highest on Tuesday and lowest on Saturday. RESEARCH METHODOLOGY Data was recorded between September 28, 2013 and October 27, 2013, by visiting the flight booking pages of three Southeast Asian low-cost airlines, i.e. AirAsia, Jetstar and Tigerair. We collected data on airfares, flight departure dates, and flight departure times. Our data set consisted of flights originating in Singapore (SIN) to Jakarta (JKT), Kuala Lumpur (KUL), Bangkok (BKK), and vice versa. In total, there were 3,450 cases in the data set. From a preliminary data analysis, we carefully designed different categories of airfares and flight departure times. We were also able to derive flight departure days from the flight departure dates. Thus, the cases were further categorized by airfare, flight departure time, and flight departure day. Table 1 shows a summary of the cases by the categories airfare, airline, flight origin and departure, flight departure time, and flight departure day. TABLE 1. SUMMARY OF CASES Variable Category N Percentage Airfare Zero fare 597 17.3% 0 and 25 505 14.6% 25 and 50 545 15.8% 50 and 75 485 14.1% 75 and 100 362 10.5% 100 and 125 347 10.1% 125 and 150 153 4.4% 150 and 175 154 4.5% More than 175 302 8.8% Airlines AirAsia 1380 40.0% Jetstar 840 24.3% Tigerair 1230 35.7% Flight Origin and Destination SIN to JKT 570 16.5% SIN to KUL 630 18.3% SIN to BKK 540 15.7% JKT to SIN 570 16.5% KUL to SIN 630 18.3% BKK to SIN 510 14.8% Flight Departure Time Before 9am 630 18.3% 9am and 12pm 630 18.3% 12pm and 3pm 660 19.1% 3pm and 6pm 450 13.0% 6pm and 9pm 750 21.7% After 9pm 330 9.6% Flight Departure Day Monday 460 13.3% Tuesday 460 13.3%
Wednesday 460 13.3% Thursday 460 13.3% Friday 460 13.3% Saturday 575 16.7% Sunday 575 16.7% Total 3450 100.0% RESEARCH FINDINGS The data analysis aimed to examine the relationship between the four independent variables; i.e. airline, flight origin and destination, flight departure time, and flight departure day; and the dependent variable, i.e. airfare. As the independent and dependent variables are all categorical and the airfare categories are ordinal, an ordinal logistic regression is considered appropriate. There were 4,197 (80.3%) cells with zero frequencies. Thus, we do not report the goodness-of-fit measures (e.g. Pearson or Deviance) as they become unreliable when there is a large percentage of cells with zero frequencies. Instead, the change in model fit (in which the full model is compared to the intercept-only model) is reported. The bigger the difference between the two models, the better the independent variables are in predicting the dependent variable. The full model was able to statistically significantly predict the dependent variable over and above the intercept-only model. (χ2(18) = 6446.159, p <.001). We tested if airline, flight origin and destination, flight departure time, and flight departure day had a statistically significant overall effect on airfare. Analyses showed that the overall effect of airline on the prediction of airfare was statistically significant (χ 2 (2) = 2907.279, p =.000); flight origin and destination was statistically significant (χ 2 (5) = 1994.387, p =.000); flight departure time was statistically significant (χ 2 (5) = 28.432, p =.000); flight departure day was statistically significant (χ 2 (6) = 61.933, p =.000). Having determined that the overall effects of airline, flight origin and destination, flight departure time, and flight departure day on airfare were statistically significant, we then examined the parameter estimates. The ordinal logistic regression model produces an equation for each of the J 1 cumulative logits, where J is the number of categories of the dependent variable. There are nine categories of airfare. Thus, there are eight cumulative logits and eight equations. For each variable, the last category is the reference category by default. For example, the effect of each of the first eight categories of airfare is compared to the last category, i.e. more than 175. Results show that some effects were statistically significant and some were not at 95% confidence level (see Table 2). The results are summarized as follows: The odds ratio of AirAsia having higher airfare compared to TigerAir is 9452.701, which is statistically significant (χ 2 (1) = 2773.854, p =.000). In other words, the odds of AirAsia having higher airfare are 9452.701 times that of TigerAir. The odds ratio of Jetstar having higher airfare compared to TigerAir is 1.230, which is statistically significant (χ 2 (1) = 5.286, p =.021). In other words, the odds of Jetstar having higher airfare are 1.230 times that of TigerAir. The odds ratio of Monday having higher airfare compared to Sunday is.650, which is statistically significant (χ 2 (1) = 12.553, p =.000). In other words, the odds of Monday having higher airfare are.650 The odds ratio of Tuesday having higher airfare compared to Sunday is.776, which is statistically significant (χ 2 (1) = 4.392, p =.036). In other words, the odds of Tuesday having higher airfare are.776 The odds ratio of Wednesday having higher airfare compared to Sunday is.786, which is statistically significant (χ 2 (1) = 3.942, p =.047). In other words, the odds of Wednesday having higher airfare are.786 The odds ratio of Thursday having higher airfare compared to Sunday is 1.269, which is statistically significant (χ 2 (1) = 3.886, p =.049). In other words, the odds of Thursday having higher airfare are 1.269 The odds ratio of Friday having higher airfare compared to Sunday is 1.491, which is statistically significant (χ 2 (1) = 10.889, p =.001). In other words, the odds of Friday having higher airfare are 1.491 The odds ratio of SIN to JKT having higher airfare compared to BKK to SIN is.133, which is statistically significant (χ 2 (1) = 251.783, p =.000). In other words, the odds of SIN to JKT having higher airfare are.133 times that of BKK to SIN.
The odds ratio of SIN to KUL having higher airfare compared to BKK to SIN is.006, which is statistically significant (χ 2 (1) = 1134.276, p =.000). In other words, the odds of SIN to KL having higher airfare are.006 times that of BKK to SIN. The odds ratio of JKT to SIN having higher airfare compared to BKK to SIN is.006, which is statistically significant (χ 2 (1) = 1042.164, p =.000). In other words, the odds of JKT to SIN having higher airfare are.006 times lesser that of BKK to SIN. The odds ratio of KUL to SIN having higher airfare compared to BKK to SIN is.001, which is statistically significant (χ 2 (1) = 1577.626, p =.000). In other words, the odds of KUL to SIN having higher airfare are.001 times that of BKK to SIN. TABLE 2. PARAMETER ESTIMATES Estimate S.E Wald df Sig. Exp(B) 95% C.I. for Exp(B) Lower Upper [Airfare = 1] -5.097.195 685.486 1.000.006.004.009 [Airfare = 2] -2.686.179 225.149 1.000.068.048.097 [Airfare = 3] -.187.161 1.341 1.247.830.605 1.138 [Airfare = 4] 1.684.170 98.146 1.000 5.385 3.860 7.514 [Airfare = 5] 3.508.188 349.734 1.000 33.379 23.111 48.211 [Airfare = 6] 5.329.204 685.422 1.000 206.268 138.409 307.395 [Airfare = 7] 6.486.216 904.695 1.000 655.583 429.625 1000.382 [Airfare = 8] 8.198.228 1296.168 1.000 3635.071 2326.370 5679.982 [Airline=1] 9.154.174 2773.854 1.000 9452.701 6723.720 13289.304 [Airline=2].207.090 5.286 1.021 1.230 1.031 1.467 [Airline=3] 0 a.. 0. 1.000 [FlightTime=1] -.241.135 3.167 1.075.786.603 1.025 [FlightTime=2].249.134 3.422 1.064 1.282.985 1.669 [FlightTime=3].239.132 3.258 1.071 1.270.980 1.646 [FlightTime=4].225.145 2.431 1.119 1.253.944 1.663 [FlightTime=5].149.130 1.311 1.252 1.160.899 1.497 [FlightTime=6] 0 a.. 0. 1.000 [FlightDay=1] -.430.121 12.553 1.000.650.513.825 [FlightDay=2] -.254.121 4.392 1.036.776.612.984 [FlightDay=3] -.240.121 3.942 1.047.786.620.997 [FlightDay=4].239.121 3.886 1.049 1.269 1.001 1.609 [FlightDay=5].400.121 10.889 1.001 1.491 1.176 1.891 [FlightDay=6] -.003.114.001 1.981.997.798 1.247 [FlightDay=7] 0 a.. 0. 1.000 [OriginDesti=1] -2.021.127 251.783 1.000.133.103.170 [OriginDesti=2] -5.184.154 1134.276 1.000.006.004.008 [OriginDesti=3].070.123.323 1.570 1.072.843 1.363 [OriginDesti=4] -5.055.157 1042.164 1.000.006.005.009 [OriginDesti=5] -6.622.167 1577.626 1.000.001.001.002 [OriginDesti=6] 0 a.. 0. 1.000 Link function: Logit. a. This parameter is set to zero because it is redundant. The ordinal logistic regression also estimates the probability of a case being in one of the nine airfare categories. A Confusion Table (see Table 3) shows the number of cases that have been classified into different airfare categories. The highlighted cells indicate the number of cases that have been correctly classified. For example, 421 (70.5%) of the zero fare cases were correctly classified; 135 (22.6%) were classified as between 0 and 25; 39 (6.5%) between 25 and 50; 2 (0.3%) between 50 and 75; and no cases for the remaining airfare categories. Overall, our model correctly predicted a case to be in the correct airfare category more than 50% of
the time. However, it is interesting to note that no cases were correctly classified for the 125 to 150 airfare category. Airfare Zero Fare 0 and 25 25 and 50 50 and 75 75 and 100 100 and 125 125 and 150 150 and 175 More than 175 Total Zero fare 0 and 25 TABLE 3. CONFUSION TABLE 25 and 50 Predicted Response Category 50 and 75 75 and 100 100 and 125 150 and 175 More than 175 421 135 39 2 0 0 0 0 597 70.5% 22.6% 6.5%.3%.0%.0%.0%.0% 100.0% 265 130 100 10 0 0 0 0 505 52.5% 25.7% 19.8% 2.0%.0%.0%.0%.0% 100.0% 5 28 279 123 61 49 0 0 545 Total.9% 5.1% 51.2% 22.6% 11.2% 9.0%.0%.0% 100.0% 0 3 125 327 2 28 0 0 485.0%.6% 25.8% 67.4%.4% 5.8%.0%.0% 100.0% 0 0 19 38 275 0 30 0 362.0%.0% 5.2% 10.5% 76.0%.0% 8.3%.0% 100.0% 0 0 2 10 18 257 0 60 347.0%.0%.6% 2.9% 5.2% 74.1%.0% 17.3% 100.0% 0 0 0 3 6 144 0 0 153.0%.0%.0% 2.0% 3.9% 94.1%.0%.0% 100.0% 2 0 0 2 0 0 134 16 154 1.3%.0%.0% 1.3%.0%.0% 87.0% 10.4% 100.0% 1 0 1 0 0 0 0 300 302.3%.0%.3%.0%.0%.0%.0% 99.3% 100.0% 694 296 565 515 362 478 164 376 3450 20.1% 8.6% 16.4% 14.9% 10.5% 13.9% 4.8% 10.9% 100.0% DISCUSSION AND CONCLUSION Our analysis shows that airline, flight origin and destination, and flight departure day have a statistically significant effect on airfare, with the exception of Saturday (flight departure day) and Singapore to Bangkok (flight origin and destination). It is interesting to note that flight time has no statistically significant effect on airfare. Overall, our model correctly predicted more than 50% of the 3,450 cases to be in the correct airfare category, except for the 0 to 25 airfare category. There are four key implications we can gather from the research findings. First, airfares vary across airlines. One possible explanation is how efficient an airline is in its operations. The more efficient one is, the greater its ability to pass the savings on to its passengers. Second, airfares vary according to the location passengers are flying from (origin) and the destination they are flying to. This may be because of the differences in airport taxes and charges. Third, although our analysis shows that Saturday does not seem to have a significant effect, generally flight departure days have an overall significant effect on airfares. It is likely that airfares are higher on certain days when demand is greater and lower when demand is lesser. Fourth, flight time has no effect on airfares. As it is common for low-cost airlines to schedule multiple flights in a day, differences in airfares are thus minimal. There are two research limitations. First, as our data was collected over a short time span, the analysis may not explain other possible fluctuations in airfares. Second, our analysis includes only such variables as airline, flight origin and destination, flight departure time, and flight departure day. However, there are other variables that may affect airfares.
REFERENCES Hsu, S.H. and Lee, J.J., Timing of sale, pricing, and cost information: Evidence from the airline industry, Accounting Perspectives, 2012, Vol. 11, pp. 197-209. Mantin, B. and Koo, B., Weekend effect in airfare pricing, Journal of Transport Management, 2010, Vol. 16, pp. 48-50. Martin, H., Lowest airfares found six weeks before flight, study says, [Online] 2012. Available from http://articles.latimes.com/2012/jan/23/business/la-fi-0123-travel-briefcase-20120123. [Accessed: 2nd May 2014] Narangajavana, Y., Garrigos-Simon, F.J., Garcia, J.S., and Forgas-Coll, S., Prices, prices and prices: A study in the airline sector, Tourism Management, 2014, Vol. 41, pp. 28-42. Pels, E. and Rietveld, P., Airline pricing behaviour in the London-Paris market, Journal of Air Transport Management, 2004, Vol. 10, pp. 279-283. Puller, S.L. and Taylor, L.M., Price discrimination by day-of-week of purchase: Evidence from the U.S. airline industry, Journal of Economic Behavior & Organization, 2012, Vol. 84, pp. 801-812. Zay, L., When is the best time to buy airline tickets?, [Online] 2011. Available from http://news.travel.aol.com/2011/01/31/when-is-the-best-time-to-buy-airline-tickets. [Accessed: 2nd May 2014]