Research Article An empirical study of price movements in the airline industry in the Indian market with power divergence statistics Received (in revised form): 29th January 216 Goutam Dutta a and Sumitro Santra b a Production and Quantitative Methods Area, Indian Institute of Management, Ahmedabad, India; and b Indian Institute of Management, Ahmedabad, India Goutam Dutta Goutam Dutta is Professor of Operations Management at Indian Institute of Management, Ahmedabad. He holds a PhD from Northwestern University, USA. He has taught at the University of Illinois at Chicago, London School of Economics and HEC Paris as visiting faculty. He is the winner of the Franz Edelman Award (1994), IFORS OR for Development Prize (1993) and was a finalist in the EURO Excellence in Practice Award (26). His papers have been published in several scholarly journals. He is on the editorial board of JORS, Interfaces and IJRM. Sumitro Santra Sumitro Santra is a Research Assistant at Indian Institute of Management, Ahmedabad. He holds a MSc in Economics from University of Calcutta, India. Before joining at IIMA he had worked at Calcutta Institute of Technology, Kolkata and IIFT, Kolkata. His area of interest in revenue management, data mining, competitive pricing strategy, time series forecasting and business analytics related to Airlines and BFSI domain. Correspondence: Goutam Dutta, Production and Quantitative Methods Area, Indian Institute of Management, Ahmedabad 3815, India E-mail: goutam@iimahd.ernet.in ABSTRACT In this article, we analyze the dynamic price dispersion of the Indian domestic airline industry. We develop the power divergence statistic (PDS) for each route and study the average PDS and average airfare movement. We analyze the effect of Average PDS based on a few selected route characteristic variables and market structure variables. Our research suggests that the competition and price dispersion steps up airfares as the departure date comes closer, when both full service carriers and low cost carriers operate in the same market. Application of Revenue Management and Dynamic Pricing is a common practice in the Indian domestic airline industry. It also shows that route characteristics affect airfare movement as well as airfare dispersion in different routes. Journal of Revenue and Pricing Management advance online publication, 11 March 216; doi:1.157/rpm.216.12 Keywords: airline industry; airfare movement; dynamic price dispersion; power divergence statistic; revenue management INTRODUCTION The market for homogeneous (or close substitute) goods and services is often characterized by price differential (between products), price dispersion (over time) and price discrimination (among customers). The airline industry, which has all these features, has been the main focus of such studies over the years. The Revenue Management and Dynamic Pricing (RMDP) (Talluri and Van Ryzin, 24) www.palgrave-journals.com/rpm/
Dutta and Santra system is an integral part of the IT system in the airline industry. It helps to update airfares dynamically based on the actual booking (till that time) and demand forecast, price sensitivity of the routes and several market and customerrelated variables. The RMDP system sets corresponding booking limits for the updated airfares for each flight. In a competitive environment, every airline operator offers airfares based on their own competitive pricing approaches (Bilotkach et al, 21). There are two types of passenger airlines carriers Full Service Carriers (FSC) and Low Cost Carriers (LCC). These carriers operate in the same competitive airline market. When the booking window for a particular departure day s flight opens (say, 6 or 9 days before departure day), we observe that the airfare is relatively low and LCCs set their fares lower compared with conventional FSC (Pels et al, 24). Then the airfare dynamically increases as the departure date gets closer. The Indian domestic airline industry is operated by both FSCs and LCCs. Domestic air traffic grew at 18.5 per cent per annum during 24 25 to 21 211 (Report June 212, Ministry of Civil Aviation, Government of India). In India, the airfare (price) dispersion varies across different domestic air routes. The literature (Gillen and Mantin, 29; Mantin and Koo, 29) suggests that airfares vary substantially over time across routes, and airfare dispersion varies according to the route characteristics, market structure, potential size of the market, supply flow and so on. The pioneering work in this area on the development of a Power Divergence Statistic (PDS) was done by Read and Cressie (1988) and was used in this context by Mantin and Koo (29) for analyzing the dynamic price dispersion of the United states (US) airline industry. In this study, we look at the dynamic airfare dispersion analysis across airfare histories among 69 randomly selected Indian domestic air routes and how these dynamic airfare (price) dispersions are explained by the relevant macroeconomic variables like route characteristics variables (distance, average population, average income between origin destination city) and the market structure variables (number of nonstop flights, proportion of LCC operation, degree of the competition index Herfindahl- Hirschman Index (HHI)), and lowest average airfare between origin destination city. Our current airfare dispersion study can help airline regulators to regulate the airfare during the booking period whereas; It can help price sensitive passenger to understand the price variation. Similarly, this study can help the existing and/or newly entrant airlines operators to understand the variation in the market and make their pricing-related decisions. In this article, we adopt the similar methodology as Mantin and Koo (29) adopted to examine the airfare dispersion of the US airlines industry. However, there is practically no literature that discusses the airfare movement or PDS in airlines in the Indian domestic market. This article is probably one of the early attempts that discuss airfare movement and airfare dispersion in the form of PDS in the context of the Indian domestic airline market. On the basis of the objective of this study and its usefulness we plan to address the following questions: 1. How does the PDS vary in different routes? 2. To what extent can PDS variation be explained by the distance of two cities, average population of two cities, average income of two cities and degree of competition? 3. What are the similarities and the differences between our study and the recent study by Mantin and Koo (29)? This article is organized as follows. In this section, we provide an introduction about the Indian airline industry with its market characteristics. In the next section, we review literature on the dynamic pricing mechanism. After that we provide the data sources. We define the PDS, route characteristic variable, and market structure variable, and develop a regression model of PDS with route characteristic variables (distance between origin destination city, average 2
An empirical study of price movements population and average income) and market structure variables (number of non-stop flights, LCC proportion, average airfare and HHI on selected origin destination cities). Then we explain the airfare movements for the selected five routes and the analysis of the PDS study with the introduction of route characteristic variables and market structure variables. We specify the usefulness of the study to passengers, service providers (Airline Companies), regulators and researchers. Finally, in the final section, we draw some conclusions about the nature of the Indian domestic airline industry. LITERATURE REVIEW While looking at the literature, we find that there are three important areas that can be studied. The first deals with the pricing strategy. Borenstein and Rose (1994), Stavins (21) and Gerardi and Shapiro (27, 29) empirically studied the relationship between market competition and price dispersion with cross-sectional data. Dynamic pricing strategies are driven by customer dynamics rather than price discrimination over an existing set of customers (McAfee and Velde, 27). In a competitive market with uncertain aggregate demand where firms pre-commit to a capacity, equilibrium price is determined by the airline market demand and supply (Talluri and Van Ryzin, 24). Often, irregular pricing strategies help firms to avoid price wars with their competitors, while price dispersion is often an outcome of uncertainty in demand and higher capacity costs in the oligopolistic market (Dana, 1998). Hernandez and Wiggins (28) studied the non-linear pricing behavior of the US airline industry. They found that there is a negative correlation between market concentration and price dispersion. The second area discusses competition and price discrimination. Gallego and van Ryzin (1994) discussed on inter-temporal price discrimination approaches in a revenue management perspective. Hazledine (26) empirically worked on the price discrimination strategies of homogenous products where the firm had charged according to the consumers willingness to pay and in this competitive market structure, average prices were independent of price discrimination strategies. Su (27) studied intrafirm price discrimination and classified consumers based on their waiting cost, and set prices dynamically based on various rationing rules. Escobari and Gan (27) studied price dispersion where capacity costs are higher with uncertain demand. They used US airline data and observed that the second degree price discrimination is applicable when offering advanced discounts on purchase. The third area deals with competition, price movement and price dispersion measure along the booking profile. Piga and Bachis (27) examine the daily change in airfares for FSCs as well as LCCs. They conclude that each airline s price distribution tends to rise as the departure date comes closer. So, there are price movements within the booking period. To capture this dynamic pricing movement, many researchers used some statistical measures. We discuss some of the major routes (a) Brynojlfsson and Smith (2) used ranges as a dispersion measure to study the pricing behavior of the online and conventional retail outlet market. It is calculated simply by taking the maximum and minimum prices only. (b) Sorensen (2) used coefficient of variation as a measure of price dispersion. It measures the degree of variability simply by taking the ratio of standard deviation to mean. But in the airfare (price) analysis context. (c) Borenstein and Rose (1994) and Gerardi and Shapiro (27) used Gini Coefficient as measure of price dispersion. It is a measure of statistical dispersion intended to represent the (cumulative) income distribution of a nation s residents. It is one of the most widely used measure of inequality or varaion of the data from uniform distribution. The Gini coefficient measures the inequality among values of a frequency distribution 3
Dutta and Santra (for example, levels of income). A Gini coefficient of zero expresses perfect equality, where all values are uniformly distributed (for example, where everyone has the same income). A Gini coefficient of one (or 1 per cent) expresses maximal inequality among values or where its difference from ideal situation is maximum. Mathematically, it is equivalent measure of the half of relative mean absolute deviation and measures degree of inequality. The mean absolute difference is the average absolute difference of all pairs of items of the population, and the relative mean absolute difference is the mean absolute difference divided by the average, to normalize for scale. (d) Atkinson index A e (Hayes and Ross, 1998) also a measure of inequality to understand the distribution of inequality. It has been used on the quarterly or monthly data only. The index can be turned into a normative measure by imposing a coefficient ε to weight incomes. For ε =, it is assumed there is no aversion to inequality and we get A ε =, and for, for ε = infinity, it is assumed that it will have infinite aversion to inequality A ε = 1. Mantin and Koo (29) studied on the daily airfare prices and emphasized the link between dynamic price dispersion in the airfare and the distance between origin destination airports, average population, average per capita income, business index, number of non-stop flights, proportion of LCC, HHI, average number of boarding, and average fare level, and explained how these factors varied over time on different routes in the US. They introduced the PDS as a measure of airfare dispersion and found that dynamic airfare dispersion depends on the demographic characteristics of the routes and the presence of low cost carriers, but not on the competition intensity. Mantin and Koo (21) extended their airfare dispersion studies by introducing the time effect and the weekend effect. Similar airfare dispersion (PDS) study was done by Obermeyer et al (213) in European airline markets. In their study, they established that efficient airlines were better able to differentiate airfares compared with their inefficient counterparts. Looking at the literature search, we find that although studies have been conducted in the European and US market, no work has been done on the price movement of competing airlines in the Indian context. We found that there is hardly any published literature in this sector. Hence we decided to take up this study. DATA SOURCE In our airfare data collection process, we collect the minimum airfare data from 69 domestic airline routes across India (for the economy class). We develop a regression analysis of price dispersion versus seven factors: distance between two airports, number of non-stop flights, proportion of LCCs, average population of the city pair, average per capita income, average airfare and HHI. The data of daily economy class airfares of 31 departure dates (1 December 31 December 212) has been collected for selected city pairs from a popular travel Website (www.yatra.com). We collect the one way minimum airfare for all the airlines operating on that particular route from 6 days before each departure date. We collect distance data (in km) between two origin destination s airports from Google Map (www.google.com). Population data collected from the latest census report 211 of the Government of India and the average per capita city income has been calculated from state level per capita income (National State Domestic Product 211 212). We obtain average state level per capita from the CMIE database States of India. PRICE DISPERSION ANALYSIS MODEL ON SELECTED ORIGIN DESTINATION CITIES In this section, our objective is to capture the airfare dispersion in Indian domestic airline markets based on relevant macroeconomic 4
An empirical study of price movements variables, and we answer the first, second and third questions stated in the section Introduction of this article. There are various types of price dispersion measures such as price range, coefficient of variation, Gini coefficient, Atkinson index and so on that we have described earlier. But to capture the dynamic nature of daily airfare data, one of the most suitable parameters is the PDS (Read and Cressie, 1988) which has been used by earlier researchers. PDS measures the degree of divergence by comparing the observed and expected level of frequencies. The PDS form is as follows: PDS r;t ¼ 2 X n λðλ + 1Þ i¼1 " ( )# ; F i r;t Fr;t i λ - 1 E r;t (1) where r denotes routes and t denotes days before departure. The PDS for route r on t days before departure is denoted by PDS r,t. Here, i F r,t is the observed airfare for route r on t days before departure on i th departure day on airfare history i, i = 1, 2,.n. E r,t is the expected average fare for r routes on t days before departure. And, λ is a real valued family parameter. If λ tends to, then the PDS statistic becomes a Log-likelihood ratio statistic. Further, if λ equals to 1 then the PDS becomes Pearson s χ 2 statistic. So if λ values lie between and 1 (, 1), then PDS values lie between the Pearson χ 2 values and the log-likelihood statistic values. The deviation of the PDS value depends on the magnitude of the observed fare and expected fare. If the observed fare is greater than the expected fare then the PDS value increases and vice-versa. In our study, we consider the average PDS values for route r and set λ as a family parameter at the level 2/3 according to Read and Cressie (1988). For the price dispersion analysis, we consider a number of relevant macroeconomic factors. We cluster these variables in the form of route characteristics variables and market structure variables for different city pairs. ROUTE CHARACTERISTICS VARIABLE To identify route characteristics for a particular origin destination city, we consider distance, average population and average income between the origin destination cities as a proxy variable. These variables help us to identify route characteristics for a particular origin destination city pair. The distance data between origin destination airports represents the operational cost between origin destination city pairs as an alternative variable. The average population of a city pair (origin destination) represents the market size for that air route. The average population has been calculated by taking a simple average between the two origin destination cities. The average income between the city pairs captures the potential market size. We calculate the average income simply by taking the arithmetic mean between two city pairs. MARKET STRUCTURE VARIABLE: The market structure variable includes the number of non-stop flights, the proportion of LCC operation, HHI and the lowest average price for each route as a proxy of the market structure variable. The Number of Non-Stop Flights between each origin destination air route shows the average number of seats for each air route as a proxy variable. This variable alternatively depicts passenger demand based on the average population and average income between city pairs. The Proportion of LCC Operation shows the ratio of LCCs operated out of a total number of flights operated (both LCC and FSC). Here we have used this variable as a proxy variable to determine the degree of market competition in the presence of LCCs. The HHI measures the degree of competition among the competing airlines. 5
Dutta and Santra Table 1: Descriptive statistics of route characteristics and Market structure variables Descriptive statistics Mean Standard deviation Number of non-stop flights 9.12 9.36 (avg. per day) Proportion of LCC (avg. per.78.23 day) Distance in km (1 km) 9.9 4.88 Average population of city pair 8.62 3.7 (million) Average per capita income 93.23 28.11 (Rs. 1) HHI.45.3 Number of Routes: 69 We calculate this index based on the frequency of flights operated on different competing airlines for the particular air routes in the given time period, 1 31 December 212. HHI r ¼ XJ j¼1 X jr P J j¼1 X jr! 2 ; (2) where X jr denotes the number of flights operated by the j th airline for the route r and assuming there are J number of airlines operating in the route r. We take the lowest average airfare for each route. It is one of the important variables in our study and it represents the airfare structure of each route. The summary statistics for these variables are shown in Table 1. From the above discussion, we develop a regression model to analyze price dispersion based on the dependent variable, that is, the average PDS, and the explanatory variables that are distance, average population, average income, number of non-stop flights, LCC proportion, HHI and price. The estimated model for route r is expressed as follows: The Model Avg PDS r ¼ β + β 1 *Distance r + β 2 *Avg Pop r + β 3 *Avg Income r + β 4 *NonStop Flight r + β 5 *LCC r + β 6 *HHI r + β 7 *Price r + ε r r ¼ 1; 2; 3 ¼ ¼ ¼ ; R ð3þ where, Distance r Distance between origin destination cities (in kilometers) Avg_pop r Average population between origin destination cities (million) Avg_Income Average income between origin r destination cities (Rs. 1) Non- Number of non-stop flights Stop_- between origin destination cities Flight r (average per day) LCC r Proportion of LCC operated (average per day) HHI r Herfindahl-Hirschman Index for the route r Price r Lowest average price between origin destination cities (in Rupees) RESULTS AND DATA ANALYSIS OF PRICE DISPERSION MODEL In this section, we discuss through graphs, how the airfares (minimum airfare, maximum airfare and average airfare) and average PDS change dynamically over the time interval for the selected routes. For any departure day in any selected route, we collect minimum airfares from 6 days to 1 day before the departure day for each of the selected routes, respectively. The process was repeated for 31 departure days Therefore, for each day before departure, we have 31 observations (that is, 31 observations for 6 days before departure) and we compute the minimum airfare, maximum airfare, and average airfare, and the average PDS. We then do a regression analysis based on model described (equation No 3) in the 69 selected routes. In the following part, we explain the airfare behavior with PDS movements on selected five origin destination cities. These routes are 6
An empirical study of price movements Airfare 2 18 16 14 12 1 8 6 4 2 6 55 5 45 4 35 3 25 2 15 1 5 Figure 1: Average price and PDS on Chennai Kolkata routes. Days before Departure Max Min Average PDS 8 7 6 5 Airfare 4 3 2 1 Figure 2: Average price and PDS on Goa Mumbai routes. 6 55 5 45 4 35 3 25 2 15 1 5 Days before Departure Max Min Average PDS operated by both FSCs and LCCs. In Figure 1, we show the airfare behavior of the Chennai Kolkata route. These two cities are metropolitan cities in India and international flights also operate from these two cities. The LCC operation for this air route is 87.5 per cent of total number of flights in figure 1, the average airfare remains stable from 6 days to 14 days before departure, but the average PDS is not so smooth compared with the average fare. Here we observe that the average PDS rises, followed by more ups and downs. For 2 weeks before departure, the average airfare rises upward steadily and the corresponding average PDS becomes more unstable. This instability follows because of the airfare divergence between the observed airfare and the expected airfare. In Figure 2, we represent the maximum fare, minimum fare, average fare and average PDS for the Goa Mumbai air route. In this city pair, Goa is a tourist destination whereas Mumbai is a metropolitan city where international flights operate. In this route structure, we observe that 7
Dutta and Santra 12 1 8 Airfare 6 4 2 6 55 5 45 4 35 3 25 2 15 1 5 Days before Departure Max Min Average PDS Figure 3: Average price and PDS on Kolkata Mumbai routes. 16 14 12 Airfare 1 8 6 4 2 6 55 5 45 4 35 3 25 2 15 1 5 Days before Departure Max Min Average PDS Figure 4: Average price and PDS on Bangalore Delhi routes. the average fare improves steadily but the average PDS fluctuates more. The average PDS initially rises from 6 days to 3 days before departure and within this time interval, the average fare remains steady, but after 28 days, the average fare increases upwards steadily. During this time interval, the average PDS starts to fall and remains stable at a certain average PDS interval of 5 2. That is, within this time interval, the price dispersion between the observed fare and expected fare becomes more stable. In Figure 3, we represent a graph for the Kolkata Mumbai air routes. Both cities are metropolitan cities where international and domestic flights operate. The LCC operation for this air route is 44.1 per cent of total number of flights. Here, we observe that the average fare and the average PDS remain stable for 6 days to 3 days before departure. Then both average fare and average PDS rise upward steadily. So, for the 28 days to 1 day before departure time interval, the average airfare rises closer to departure day and at the same time the rate of average PDS also tends to rise. So, within this time interval, the rate of increase in observed fares is higher compared with the rate of increase in expected fares. In Figures 4 and 5, we have shown price movement against average PDS for the air routes of Bangalore Delhi and Delhi Mumbai. We draw a graph of average price and average PDS for all air routes (69 routes). In Figure 6, we observe that the average fare 8
An empirical study of price movements Airfare 9 8 7 6 5 4 3 2 1 6 55 5 45 4 35 3 25 2 15 1 5 Days before Departure Max Min Average PDS Figure 5: Average price and PDS on Delhi Mumbai routes. 14 12 1 Airfare 8 6 4 2 6 55 5 45 4 35 3 25 2 15 1 5 Days before Departure Average Price for all routes Average PDS for all routes Figure 6: Average price and PDS on all air routes (69 routes). Table 2: Maximum airfare ratio of selected 5 routes Route (D 1 /D 6 ) ratio of maiximum airfare Chennai 3.527 Kolkata Goa Mumbai 2.42 Kolkata 2.42 Mumbai Bangalore Delhi 2.345 Delhi Mumbai 1.765 remains stable from 6 days to 32 days before departure and after that it tends to rise. The average PDS follows a stable pattern from 6 days to 2 days before departure and after that it diverges more. After 2 days before departure, we find that the rate of increase in observed fare is greater compared with the rate of increase in the expected fare for all routes. The average maximum airfare movement of the selected five routes increases as the departure day gets closer. In our study, we observe that the average maximum airfare decreases within the interval 5 days to 1 day before departure for the Goa Mumbai and Bangalore Delhi air routes. Thus, in spite of the application of RMDP strategies, airline operators often step down maximum airfare as the day before departure gets closer. The possible causes behind the airfare step down are lower passenger demand and fewer tickets sold. In Table 2, we show the ratio of (D 1 /D 6 ) for the five air routes. This ratio is significant to 9
Dutta and Santra understand the maximum extent of airfare for each route. It shows that the maximum airfare at 1 day before departure compared with 6 days before departure. Out of five air routes, our study suggests that the maximum extent of airfare at D -1 is 3.5 times of D -6. The airline operators often form a cartel and offer airfare dynamically according to their pricing strategies. Thus, the ratio (D 1 /D 6 ) can help to regulate the maximum airfare and the passenger can understand the airfare variability based on the days before departure. From the above discussion, we conclude the following: 1. An important justification for the fluctuation of average PDS for a route, as the departure day comes closer, is that, both FSCs and LCCs operate in the same air route with different pricing strategies. 2. The dynamic increase of average airfare and the average PDS fluctuation takes place from 14 days (2 weeks) before departure for all routes. 3. On the basis of the route characteristics, percentage of LCC operations and time (days before departure) factor, each route follows a different pattern of average airfare and airfare (price) dispersion. 4. Average PDS movement for all the routes airfare seems to be more predictable than the average PDS movement for an individual route s airfare. 5. The ratio of (D 1 /D 6 ) will be able to identify the extent of maximum airfare level for each route. In the following part, we discuss price dispersion analysis in terms of the average PDS based on the route characteristics variables and market structure variables. The estimated regression model s (Model 1) result is given in Table 3. In our regression analysis, we take 69 routes data from different Indian domestic air routes and we consider only non-stop flights between origin destination city pairs. The average PDS is calculated by taking 6 days before departure to 1 day before departure data. In our sample study, our dependent variable is Average PDS and the independent variables are average population, distance, number of non-stop flights, price, LCC proportion, HHI and average per capita income. In Table 3, the independent variable, Average population has the least (absolute) negative significant impact on the dependent variable Average PDS and we find that potential market size has an reverse effect on the Average PDS. This finding supports Mantin and Koo s (29) study. In our study we have assumed that the market size is proportional to the average population. Thus, as the average population (market size) goes up, there is a decrease in the airfare dispersion. In our study, we find that a negative relationship exists between Average PDS and Average Table 3: Estimated regression equation Avg PDS ¼ 4677:9252 + :7717*Distance - :5*Avg population - :49*Avg Income ð2262:4867þ * ð1:1276þ ð:2þ ** ð:12þ + :5982*Non - stop flights - 3742:427*LCC - 226:819*HHI + 1:3986*Price ð:5982þ ð1285:9812þ ** ð126:5929þ ð:5264þ ** ð4þ N = 69 R 2 =.4451 R 2 adj =.3814 F = 6.99 ** **significant at 1 per cent level, *significant at 5 per cent level. Standard error values are in parentheses. 1
An empirical study of price movements Per Capita Income. But our study does not find any significant relationship between Average Per Capita Income and Average PDS. In this context our result is different compared with Mantin and Koo s (29) study. There are no significant relationships between Average PDS and distance. This finding is similar to the study by Mantin and Koo s (29). Thus controlling other independent variables, operating cost does not explain the variation of the Average PDS. The market structure variable LCC proportion has a large (absolute) negative significant impact on the Average PDS (controlling other independent variables) but this negative relation does not hold true in the Mantin and Koo (29) study. Thus, with the introduction of more LCC operation (where FSCs are already operated) for a particular air route, competition among airline operators becomes more intense. Since, FSCs set their airfare higher compared LCCs and dynamically increase their airfare as the day of bookings come closer to operations. Hence, to stay in the airline business, FSCs also frequently change their pricing strategies and compete with LCCs (Gillen and Mantin, 29). But when we consider the market structure variable Number of non-stop flights, we do not find any significant relationship with the dependent variable Average PDS. Hence, the passenger demand for a city pair does not capture any effect on the Average PDS. This, again, is different from Mantin and Koo s (29) study in the US context. The market competition index HHI does not hold any significant relationship vis-à-vis Average PDS in our study. Hence, the degree of market competition, as explained by HHI index, does not capture the average price dispersion that is developed according to the daily prices (airfare). The same relationship holds in Mantin and Koo s (29) study also. We also find that the day to day frequency of flight operation (competition intensity) does not capture the variation of the Average PDS. However, the LCC proportion negatively relates to the Average PDS and it reflects the extent of competitive intensity among the airlines operators for a particular air route. Hence, these two results give us contradictory conclusions regarding the competition intensities among the airlines operators for a particular route. The average lowest airfare (price) has a positive significant impact on the Average PDS (Table 3). Since the price variable implies the overall airfare structure of the Indian domestic airline market, getting a significant relationship with average PDS helps us to capture the airfare dispersion. This result is completely supported by Mantin and Koo s (29) study. Thus, the increment of average airfare implies that the FSCs offer higher airfares compared with LCCs airfare or that both operators raise their airfare. Airline companies frequently revise their prices based on their estimated forecasted demand and their past experiences. They generally upgrade their prices once or twice a week. These strategies help airline operators to improve their revenues. But how do these weekly price revisions affect the Average PDS in the Indian domestic airline industry? To answer this question we analyze the entire sample on a weekly basis. The weekly average price dispersion measures are based on the route characteristics and market structure variables and consider the average weekly airfare, where Week 1 represents 1 7 days before departure, Week 2 represents 8 14 days before departure and so on. The results are shown in Table 4, and in Figure 7. In Figure 7, we plot only the three significant β coefficients of the variables: LCC, average population and price for 8 weeks. Hence in this figures, we find how the β coefficients change as we move from 8 weeks before departure to 1 week before departure. In Figure 7, we find that the coefficients of all the three variables decrease as the weeks decrease (days before departure gets closer). During this time, airline operators revise their prices weekly, and the corresponding coefficient of the route characteristics and market structure variables also change accordingly. 11
12 Table 4: Weekly price dispersion Variables Week 1 Week 2 Week 3 Week 4 Week 5 Week 6 Week 7 Week 8 Average population of city pair.13**.1**.5*.3*.2.1.3*.4** (.4) (.4) (.2) (.2) (.2) (.4) (.2) (.2) Distance in km 3.454 4.5549 #.6584.5128 1.2334 3.716 1.112 1.598 (3.5136) (2.7117) (1.6134) (1.2) (1.1986) (2.3922) (1.598) (.9915) No. of non-stop flights 4.3253 3.9685.327.53 1.9234 1.164.3533.134 (3.3756) (3.825) (1.8544) (1.1793) (1.5159) (2.9561) (1.2513) (1.158) Price.555.2411 1.1533* 1.2228** 1.9178** 2.8493** 1.9439** 2.4328** (.9216) (.8817) (.5815) (.3781) (.4741) (.989) (.457) (.4497) LCC 7491.865* 6395.6477* 1461.398 3189.781** 159.516 9441.4** 548.474 174.6327 (33.989) (353.612) (1818.2447) (1153.3464) (1488.292) (2895.9591) (1228.794) (1135.454) Average per capita income.532*.411 #.12.26.16.186.21*.199* (.251) (.231) (.142) (.9) (.115) (.224) (.97) (.92) HHI 835.7441 2.7987 2153.72 114.2113 228.336 3417.3644 98.7247 1144.616 (3258.879) (2986.2399) (182.1149) (1141.2114) (1453.8157) (2833.12) (12.3678) (111.971) No. of observations 69 69 69 69 69 69 69 69 R 2.3364.266.2683.3685.3547.3316.4519.5497 Adjusted R 2.263.1818.1843.296.286.2669.389.4981 **1 per cent significant, *5 per cent significant, #1 per cent significant. Standard error values are in parentheses. Dutta and Santra
An empirical study of price movements Coefficient Coefficient Coefficient LCC 1 2 3 4 5 6 7 8 9-2 -4-6 -8-1 Week 1 2 3 4 5 6 7 8 9. -.5 -.1 -.15 Average Population Week Price 3 2.5 2 1.5 1.5 1 2 3 4 5 6 7 8 9 Week Figure 7: Change in regression coefficient with respect to week. Note: In the above figure we plot only significant coefficients against week.+: Significant at 1 per cent; X: Significant at 5 per cent. Since the Indian domestic airline market is operated by both FSCs and LCCs, the weekly airfare revision has a significant impact on the Average PDS. The signs of the significant variables coefficients of the regression model are the same for the overall Average PDS (Table 3) and the separate weekly Average PDS models (Table 4). In our study, the coefficient sign of the independent variables LCC with respect to each week is totally different compared with Mantin and Koo s (29) study in the context of US airlines. In their study, the coefficient sign of the independent variable price (airfare) in the overall study is different compared with the coefficient sign of airfare of the weekly study. However, in our sample study, we find the same negative coefficient impact of the variables LCC proportions and the Average populations for all the weekly cases. The weekly analysis of the Average PDS based on the market structure variables and route characteristics variables illustrates that the every operating airline (both LCCs and FSCs) increases their airfare dynamically in a competitive environment. In our study, out of seven explanatory variables, we have found that a few variables are significant, and these variables have collectively captured a low percentage of variations of the average PDS. Although our sample study does not capture the maximum price variability of the collected sample data set based on the seven explanatory variables, these seven variables are very important in the Indian domestic airline market. In this context, we can possibly include a few relevant explanatory variables like ;business index, average number of boarding and so on but because of lack of data availability, we have not been able to introduce these. Usefulness of the analysis for airlines operators, passengers, airlines regulatory authority and researchers This analysis can be useful to all the stakeholders (passengers, new entrants and so on) that are crucial to airlines. First, if we can set a good price dispersion index and set that index across various seasons, the regulators can prevent an unusual price hike. The airfares become very high 1 or 2 days before the Diwali festival in November. In India, the DGCA (Director General of Civil Aviation) is the nodal agency as a regulator. The DGCA can use this study as a reference to control unusual increases in airfares. Second, the airline industries will benefit from this study by looking at the PDS history of the last few years while setting the initial price. This will help both the existing and new airline companies to understand their competitor s price behavior, route characteristics of the each route with respect to the market structure variables and route characteristics variables. Third, it may be useful for price sensitive passengers when they are taking a decision to buy their air ticket. If this rule is strictly regulated by the DGCA, the airline operators can increase or decrease airfares according to the demand, but 13
Dutta and Santra only up to a certain limit. This study may be useful to break the cartel among the airlines operators and may set the upper limit of the mercenary pricing strategies for each route. Fourth, researchers should get an idea about the Indian domestic airline s airfare market and discover how airline operators apply dynamic pricing strategies in a competitive airfare market. The ratio of (D 1 /D 6 ) for each route can identify the extent of maximum airfare and it helps all the stakeholders to identify the ceiling level of airfare. CONCLUSION All the results compiled together indicate that there is similarity price movement analysis in the Indian and US markets. In general as the days before departure come closer, airfares increased evenly and airfares increased at regular intervals, which indicate a block pricing strategy. Each route is characterized by its own significant route characteristic behavior and the competing airline operators set their prices according to these route characteristics. We form a regression analysis on an overall and weekly basis where the average PDS relates to the route characteristics variables and market structure variables. We compare our findings with that of Mantin and Koo (29). These explanatory factors provide details about the Indian domestic market structure. The regression analysis provides us with a clearer representation of the dynamic pricing movement of the Indian domestic airline industry based on these macroeconomic factors. 1. Although our data set is small, our study confirms most of the results suggested by Mantin and Koo (29) and we can say the generic conclusions drawn about price dispersion in the Indian market are similar to those in the US market. In fact we have found that even our data set is small we get similar result to that of Mantin and Koo (29). 2. We also find that dynamic airfare dispersion depends on the time (days before departure) and LCC operations in the presence of FSCs. 3. Each route is characterized by its unique price dispersion features with the relevant macroeconomic factors like LCC proportion, average population between city pairs, average airfare. 4. Average PDS movement for all the routes airfare combined seems to be more predictable than the average PDS movement for an individual route s airfare. 5. The airfare movement depends on the degree of airfare competition (strategies) between LCCs and FSCs and every competing airline applying RMDP strategies. 6. The ratio of airfare 1 day before departure to 6 days before departure can identify the ceiling level of airfare for each route. This can be done to restrict exceptionally high airfare on certain days like Diwali days. REFERENCES Bilotkach, V., Gorodnichenko, Y. and Talavera, O. (21) Are airlines price-setting strategies different? Journal of Air Transport Management 16(1): 1 6. Borenstein, S. and Rose, N.L. (1994) Competition and price dispersion in the U.S. airline industry. Journal of Political Economy 12(4): 653 683. Brynjolfsson, E. and Smith, M.D. (2) Frictionless commerce: A comparison of internet and conventional retailers. Management Science 46(4): 563 585. Dana, J. (1998) Advance-Purchase discounts and price discrimination in competitive markets. Journal of Political Economy 16(2): 395 422. Escobari, D. and Gan, L. (27) Price dispersion under costly capacity and demand uncertainty. NBER Working Paper No. 1375, National Bureau of Economic research, Cambridge, MA. Gallego, G. and van Ryzin, G. (1994) Optimal dynamic pricing of inventories with stochastic demand over finite horizons. Management Science 4(8): 999 12. Gerardi, K.S. and Shapiro, A.H. (27) The effect of competition on price dispersion in the airline industry: A panel analysis. Working Paper No. 7-7, Federal Reserve Bank of Boston. Gerardi, K.S. and Shapiro, A.H. (29) Does competition reduce price dispersion? New evidence from the airline industry. Journal of Political Economy 117(1): 1 37. Gillen, D. and Mantin, B. (29) Price volatility in the airlines markets. Transportation Research Part E 45(5): 693 79. Hazledine, T. (26) Price discrimination in Cournot-Nash oligopoly. Economics Letters 93(3): 413 42. 14
An empirical study of price movements Hernandez, M.A. and Wiggins, S.N. (28) Non-linear pricing and market concentration in the US airline industry. Texas A&M University Working Paper, Texas A&M University, TX, USA. Hayes, K.J. and Ross, L.B. (1998) Is airlines price dispersion the results of careful planning or competitive forces? Review of Industrial Organization 13(5): 523 541. Mantin, B. and Koo, B. (29) Dynamic price dispersion in airline markets. Transportation Research E 45(6): 12 129. Mantin, B. and Koo, B. (21) Weekend effect in airfare pricing. Journal of Air Transport Management 16(1): 48 5. McAfee, P.R. and Velde, V.T. (27) Dynamic pricing in the airline industry. In: T.J. Hendershott (ed.) Handbook on Economics and Information Systems. Amsterdam: Elsevier Handbooks in Information Systems, Volume I. Obermeyer, A., Evangelinos, C. and Puschel, R. (213) Price dispersion and competition in European airline markets. Journal of Air Transport Management 26(c): 31 34. Pels, E. and Rietveld, P. (24) Airline pricing behavior in the London Paris market. Journal of Air Transport Management 1(4): 277 281. Piga, C. and Bachis, E. (27) Pricing strategies by European traditional and low cost airlines: When is it the best time to book online? In: D. Lee (ed.) Advances in Airline Economics: The Economics of Airline Institutions, Operations and Marketing. Amsterdam: Elsevier, pp. 319 344. Read, T.R.C. and Cressie, N.A.C. (1988) Goodness of Fit Statistics for Discrete Multivariate Data. New York: Springer-Verlag. Report of working group on civil aviation sector (212) National Transport Development Policy Committee, Ministry of Civil Aviation. India: Government of India. Sorensen, A.T. (2) Equilibrium price dispersion in retail markets for prescription drugs. Journal of Political Economy 18(4): 833 85. Stavins, J. (21) Price discrimination in the airline market: The effect of market concentration. Review of Economics and Statistics 83(1): 2 22. Su, X. (27) Intertemporal pricing with strategic customer behavior. Management Science 53(5): 726 741. Talluri, K.T. and van Ryzin, G.J. (24) The Theory and Practice of Revenue Management. International Series in Operations Research and Management Science, Vol. 68. New York, USA: Springer. 15