CasereportGroup12.docx

1、CasereportGroup12Case ReportDeltas New SongBy Group 122011/3/29Contents1. Background2. Main issue3. Problems4. Analysis1) High-low for Delta2) Simple Regression for Delta3) Multiple Regression for Delta4) Conditions and assumptions of cost functions5) High-low for JetBlue6) Salary cost estimation fo

2、r Song5. Solution and suggestionBackgroundThe entire US airline business is facing the challenge of operating within a low-margin, high-fixed-cost environment. Its profitability is particularly sensitive to decreases in volume, either from environmental factors or from competition. Moreover, the air

3、line business is labor-intensive. Labor costs as a percentage of revenues ranges from a low of about 25 percent for the low-fare airlines to almost 50 percent for the large, full-service airlines such as United. Furthermore, for many airlines, labor unions are strong, presenting an additional challe

4、nge in the management of cost. Salaries are largely fixed in the short term for unionized employees. In recent years, some major airlines have won concessions from labor unions about reducing the staff and cutting down salaries.Delta Airline is the third largest U.S. airline in operating revenues an

5、d revenue passenger miles flown. Traditionally, the competition came from the other full-service airlines such as United Airlines and American Airlines. However, in recent years, the major airlines have been forced to compete with low-cost, no-frill airlines pioneered by Southwest Airlines. Salaries

6、 are a significant component of Deltas cost structure. Delta pilots are the highest paid in the industry. Meanwhile, it is the least unionized of the major airlines. Delta pilots are the only unionized employee group. Its flight attendant and ticket agents are not under union contract. Consequently,

7、 their salaries present flexibility and variability in nature. In November 2002, despite its previous failure in entry into the low-fare market, Delta decided to form a new low-cost carrier, Song, which is targeted to compete with successful newcomer JetBlue. Main issueHow can Delta create a differe

8、nt cost structure and business model in order to succeed in the low-cost carrier market?ProblemsDelta is in a position of evaluating entry into the low-cost carrier market. The success of Deltas Song depends on the following issues: Can Delta create a very different cost structure and a new business

9、 model to compete with JetBlue? How to predict future salaries of Delta and JetBlue? How to deal with the strong work union?Analysis1. High-low methodFirstly, we identify some possible cost drivers to estimate Deltas salaries: available seat miles; available ton miles; number of departures; revenue

10、passenger miles; revenue ton miles. After applying simple regression using each of the possible drivers and comparing and residual errors of each driver, we choose two most reasonable cost drivers: revenue passenger miles and available ton miles. The salaries consist of payments to pilots, flight at

11、tendants and ticket agents. They are determined not only by the number of passengers and cargoes but also the miles or hours flown. In fact, miles and hours are correlated. So we choose revenue passenger miles and available ton miles as cost drivers. Revenue passenger miles is a major indicator in t

12、he airline industry, so its reasonable to be a driver. Available ton miles, however, seems not so good. But after calculation we find that of the former is 0.1764, and of the latter is 0.5577. For more obvious comparison, we draw the following scatter plots.The scatter plot between revenue passenger

13、 miles and salaryThe scatter plot between available ton miles and salaryWe can see that the latter scatter plot shows a more linear relationship between the two variables. From the perspective of either numerical analysis or visual judgment, available ton miles is more accurate for estimation. So we

14、 choose available ton miles as the cost driver.Low point (3132, 1145), high point (4029, 1514) Salary=0.4114available ton miles-143.50This technique has advantages: only two data is needed, so its quite convenient. Its easy to apply and illustrate mathematically how a change in a cost driver can cha

15、nge total cost.However, not all the information is used, which is regarded as inefficient. Because it bases cost function on only two periods cost experience, regardless of how many relevant data points have been allocated. In a word, its less accurate.2. Simple RegressionWe use simple regression to

16、 estimate the salary cost with available ton miles as the cost driver. Results are as follows:CoefficientsStandard deviationIntercept-682.643282.6033X Variable 10.5516930.079698Salary=0.5517available ton miles-682.64=0.5577, and standard deviations are much smaller than coefficients, so its statisti

17、cally valid and significant.Regression analysis measures cost behavior more reliably than other cost measurement methods, since this technique uses statistics to fit a cost function to all the historical data. Compared with the high-low method which only contains two groups of data, its an improveme

18、nt. Whats more, regression analysis yields important statistical information about the reliability of the cost estimates, which allows analyst to assess confidence in the cost measures and select best cost driver. But we should notice that only one cost driver is considered, so it cant explain the v

19、ariation of salaries completely.3. Multiple RegressionBecause of the reasons mentioned in Question 1, we choose revenue passenger miles and available ton miles as cost drivers and use multiple regression to estimate the salary cost. Results are as follows:CoefficientsStandard deviationIntercept-1144

20、.55243.2101X Variable 11.0519370.120829X Variable 2-72.295514.88974Salary=-1144.55+1.05available ton miles-72.30revenue passenger miles=0.5577, and standard deviations are much smaller than coefficients, so its statistically valid and significant.This technique takes more cost drivers into considera

21、tion, and the results calculated are more close to the data given, so its an improvement over the model estimated in Question 2. The accuracy of estimation thus serves as its top advantage. However, it might be more costly, time-consuming and complicated to be implemented than other techniques.4. Th

22、e usefulness of analysis in Question 1-3The cost functions estimated in Question 1-3 are based on the assumption that the wages per hour remain the same and there is no additional labor needed, so its useful only under certain conditions. Taking the background of the industry and the companys circum

23、stance into consideration, we think these are important: The first one is that the present equilibrium between Delta and the labor unions is not interfered. That is, endeavors concerning lowering pilots salaries to industry level will not be obstructed by union forces, and employees other than pilot

24、s will not join labor unions to require higher payments. If not, adjustments on cost structure will be futile and Song will only turn out to be another Delta Express. Secondly, no harsh regulations regarding reducing staffs or cutting salaries are to be formulated. Restrictions about layoffs will di

25、rectly lead to weak control over budgets, and in turn creates similar problems as high salaries do. Nevertheless, regulations of this kind are almost inevitable. According to precedents, large scale furloughs have already been blocked once by ALPA, thus we have no reason to remain optimistic as long

26、 as the recession of the general economics stay as a fact. The last point we come up with is the new fixed cost caused by new security directives after the September 11 terrorist attacks. However, since security costs can be expected amid the whole industry, it shall not become a major concern for S

27、ong, although our prediction model may overall shift upwards.If the conditions are not met, the cost functions will be less useful.5. Estimate the salary cost for JetBlueAccording to Question 1, available ton miles should be used to estimate the salary cost. However, available ton miles of 2002Q3 is

28、 eccentrically low. So we draw a scatter plot:In this situation, available ton miles and salaries are not linear. The scatter plot of revenue passenger miles and salaries is as follows:Revenue passenger miles and salaries are quite linear. We use the high-low technique to estimate the salary cost wi

29、th revenue passenger miles as cost driver. Low point (599.4, 16000), high point (2016.2, 49000)Salary=23.29revenue passenger miles+2038.836. Estimate the salaries cost for Song in its first yearTo estimate the salaries cost of Song is quite difficult because there is no historical data for reference

30、. Because JetBlue is a successful example in the low-cost market, we use its historical salaries cost to predict Songs salary. For simplicity, we make some assumptions: Song can achieve the same revenue passenger miles as JetBlue in every quarter JetBlues salaries are linear with time seriesThough t

31、he first assumption is very strong, the second one can be easily verified. We number each quarter in 2001 and 2002 from 1 to 8, and make simple regression between the time series and salary.QuarterNumber of time seriesJetBlues Salary2001 Q11160002001 Q22190002001 Q33220002001 Q44280002002 Q153400020

32、02 Q26380002002 Q37420002002 Q4849000Simple regression between the time series and salary:CoefficientsStandard deviationIntercept9571.429986.5881X Variable4761.905195.3735JetBlues salary=$4761.91number of time series+$9571.43The scatter plot of number of time series and salary is as follows:=0.9900 and standard deviations are much smaller than