Chapter 3 Cost Behavior Analysis Leaming Objectives Aer studying this chapter, you willbe able to: 1. Classify an expenditure as fixed, variable, or semivariable. List reasons for separating fixed and variable costs. ‘Compute the fixed and variable components of costs by three methods. Define, compute, and explain the use of the coefficient of determination, Define, compute, and explain the use of the sandard error of the estimate. ‘Compute an estimate of cost using the equation developed by the method of least squares and then compute a confidence intel for that estimate ave ‘Some costs vary in total directly with changes in activity, while others remain relatively ‘unaffected Because ofthe dynamic nature of business, companies often are faced withthe need to make changes in the level and mix oftheir business activities. In order for man- ‘agement to plan a company’s activities inteligently and control its costs effectively, the relationship between cost incurrence and changes in activity must be thoroughly under- stood, This chapter discusses the effect of changes in business activity on costs and classi- fies costs as fixed, variable, or semivariable with respect to activity. In addition, techniques used to segregate the fixed and variable components of costs are illustrated. Although the discussion centers on production costs and activities, the concepts and techniques are ‘equally applicable to marketing and administrative activities. Classifying Cost Success in planning and controlling cost depends on a thorough understanding of the rela- tionship between the incurrence of cost and business activity. Careful stady and analysis ofthe effect of business activities on costs generally will result in the classification of each type of expenditure asa fixed, variable, or semivariable cost, Fixed Cost A fixed cost is defined as one that does not change in total as business activity increases or deereases. Although some kinds of costs have the appearance of being fixed, all costs are variable in the long run. If all business activity decreases to zero and there is no prospect for an increase, a firm will liquidate and avoid all costs. If activity is expected 10 ‘increase beyond the capacity of current facies, fixed costs must be increased to handle the expected increase in volume. For example, factory overhead includes items such as Supervision, depreciation, ent, property insurance, and property taxes—all generally con- sidered fixed costs. If management expects demand forthe company’s products to increase beyond present capacity, it may acquire additional plant, equipment, indirect labor, and aPart 1 > Costs: Concepts and Objectives possibly supervisors to produce the level of output necessary to meet demand. Such addi- tions increase the level of expenditure foreach of these items of factory overhead. For this reason, a particular kind of expenditure shouldbe classified as a fixed cost only within a limited range of activity. This limited range of activity is refered to asthe relevant range. Total fixed cost will change outside the relevant ange of activity. igure 3-1 depicts ‘changes in fixed cost at different levels of activity and the relevant range. FIGURE 3.1 Fixed Cost ‘Some expenditures are fixed asa result of management policy. For example, the level of advertising expenditure and the amount of charitable contribution are determined by ‘management and are not directly related to sales or production activity. Such expenditures are sometimes refered to as discretionary fixed costs or programmed fixed costs Expenditures that require a series of payments over a lng.term period of time are often, called committed fixed costs. Examples include interest on long-term debt and long-term lease rentals Variable Cost ‘A variable cost is defined as one that increases in total proportionately with an increase in activity and decreases proportionately with a decrease in activity. Variable costs include the cost of direct material, direct labor, some supplies, some indirect labor, small tools, rework, and normal spoilage. Variable costs usually can be directly identified with the activity that causes the cost In practice, the relationship between a business activity and the related variable cost ‘usually is treated as if it were linear, that is, total variable cost i assumed to increase by a constant amount for each unit increas in activity. However, the actual relationship is rarely perfectly linear over the entire range of possible activity. Productive efficiency usually cchanges when the work load is very light or very heavy. When the volume of activity increases 1o a certain level, management may add newer, more efficient machinery or replace existing machinery with more productive machinery As a result ofthese factors, the cost per unit of activity usually is differen at widely varied levels of activity. Nevertheless, Within a limited range of activity, the relationship between an activity and the related costChapter 3 > Cost Behavior Analysis 33 ‘may closely approximate linearity. Ths relationship is illustrated in Figure 3-2. The solid line (line B) represents actual variable costs at all levels of activity, and the dashed line (ine A) represents the calculated variable cost a all levels of activity as determined from ‘observations within the relevant range of act w- tort ro ¢ % & » © © © m om w Wo {ASE EXPRESSED A A RELEVANT MEASURE OF ACTIVITY FIGURE 3-2 Variable Cost In cases such asthe one illustrated in Figure 3-2, a constant variable-cost rate usually is a sufficient approximation of the relationship between the variable cost and the related ‘activity within the relevant range. However, to plan and control variable costs effectively, the’ underlying conditions that cause the incurrence of cost should be reviewed frequently to determine whether or not the variable cost per unit of activity has changed. When con- ditions change or the level of activity is outside the relevant range, a new variable-cost rate should be computed. Semivariable Cost A semivariable cost is defined as one that displays both fixed and variable characteristics. [Examples include the cost of electricity, wate, gas, fuel oil, coal, some supplies, mainte- nance, some indirect labor, employee group-tem life insurance, pension cost, payroll taxes, and travel and entertainment. “Two reasons for the semivariable characteristic of some types of expenditures are: L.A minimum of organization may be needed, or a minimum quantity of supplies or services may need to be consumed, t9 maintain readiness to operate, Beyond this minimum level of cost, which is essentially fixed, additional cost varies with volume, 2. Accounting classifications, based on the object of expenditure or function, com- ‘monly group fixed and variable items together. For example, the cost of steam sed for heating, which is dependent on the weather, and the cost of steam used34 Part 1 > Costs: Concepts and Objectives for manufacturing, which is dependent upon the volume of production, may be charged to the same account, resulting in a mixture of fixed and variable costs in the same account! ‘A semivariable cost i ilustrated in the graph in Figure 3-3. The solid line in Figure 3.3 represents actual costs at all levels of activity. In ths illustration, the actual-cost line (ine C) is nonlinear. This situation could occur because of the use of different production techniques or equipment and/or because of different degrees of capacity utilization at. ferent levels of activity. The dashed lines are linear and represent calculated fixed-cost -cost components (line A and line B, respectively) at all levels of activity, as from observations within the relevant range. Estimated total variable cost is the difference between the points on line B and line A. Where line B and line C coincide, the linearity assumption closely approximates the actual relationship. This area of coin- cidence is the relevant range. The use of the caiculated fixed cost and the variable cost rate to estimate cost at any level of activity outside ofthe relevant range would result in unreliable estimates. i a ob wo & & % © w wo ‘ASE EXPRESSED AS A RELEVANT MEASURE OF ACTIOTY FIGURE 3-3 Semivariable Cost Separating Fixed and Variable Costs To plan, analyze control, measure, o evaluate costs at diferent levels of activity, fixed and variable costs must be separated, Those costs that are entirely fixed or entirely variable within the range of activity anticipated must be icentfed, andthe fixed and variable com- ponents of semivariable costs must be estimated. As discussed in later chapters, the Sepa- ration of fixed and variable costs is necessary forthe following purposes: AICI Bu, NL 90, No. 20, pp. 1224-1226,Chapter 3 > Cost Behavior Analysis 3s Predetermined factory overhead rate computation and variance analysis, Fiexible budget preparation and variance analysis. Direct costing and contribution margin analysis. Break-even and cost-volume-profit analysis. Differential and comparative cost analysis. Short-run profit maximization and cost minimization analysis. 7. Capital budgeting analysis. 8. Marketing profitability analysis by territories, products, and customers. an aey In practice, managerial judgment often is used to classify costs as fixed or variable? In such eases, classification is based on te personal experience of management. Although such an approach is expedient, it often results in unreliable estimates of cost. The behav- jor of a particular type of cost is not always readily apparent from casual observation. Furthermore, managers often attempt to simplify the process by classifying each cost as either entirely fixed o entirely variable, thereby ignoring the fact that some costs are semi variable. Generally, more reliable classifications and cost estimates are obtained by using ‘one of the following computational methods al of which ae illustrated in this section: 1) the high and low points method, (2) the seatergraph method, oF (3) the method of last squares. These methods are used not only to estimate the fied and variable components of semivariable costs, but also to determine whether a cost is entirely fixed or entirely vri- able within the relevant range of activity ‘Although the use ofa computational method typically results in amore reliable analy- sis of cost behavior than the simple use of managerial jadgment, the analyst should keep in mind that the results obtained are dependent on historical data If abnormal or unusual ‘conditions occurred during one or more of the periods included in the database, the obser- ‘ations reflecting such abnormalities should be removed from the sample. In this respect, ‘managerial judgment can and shoul play an important roe in cost behavior analysis. For example, taining new employees, a labor strike or work slowdown, a temporary equip- ‘ment failure o te purchase of a batch of substandard materials could distor the relation- ship between an activity and a related cost. To enhance prediction accuracy, the historical database shouldbe inspected by experienced managers, and abnormal observations should. be removed from the sample. Fixed and variable cost estimates based on historical data shouldbe adjusted to reflect, changes that ae expecied to occur during the forecast perio, Technological improvements in production techniques o facilites can affect the behavior of costs. For example, if man- agement acquires (or plans to acquire) new machinery that is expected to operate more efficiently than machinery used during the sample period, cos-behavior estimates based ‘on historical data shouldbe adjusted to reflect the expected improvement in efficiency. Product design changes, as well as changes in production technology, may affect cost behavior, For example, changes in the kinds of materials used in a product may make it necessary to operate machinery ata different speed than that required during the sample period. Tis, in tur, may affect the machinery’ rate of energy or fuel consumption and possibly the amount of preventive maintenance required. To the extent these changes are anticipated, cos-behavior estimates should be adjusted. IF the historical data base includes observations from several diferent years, the ana- lyst should consider the potential distorting effects of inflation. If the rte of inflation was Zana Mone, Acoring te Cote at Ftd td Viti (ree, NS nie ot Managemen Acs (omer Asoo Acta) 68D po OaPart 1 > Costs: Concepts and Objectives substantial during one or more of the periods in the sample, fixed and variable cost estimates are likely to be unreliable. One way to compersate for this problem would be first to ‘estate the cost for each period in the sample to current dollars, and then to perform the analysis on the inflation-adjusted costs. ‘To illustrate the three computational methods of determining the fixed and variable elements of cost, assume that the data presented in Exhibit 3-1 are taken from Barker ‘Company's records forthe preceding year. Electricity Cost and Labor Hour Data | i ge 38 BRGEEE RS. 888 ER ssscsecessestl GEE EXHIBIT 3-1 High and Low Points Method In the high and low points method, the fixed and variable elements of a cost are com- puted from two data points. The data points (periods) selected from the historical data are the periods of highest and lowest activity. These periods usually, but not always, have the highest and lowest figures for the cost being analyzed. Ifthe periods of highest or lowest ‘activity levels ae not the same as those having the highest or lowest level of cost, the activ- ity level should govern the selection, because activity is presumed to drive cost. The high and low periods are selected because they represent conditions for the two activity levels that are the farthest apart. However, care must be taken not to select data points distorted by abnormal conditions. ‘With the data provided in Exhibit 3-1 for Barker Company, the fixed and variable ele- ments are determined as follo Cost ‘Activity Level High. $680 48,000 hours Lam nn 500 28,000 hows Ditterence ——— ze 22,900hous ‘Variable rat: $180 + 22,000 hours = $,00818 per direct labor hourChapter 3 > Cost Behavior Analysis 31 High Low Tota cot 6008500 Taras st ound 3 Fred cost a Direct bor hous x 00818 ‘The high and low activity levels differ by 22,000 direct labor hours, witha cost differ- ‘ence of $180. The assumption is thatthe difference in the costs at the two levels of activ- ity occurred because of differences in the activity being measured and, therefore, is pure variable cost. The variable rate is determined by dividing the difference in cost, $180, by the difference in activity, 22,000 direct labor hours. In this example, the variable rate is Costs: Concepts and Objectives by the majority of data points. Generally, there should be as many data points above as ‘below the line. Another line (line A) is drawn parallel to the base line from the point at Which line B intersects the y-axis, which is read from the scattergraph as approximately ‘$440, This line represents the fixed element of the electricity cost forall activity levels ‘within the relevant range. ‘The area bounded by lines A and B shows the increase in electricity cost as direct labor hhours increase. This increase is computed as follows: erage monty ~ Fed, = Average mony vale element elemento! cost So “Seo $190 , cot cost _ Variable cost per "Average monthly dct aber hous ~ dct labor hour —soflthags ~ $0007 pret borne “Thus the elec cont consits of $40 fed cost er month and variable cost of $0037 pr det abr hour Inthe scaeraraph in Figure 3.4 Bi drawn a «stig ne evn hough te rin dono flow perfect near aterm In most atalyes «stright ne is adequate, Because itis a easonabe approximation of cost havior within the elvan range. 00-4 004 s600-| 0 00 ‘500 xectnerr cost i ‘200 ELEMENT ‘100 ». ‘mECT LABOR HOURS FIGURE 3-4 Scattergraph Representing the Fixed ‘and Variable Elements for Electricity Costs ‘The scattergraph method is an improvement over the high and low points method because it utilizes all available data, not just two data points. In addition, the method ‘makes it possible to inspect the data visually o determine whether r not the cost appears to be related to the activity and whether or not such a relationship is approximately lin- eat. Visual inspection also facilitates detection of abnormal data points (sometimes referred to as outliers). Nevertheless, a cost behavior analysis using the scattergraph ‘method is stil likely to be biased because the cost line drawn through the data plot is based on visual interpretation,(Chapter 3 > Cost Behavior Analysis 39 Method of Least Squares ‘The method of least squares, sometimes called regression analysis, determines mathe- matically a line of best fit, o linear regression line, through a set of points. The regression line minimizes the sum of the squares ofthe deviations of each actual plotted point from the point directly above or below it on the regression line, (Development of the regression ‘equations Barker Company data from Exhibit 3-1 ‘beyond the scope of this textbook.) Exhibit 3-2 illustrates this method using ‘Preparing the table in Exhibit 3-2 requires the following step: 1. Firs, determine the average electricity cost, J, andthe average direct labor hours, 5. Add the observations in columns 1 and 3, and then divide by the number of ‘observations. The average for electricity cost, F, is $570 ($6,840 total electricity cost + 12 months). The average for direct labor hous, ¥, is 35,000 (420,000 total direct labor hours + 12 months). 2 Next, compute the differences by comparing actual monthly electricity cost, y, ‘and direct labor hours, x; to ther respective monthly averages; the monthly aver- ages ae the F and ¥ computed in step 1. These differences are entered in columns 2 and 4 and should sum to zero, unless there is some rounding error. 3. Next, two multiplications must be performed. First, square each ofthe figures in column 4 (x, ~ 3); enter the results in column 5, (x, ~ 2; and total column 5. Second, multiply each ofthe figures in column 4, (x, — x). by the corresponding figures in column 2, (y,~3); enter the products in column 6, (x,~) (,~ 9) and {otal column 6. (Notice that the figures in column 2, (y,~F), are squared also; the results are entered in column 7, (),~J; and column 7 is totaled. Column 7's total will be used in the next section to compute the coefficient of correlation.) nl re Ebsseescessesygt <= 7 EXHIBIT 32310 Part 1 > Costs: Concepts and Objectives ‘The variable rate for electricity cost, b, is cemputed as follows: $2,279,000. 5.9944 por crac labor hour ‘Column 5 total ~ 512,000,000 “The fixed cost, a, can be computed using the formula fora straight line as follows: a+r $570 = 2+ (6.0044)(95,000) $570 = a4 $154 12 = S416 xed element of electricity cost per month ‘These results differ from those determined ty the scatergraph method because fitting 4 line visually through the data points is not as aécurate as fitting a line mathematically ‘The mathematical precisenes ofthe method of last squares injects a high degree of objc- tivity into the analysis. However, itis stil useful to plot the data to verify visually the ex tence ofa linea relationship between the dependent variable and the independent variable. Potting the data makes it easier to spot abnormal data which can distor the least-squares, estimates. If abnormal data are found, they should be removed from the sample data set before using the least-squares formulas. In this illustration, the sample size was small to simplify the computations. In practice, the sample size should be sufficiently lage to rep- resent normal operating conditions Correlation Analysis. The use of the scattergraph method makes it possible visually to determine whether or not there i reasonable dese of corti between the cost and the fctivity being analyzed. In-a statistical sense, creation is a measure ofthe covariation between two variables—the independent variable (x, or direct labor hours in the ilstration) and the dependent variable (yor electricity cost in the istration), In aditon to computing the fied cost and the variable rate fr semivarible costs or the variable apf entirely vari- he costs, the comelation between the independent variable andthe dependent variable should be assessed. Ill pote pont fll on the regression line, perfect coeaton exists. If core- lation is high and the past lationship between the two variables continues inthe Future the activity chosen wl be usefl for predicting future levels ofthe cost being analyzed Mathematical measurements can be used to quantify correlation, In statistical theory, the coefficient of correlation, denoted isa measure ofthe extent to which two variables are related linearly. When r =O, there is no correlation. When r=. the corelaton i per- fect. Ifthe sign of ris positive, the relationship between the dependent variable y, and the independent variable, x, is positive. A postive relationship means the value of y increases a the value of x increases, and the regression line slopes upward tothe right. Ifthe sign of rs negative the relationship between the degendent and independent variables is neg- itive or inverse, which means the value of y decreases asthe value of x increases and the regression line slopes downward to the right. ‘The coefficient of determination is found by squaring the coefficient of correlation. (Derivations ofthe equations forthe coefficient of determination and the coefficient of co- relation ae beyond the scope of this textbook.) The coefficient of determination, denoted a7 is considered easier to interpret than the coefficient of corelation because repre- Sents the percentage ofthe dependent varable's variance thats explained by the independ- ent variable. In this sense, the word explained means the variations in the dependent Variable are elated to, but ot necessarily causedby, the variations in the independent vari- able. Although the cofficient of corelation and he coefficient of determination are math- ematical measures of covariation, they do not establish a cause-and-effect relationshipChapter 3 > Cost Behavior Analysis au between the dependent variable and the independent variable. Such a relationship must be theoretically developed or physically observed. ‘The formula for calculating the coefficient of correlation is: iv tx, — FB, where (x, ~ #) is the difference between each observation of the independent variable (direct labor hours inthe Barker Company illustration) and its average; and (,~ 9) is the difference between each observation of the dependent variable (electricity cost) and its average, The coefficient of correlation, r, and the coefficient of determination, °, forthe data in Exhibit 3-2 are calculated as follows: pe NP), _Coturm tta_ [Eu ca ay,- pF Aol tai coum 7 ta) 2270,000 = 2.0. 49655 [i200 000540, 6005 = 24087 ‘A coefficient of determination that is less than 0.25 means that less than 25 percent of the change in electricity cost is related to the change in direct labor hours. Apparently, the ‘cost in this case is related not only to direct labor hours but to other factors as well, such ‘asthe time of day or the season of the year. Furthermore, some other activity, such as ‘machine hours, may be more closely correlated with electricity cost, thereby providing a better basis for predicting electricity cost. ‘To illustrate a case in which a high degree of correlation exists, the data for electricity cost listed in Exhibit 3-1 are slightly altered in Exhibit 3-3, with direct labor hours remain- ing unchanged. o @), ° , Ce © @, OR mR) EF MD UIP iterenca fom inerence fom ‘average Direct Average ‘18685 Labor of 35,000 Cost Hours @x2) 2) Squared S35 34000 (6.000) 25 (65) 30,000 325000 4.225 5 34.000 (6.000) 25 28 100000 _ 625 as 595000 7225 (a5) 195000 21005 (75) (675,000, 5,625, (105) 948,000 1028 (23) 00000 825, 3) 0228 3 760.900 9.085, a8 1,495,000 13,225 = Suzo.o 53900 EXHIBIT 3.332 Part 1 > Costs: Concepts and Objectives Based on these altered data, the following solution indicates a very high correlation between the two variables, which means that this relationship could be accepted as the bass for calculating electricity cost for planning and control. Box, = My. - 7) Column 6 total Bie, =RF 3y,—9F Clarins total Colune 7 ta —_8120,000__. grag [652,000,000}658 900) = 94991 ‘Standard Error of the Estimate, The regression equation, which in the Barker Company illustration is y/' = $416 + $.0044r,, can be used to predict cost at any level of activity thin the relevant range. However, beceuse the regression equation is determined from & limited sample and because variables that are no: included in the regression equation may hhave some influence on the cost being predicted, the estimated cost will usually be dffer- ‘ent from the actual cost at the same level of activity. The visual scatter around the regres- sion line portrayed in Figure 3-4 illustrates that he actual electricity cost will likely vary from what might be estimated using the calculated fixed cost and the variable cost rate. Because some variation can be expected, management. should determine an acceptable range of tolerance for use in exercising control cver costs. Costs within the limits of vari- ation can be accepted. Costs beyond the limits should be investigated, and any necessary corrective action should be taken. ‘The standard error of the estimate, denoted s', is defined asthe actual data points’ standard deviation from the regression line. A small Value forthe standard error of the esi mate indicates a good fit. When r equals one, the standard error equals zero. Management can use this concept to develop a confidence interval which in tun, can be used to decide “whether a given level of cost variance is likely torequire management action, To illustrate, the table in Exhibit 3-4 can be prepared from the data in Exhibit 3-1 °. GL, Op ewebe one Direct Acta! “bred Prbaihon eieticly Berity ror Squared ‘Cost Cost? (4) Squared seo $565 $5476 eo Ed Stee eo ies Zone 50 ee < io oo toa0t Eo Sr 79 eo 50 oo So So fo = ie ioe 0 a ‘co So tos ees 0 827 2.009 son S20 Sos * Caloaied regression Ine, y. value, (rect bor hours» $ 0044) + $476, are rounded to te nearest dolar. “Th nn ol coke (ee 20, wet rng EXHIBIT 3-4(Chapter 3 > Cost Behavior Analysis 313 Based on the computations in Exhibit 3-4, the standard error of the estimate is then calculated as follows: fu [Column 5 total ot. Tepes th ey oe lnenrad cbc Yano, for smal samples, the Student’ itribuion is a more appropriate assumption. table of selected vales, based onthe assumption that two tal ofthe distribution are of concer, ie. that managers are concemed about both favorable and unfavorable variances, is pro- vided in Exhibit 3-5, $55.34 “Table of Selected Values of Student's Distribution Desired Confidence Level Degrees of Freedom 0% 2% ome 98.8% 1 bar 2706 jaesr 8310 2 2920 ‘4303 ‘9925 "22.308 3 2353 3.182 5841 10213 4 Dis 2776 4604 7473 5 2015 2sr ‘ose 5.903 6 1943 2aa7 3707 5.208 7 1805 2365 3.499 4785 8 11360 2308 3.955 ‘4501 9 1835 2282 3.250 4297 10 isi2 2208 3.169 ata 1 1796 2a01 3.108 4.025 2 1782 2179 3.055 3.890 3 azn 2160 3012 2.052 4 1761 2145 2or7 3787 15 1753 2i31 2087 3733 2 1705 2.086 2oas 3.882 25 1708 2.060 2787 3.450 30 1897 2012 2750 3.385, «0 1684 2001 2704 3307 60 tert 2.000 2.560 3.232 120 1858 11980 2817 3.160 ea 1645 1.960 2578 3.000 EXHIBIT 3.5 ‘The acceptable range of actual cost around the predicted cost would be computed fora sample of size n by multiplying the standard error ofthe estimate by thet value for n ~ 2 degrees of freedom atthe desired confidence level ,. (Degrees of freedom, denoted df, refers othe number of values that ae free to vay after certain restrictions ae placed on the data. In general, if a regression equation involves p unknown parameters, then df = n ~ p. {In linear bivariate regression there are two unknown parameters, a and b; thus, df= n — 2.) For small samples, the results then multiplied by a correction factor, calculated as follows nda i fed 1, oo34 Part 1 > Costs: Concepts and Objectives where all variables are as previously defined. To illustrate the computation and use ofthe confidence interval, assume that the actual level of activity for a period is 40,000 direct labor hours. The electricity cost computed for tne budget from the regression equation determined in the previous example is $592 [S416 + (S.0044 x 40,000)]. Assume further that management wants to be 95 percent confidert that the actual electricity cost is within, acceptable tolerance limits. Based on the table factor of 2.228 for fat the 95-percent con- fidence level, with df= 12 ~ 2, and on the standard error of the estimate computed above '$55.34), the confidence interval would be: iw PEE rac se000=(2zns8.an foo OOO ROE $582.00 = (2228),$55.34)(1.064) $592.00 2 $191.19 “Management can expect the actual electricity cost to be between $460.81 ($592.00 — $131.19) and $723.19 ($592.00 + $131.19) about 95 percent of the time. Electricity cost ‘outside these limits will occur at random only 5 percent ofthe time. Ifthe actual electric- ity cost is less than $460.81 or greater than $723.19, management should investigate the cause and take any necessary corrective action. For large samples, the Student's «distribution approaches the normal distribution and the correction factor for small samples (the square-root term) approaches one. For large samples, therefore, the computation of the acceplable range of actual cost around the pre- dicted cost may be simplified by omitting the correction factor and using the appropriate ‘value for the normal distribution. (The values presented in Exhibit 3-5 for df'= co are ‘equal tothe z values for the normal distribution.) If the sample size used in computing the regression equation and the standard error of the estimate in the illustration were large, the ‘95-percent confidence interval for electricity cost at 40,000 direct labor hours would be: $592.00 + (1.960(855.34) ‘$662.00 + $108.47 After the fixed and variable components cf cost have: been computed using the ‘method of least squares, itis useful to plot the regression line against the sample data, ‘0 that the pattern of deviations ofthe actual observations from the corresponding esti- ‘mates on the regression line can be inspected. Normally, the distribution of observations around the regression line should be uniform for all values of the independent variable (referred to as homoscedastic) and randomly dstributed around the regression line as depicted in Figure 3-5. However, ifthe variance differs at different points on the regres- sion line (referred to as heteroscedastic), as ‘depicted in Figure 3-6, or if the observa- tions around the regression line appear to be correlated with one another {referred to as serial correlation or autocorrelation), as depicted in Figure 3-7, the standard error of the estimate and the confidence intervals based on the standard error are unreliable smeasures.? 3 nao. rennan and Thomas M, Carol, Petace fo Quattatne Economics and Econometric (Cnn Sout Western, 1987), Chaplo 19.Chapter 3 > Cost Behavior Analysis. 31s FIGURE 3-5 Homoscedasticity (Random Error Term) FIGURE 37 Serial Correlation3.16 Summary Part 1 > Costs: Concepts and Objectives Method of Least Squares for Multiple Independent Variables ‘Typically, cost behavior is treated as being dependent on a single measure of volume or on ‘some other independent variable. Inthe preceding discussion, for example, the behavior of, the dependent variable, electricity cost, was described by the independent variable, direct labor hours. However, a cost can vary because cf more than one factor. “Multiple Regression Analysis is a further application and expansion of the method of least squares, permitting consideration of more than one independent variable, In ‘multiple regression analysis, the least squares equation for a straight line, y, = a + 8.x + js expanded to include more than one independent variable. For example, in the equation 1y, + bx; cz, + ¢ ¢is the variable rate for an additional independent variable z. Although the cost relationship can no longer be shown on a two-dimensional graph and the arith- ‘metical computations become more complex, the widespread availability of computer pro- ‘grams makes the use of multiple regression analysis feasible. ‘The least-squares concept is fundamentally the same when there are two or more inde- pendent variables as when there is only one. The assumption of normality still applies. In the case of multiple regression, however, itis the joint probability distribution of the var ables that is assumed to be normally distributed (sometimes referred to as multivariate nor- ‘mal). One additional assumption is that the independent variables are not correlated with ‘one another. When the independent variables are correlated with one another, they are said to be collinear, a condition referred to as multicollinearity. When the degree of muti collinearity is high, the relationship between one or more of the independent variables and the dependent variable may be obscured The presence of multcollinearity would not affect the estimate of cost unless one or more important independent variables—activity ‘measures—were omitted from the regression model because of an apparent lack of rela- tionship tothe dependent variable (cost). Omitting important variables from the multiple- regression model is referred to as specification error and is more of a problem in analyzing the sources of cost incurrence than in estimating cost If the cost behavior of a group of expenditures in one or more ledger accounts is being. described, an alternate to multiple variables (and hence to the considerations necessitated bby multiple variables when applying the least squares method) may be possible. That is, expenses may be grouped and classified in sufficient detail so that expenses ina particular {group are all largely related to only one independent variable. This would allow the use of the method of least squares as earlier illustrated, that i, simple regression analysis. If this approach is not feasible, as when more than one independent variable is still requi describe the cost behavier, then multiple regression analysis should be employed.> Because of the dynamic nature of business, companies often are faced with the need 10 ‘make changes in the level and mix of their business activities. For management to plan a company's activites intelligently and contol its costs effectively, the relationship of cost incurrence to changes in activity must be understood thoroughly. This chapter discussed the effect of changes in business activity on costs and classified costs as fixed, variable, or semi- ‘variable with respect to activity. Three techniques for segregating the fixed and variable i oe(Chapter 3 > Cost Behavior Analysis, 37 Key Terms components of costs were illustrated: the high and low points method, the scatergraph ‘method, and the method of least squares. The most accurate method is the method of least- squares; however, the reliability of least-squares cost estimates depends on the correlation between the activity and the cost being analyzed. The coefficient of corelation should be used to find the most accurate predictor. In additio, the standard error ofthe estimate can be ‘computed on the basis of least-squares estimates and used to develop a confidence interval for cost contol. Although the discussion centered on production costs and activities, the con- cepts and techniques are equally applicable to marketing and administrative activities. fixed cost (3-1) relevant range (3-2) discretionary fixed costs or programmed fixed costs (3-2) ‘committed fixed costs (3-2) variable cost (3-2) semivariable cost (3-3) high and low points method (3-6) seattergraph method (3-7) ‘method of least squares or regression analysis (3-9) correlation (3-10) ‘coficient of correlation (3-10) coefficient of determination (3-10) standard error ofthe estimate (3-12) hhomoscedastic (3-14) hheteroscedastic (3-14) serial correlation or autocorrelation (3-14) ‘multiple regression analysis (3-16) ‘ulticollinearity (3-16) specification error (3-16) Discussion Questions Qt 32 33 B+ Bs 03-6 Explain the difference between fixed, variable, and semivariable costs ‘What is meant by the term relevant range? ‘Why should semivariable costs be segregated into their fixed and variable components? Discuss the advantage and the disadvantage of using managerial judgment to separate fined and variable cost. ‘What computational methods are available for separating the fixed and variable compo- nents of semivariable costs? ‘What are the advantages and disadvantages of exch of the three different methods of sep- araticg fixed and variable components of costs?38. Qs 39 93-10 ou Q3-12 Exercises Part 1 > Costs: Concepts and Objectives Explain the meaning of the $200 and the $4 inthe regression equation, y= $200 + Sx, ‘where y, denotes the total monthly cost of indirect supplies and x, is machine hours per ‘month, (ICMA adapted) Define and explain the difference between the coefficient of correlation and the coeffi- cient of determination ‘What i the standard error ofthe estimate? With respect to the method of least squares, what is meant by the term heteroscedasticity, and what problem is likely to occur if itis present? Define and explain the significance of the term serial correlation ‘What is meant by the term multcollinearity asi is applied in multiple regression analysis? ‘High and Low Points Method. Maddleson Incorporated wants to segregate the fixed and variable portions of maintenance cost, believed to be a semivariable cost, as meas- ‘ured against machine hours. The following information has been provided for the most recent six months: Machine Maintenance Month. Hours Goat March 2.550 $1275 owl. 2.300 1,200 May. = 2.100 4400 tone: 2600 1300 uy. 2.350 1205 Ugur 21460 11250 Required: Using te high and low pois method, compute the variable cost rate and the fixed cost for maintenance cost Statistical Seattergraph. Femwood Company's production management is interested in determining the fixed and variable components of supplies cos, a semivariable cost, as meas- tured against direct labor hours, Data forthe first ten months of the current year follow: Labor Supptes Monn Hours oat sanuary 450 $600 February a5 700 March 500 750 ‘ar. 550 650 May 75 800 shine : 750 800 ay 05 825 August 325, 75 Sopiombar =. 600 75 sober. 025 850(Chapter 3 > Cost Behavior Analysis, 319 Required: supplies cost raph the data provided and determine the fixed and variable components of “Method of Least Squares. Poplar Incorporated's management is imerested in predicting travel and entertainment expense based on the number of expected sales calls on cus- tomers. Over the past 50 weeks, the company's Sales Department reported that its sales force made 6,250 calls on customers, an average of 125 calls per week. Travel and enter- tainment expenses over the same period totaled $500,000, an average of $10,000 per week. ‘The travel and entertainment expense deviations from its average multiplied by the sales call deviations from its average and summed (S{ x, - 31, ~ 9)] is 87,000, and the sales call deviations from its average squared and summed (4, — 9] is 1,450. Required: Using the method of least squares, estimate the total cost of travel and enter- tainment for a week in which the company’s sales personnel make 200 sales calls. Method of Least Squares. The following data were collected for the most recent twelve- ‘month period by the cost accountant in one of Richardson Company's manufacturing plants. Machine Month sanuary st February 1510 March 11500 2600 Apa 1450 2.500 May 460 2510 sure. 1520 2610 ‘by. 1570 2750 ‘august. 1530 2700 ‘Septombor 1480 2530 ‘October. 1470 2520 November 1450 2.490 December 1460 2'520 Required: Using the method of least squares, compute the fixed cost and the variable cost rate for electricity cost. Round estimates to the nearest cent. Correlation Analysis. The controller of Yazoo Electronic Company would like to know hhow closely the incurrence of factory overhead in the company's Semiconductor “Manufacturing Department correlates with machine hours. Total factory overhead over the ‘past 12 months is $108,000, and 5,760 machine hours were logged in fr the same period. ‘The sum of the squares of the machine hour differences from average (3(x,—)"] is 850, and the sum of the squares ofthe factory overhead cost differences from average [S.(, — ¥?’] is 3,400. The machine hour differences from average multiplied by the factory over- head cost differences from average and summed [( x= 3), ~y)] is 1.564. Required: Compute the coefficient of corelation, r, andthe coefficient of determination, 2 for factory overhead and machine hours in the Semiconductor Manufacturing Department. Correlation Analysis. The management of Gibralter Inc. collected the following data over the most recent 12 months:Part 1 > Costs: Concepts and Objectives Shipping —_Salen Month ‘Expenses Revenue druary $500 «$25,500 February 00 20,000 arch 00 2000 20 38.000 vy 570 F000 ne 0 35.500 sy = 30.000 ogust so son September 30 ¥ ‘dober = 200 November 0 20800 December 0 500 1 of correlation, r, and the coefficient of determination, 7? for shipping expense and sales revenue Cost Behavior and Correlation Analysis. Parsons Company's total maintenance cost for the past 10 months is $50,000. Some of the maintenance activity appears related to the ‘operation of machinery, and the accountant wants to determine whether machine hours should be used as a basis upon which to estimate maintenance cost at various levels of ‘capacity. Machine hours for the same period totaled 40,000 hours. The machine hour dif- ferences from average multiplied by the maintenance cost differences from its average and summed (3 x, ~ 7), ~J)] is 2,400. The machine hour differences from average squared and summed (5x, ~ 3°] is 6.250, and the maintenance cost differences from average squared and summed (10, ~ 3) is 1,000. Required: (1) Compute the coefficient of correlation, r, andthe coefficient of determi ‘maintenance cost and machine hours. (2) Compute the variable maintenance cost per machine hour, using the method of least squares. (3) Compute the fixed maintenance cost, using the method of least squares. 1 P. for ‘Choosing Appropriate Activity Measure and Separating Fixed and Variable Costs. ‘Thompson Company's total electricity cot forthe past 20 months is $42,000. An activity measure on which to base estimates of electricity cost is needed. The two activity measures being considered are direct labor hous and machine hours. Direct labor hours for the period totaled 180,000, and machine hours totaled 120,000. The direct labor hour differences from average multiplied by the electricity cost differences from its average and summed [3ts, -0Q, —7)] i 5,700. The direct labor hour differences from average squared and summed (3x, -¥?] is 28,500. The electricity cost differences from average squared and summed (530, ~ 371 is 1.264. The machine hour differences from average multiplied by the electricity cost differences from its average and summed (3x, ~ 3), ~ J)] is 7,000. ‘The machine hour differences from average squared and summed (5(x,~ 7) is $0,000. Required: (1) Compute the coefficient of correlation, r, and the coefficient of determination, +, for direct labor hours and electricity cost.Chapter 3 > Cost Behavior Analysis 321 E310 Problems P34 2) Compute the coefficient of correlation, r, and the coefficient of determination, 2, for ‘machine hours and electricity cost. (G3) Which activity measure should be used for the estimation of fixed and variable elec- tricity cost? Explain, (4) Compute the variable electricity cost rate andthe fixed cost, using the method of least squares. Standard Error of the Estimate. The controller of Newsborg Company used the method of least squares to compute the fired and variable components of utility cost from the fol lowing data: vty Gost ‘$3,600 "4.000 +4000 3.800 3,700 31500 3.900 4,100 4500 4200 4.300 4400 Required: Assuming thatthe leat squares estimate of fixed cost is $1,000 and the vari- able rate is $1.00, compute the standard error of the estimate to three decimal places. Standard Error of the Estimate and Confidence Interval Estimation. The production supervisor of Zeus Inc. woud like to know the range of maintenance cost that should be ‘expected about 90% of the time at the 1,500-machine-hour level of activity. The least squares estimate of maintenance cost at that level of activity is $500. The least squares parameters estimates—the estimates of fixed cost and the variable cost rate—were derived from a sample of data for a recent 15-month period. The machine hour average for the sample period is 1,300, and the machine hour deviations from its average squared and summed [5( x, -¥)1 is 150,000. The prediction error squared [(),~y,!] over the same sample petiod is $49,972. Required: Compute the standard error of the estimate and the 90% confidence interval estimate for maintenance costa the 1,500-machine-hour level of activity Correlation Analysis. The Cost Department of Kemp Supply Company altemps to estab- li budget ost inthe cond of marking expenses. An examination of niu expenses shows32 Part 1 > Costs: Concepts and Objectives Flood Variable ‘tern Portion Portion. ‘Sales stat: Salaries ‘$1,200 none Retainers 2.000 none Commissions none Ybor sales savertsing, 5,000 none ‘Travel expence 2 ° ‘Statistical analysis is needed to split the travel expense satisfactorily into its fixed and variable portions. Before such an analysis is begun, itis thought that the variable portion of the travel expense might vary in accordance ether with the number of calls made on ‘customers each month or the valve of orders received each month. Records reveal the fol- lowing details over the past twelve months: geeaseeeaes Required: (1) Compute the coefficient of correlation, r and coefficient of determination, 7, between (@) the travel expense and the number of calls made and (b) the travel expense and ‘orders received. Round to four decimal places. (2) Compare the answers obtained in requirements la and 1b, ‘Choosing Appropriate Activity Measure; Cost Behavior Analysis. The controller of ‘Terrance Corporation has asked for help in the selection of the appropriate activity meas- ture to be used in estimating variable supplies cost forthe company's budget. The follow- ing information about past expenses and two poteatal activity measures has been supplied: Supplies Labor Machine Month Gost Hours ‘Hours anuary $ 1.505, 5,000 2.000 February, = 1395 4.800 4900 March a 1565 5,160 2140 Apt et 1515 x May. 145 ar was sy 1465 ‘August. 11505 September 1575 dtober. 1595 November 1500 December 4 biChapter 3 > Cost Behavior Analysis 33 Required: (1) Compute the coefficient of correlation, r, and the coefficient of determination, 2, between supplies cost and each ofthe two activity measures. (2) Identify which of the two activity measures should be used as a basis upon which to estimate the allowable supplies cost, (3) Using the activity measure selected in requirement 2 above, determine the fixed cost and the variable cost rate by the method of last squares. ‘Choosing Appropriate Activity Measure; Cost Behavior Analysis. The management of the Randolph Corporation has asked for help inthe selection ofthe appropriate activity meas- tre to be used in estimating electricity cost in one of its manufacturing plants. The following information about past expenses and two potential activity measures has been supplied: (Q) Compute the coefficient of corelation, r, and the coefficient of determination, 77, between the cost of electricity and each of the two activity measures. (2) Identify which ofthe two activity measures should be used as a basis on which to esti- ‘mate the allowable cost of electric (3) Using the activity measure selected in requirement 2, compute an est electricity cost and the variable electricity rate by the method of least squares. Cost Behavior Analysis; Correlation Analysis; Standard Error of the Estimate. The following data have been collected by the controller of Xanthan Corporation over the past 12. months: Waintenance Machine oat ‘Hours, $ 2200 2.500 2190 2380 2.000 2.000 2.170 2.400 2.050 2.100 2720 21600 2150 2.450 2250 2.550 2200 2,700 2.180 2.450 2210 2400 21100 2.300 sso 28,80034 P35 Part 1 > Costs: Concepts and Objectives Required: (1) Using the high and low points method, determine the average of fixed maintenance cost per month and the variable maintenance rate per machine hour. (2) Determine the fixed cost and the variable cos rate, using the method of least squares. (3) Compute the coefficient of correlation, r, and the coefficient of determination, 72, ‘between the machine hours and the maintenance cost. (4) Compute the standard error ofthe estimate. (5) Determine the 95% confidence interval for maintenance cost at the 2,500-machine- hour level of activity ‘Three Methods of Cost Behavior Analysis; Correlation Analysis; Standard Error of the Estimate. The management of Welcome Hotel is interested in an analysis ofthe fixed ‘and variable costs in the electricity used relative to hotel occupancy. The following data Ihave been gathered from records forthe year: sf B8seess Ebazee Required (1) Determine the fixed and variable elements of lectricity cost using each of the following methods. Round the fixed cost to the nearest dolar andthe variable rat to four decimal places. (a) the method of least squares (©) the high and low points method (©) a scattergraph with trend line fited by inspection (2) Compute the coefficient of correlation, r, andthe coefficient of determination, 2, for ‘guest days and electricity cost. (3) Compate te standard error ofthe estimate. (4) Compute the 90% confidence interval for elecicity cost at the 2,000-guest-days capacity ‘Three Methods of Cost Behavior Analysis; Corelation Analysis; Standard Error ofthe Estimate. Sipenstein Company makes tubing from aluminum billets using a process in ‘hich the billets are heated by induction toa very high temperature before being pu through an extruding machine that shapes the tubing from the heated billets. The inducer, avery large coil into which the billet is placed, must sustain 2 great flow of electric current to heat the billets tothe necessary temperature. Regardless ofthe numberof billets to be processed, the coil is kept on during the entie operating day because of the time involved in starting it up. ‘The Cost Accounting Department wants to charge the variable electricity cost to each billet and the fixed electricity cost to factory overhead. The following data have been assembled:Chapter 3 > Cost Behavior Analysis 3.28 Number _Costot of Billets, Electricity 2.000, 55 ‘900 “0 900 ss 2300 es 200 a 200 oe ‘400 ‘co 1900 ‘so 1500 3 200 00 20 rr 2200 40 2 Required: (1) Determine the fixed and variable elements ofthe electricity cos using each ofthe fol lowing methods: (a) the high and low points method (b) a scattergraph with (rend line fitted by inspection (6) the method of least squares (2) Compute the coefficient of correlation, r, and the coefficient of determination, 7. (3) Compute the standard error ofthe estimate, and the 95% confidence interval for elec- Uwicty cost atthe 2,200-billets level of activity P3-7 Cost Behavior Analysis; Correlation Analysis; Standard Error of the Estimate. Gibson Company manufactures a wide range of electrical products at several different plant locations. Due to fluctuations, its Lanford plant has been experiencing difficulties in estimating the level of monthly overhead. Management needs more accurate estimates to plan its operational an financial needs. ‘A trade association publication indicates tat for similar plans. overhead tends to vary with direct labor hours. Based on this information, one member of the accounting sta proposes thatthe overhead cost behavior pattem be determined to calculate the overhead cos in rela- tion to budgeted direct labor hours. Another member of the accounting staff suggest that a ‘good starting place for determining the cost behevior pattern of the overhead cost would be an analysis of historical data to provide a bass for estimating future overhead costs. Direct labor hours and te respective factory overhead costs forthe past three years are as follows: — 208 ___0¢__ Direct Direct Direct Factory Labor Labor Labor Overhead Month. Hours. Hours, Hours “Coste January. 2,100 2000 $8,600 February 400 25300 200 "9.300 Mare 200 2.200 2300 9.400 Apa 200 2200 2200 8,700 May. 000 2000 2000 8.100 ne 11900 41800 900 7,600, ay 1400 11200 4300 7000, ‘maga 1000 1300 120 6:00, September. 120 11500 ‘300 7.100, ‘Octobe 1700 4700 1800 7500, Novena 1500 11500 41500 7,000 December 1900 1800 1900 7,600Part 1 > Costs: Concepts and Objectives Required: (1) Compute the amount of fixed factory overbead and the variable cost rate, using the method of least squares. Round fixed factory overhead to the nearest dollar and the variable cost rate fo the nearest cent. (2) Compute the coefficient of correlation, r, and the coefficient of determination, 2, for factory overhead costs and direct labor hours. (Round to four decimal places.) (@) Compute the standard error ofthe estimate. (Round to the nearest dollar.) (4) Compute the 95% confidence interval for factory overhead costs at the 2,200-direct- labor-hour level of activity. (Assume that the sample size is sufficiently large thatthe ‘normal probability distribution can be assumed and thatthe correction facto for small ‘samples can be omitted. Round to the nearest dollar.) (Choosing Appropriate Activity Measure; Cost Behavior Analysis; Standard Error of the Estimate. Based on observation and knowledge of the production process, the con- troller of Nellysford Company believes thatthe bulk of variable maintenance cost is driven by either labor hours or machine hours. Data gathered for the past 24 months follows: Machine ‘Hours ‘209 78 oer a9 987 1618 1,188 1525, 1154 1887 01 11550 1238 1585 13186 1875 1208 1405 ‘oor 1281 eat ‘350 502 1175 733 1438 +4080 11506 1138 11508 174 11653 1248 1675 1266 1728 1333 1628 1305 11230 1575 166 1365 1853 1373 1237 1418 128 11935 (1) Compute the correlation coefficient, x and the coefficient of determination, 12, for ‘maintenance cost with each ofthe (wo activiy measures suggested by the controller, (2) Using the method of least squares and the activity measure most highly correlated with ‘maintenance cost, compute the variable maintenance cost rate and the estimated fixed ‘maintenance cost. (G) Compute the standard error of the estimate using the variable rate and estimated fixed cost computed in requirement 2 (4) Compute the 95% confidence interval estimate at the 1,100-hour level of activity. ‘Student’ 1 value at the 95% level of confidence fora two-tal test is 2.074 for 22 degrees of freedom, 2.069 for 23 degrees of freedom, and 2.064 for 24 degrees of freedom.Chapter 3 > Cost Behavior Analysis 37 Cases C31 Cost Behavior Analysis Using Method of Least Squares. Samuelson Inc. is a major book distributor that ships books throughout the United States. Samuelson's Shipping Department consists of a manager plus 10 other permanent postions —four supervisors and six loaders ‘The four supervisors and six loaders provide the minimum staff and frequently must be sup- plemented by additional workers during weeks when the volume of shipments is heavy. Thus the numberof persons shipping the orders frequemaly averages over 30 per week —10 perma- nent plus 20 temporary workers. The temporary workers are hired through a local agency. Samuelson must use temporary workers to maintain a minimum daily shipment rate of 95% of orders presented for shipping. The loss of efficiency from using temporary workers ‘minimal, and the $10.00-per-hour cost of temporary workers is less than the $15.00 per hour pad forthe loaders and $22.50 per hour paid forthe supervisors on Samuelson's per- ‘manent staff. The agency requires Samuelson to atilze each temporary worker for atleast four hours each day. LLocter, shipping manager, schedules temporary help based on forecasted orders for the coming week. Supervisors serve as loaders until temporary help is needed. A super- visor stops loading when the ratio of loaders to supervisors reaches 7 to 1. Locter knows that he will need temporary help when the forecasted average daily orders exceed 300. LLocter frequently requested from two to four entra temporary workers per day to guard against unexpected rush orders. If there is not enough work, he dismisses the extra people at noon after four hours of work. The agency is not pleased with Locter’s practice of over- hiring and notified Samuelson that itis changing its policy. From now on, if « worker is dismissed before an eight-hour assignment is completed, Samuelson will still be charged for an eight-hour day plus mileage back to the agency for reassignment. This policy is to £0 into effect the following week. Paula Brand, general manager, called Jim Locter to her office when she received the notice from the agency. She told Locter, “Your staffing levels have to be better. This penalty could cost us up to $500 per week in lator cost for which we receive no benefit. ‘Why can’t you schedule more accurately?” LLocter replied, “I agree thatthe staffing should be better, but I can't do it accurately ‘when there are rush orders. By being able to lay off people at noon, I have been able 10 adjust for the uncertain order schedule without cost to the company. Of course, the agency's new policy changes this” LLocter and Brand contacted Samvelson's controller, Mitch Berg, regarding the prob- Jem of how to estimate the number of workers needed each week. Berg realized that Locter needs a quick solution until he can study the work flow. Berg suggested a regression analy- sis using the number of orders shipped as the independent variable and the number of ‘workers (permanent plus temporary) asthe dependent variable. Berg indicated that data for the past year are available and that the analysis could be done quickly on the Accounting Department's microcomputer. ‘Berg completed the two regression analyses that follow. The firs regression is based (on the data for the entire year. The second regression excludes the weeks when only the 10 permanent staff persons worked; these weeks were unusual and appear to be out of the rel- evant range.Part 1 > Costs: Concepts and Objectives occ ee Date, ‘Dally Dats zoe ae 3 ‘ee 2ore ase see ove Locter is not familiar with regression analysis and is, therefore, unsure how to imple ‘ment this technique. He wonders which regression data he should employ; that is, which data are better. When Locter recognized that the regression is based on actual orders ipped by week, Berg told him he can use the forecasted shipments for the week to deter- mine the number of workers needed, Required: (J) Using Regression 1 based on data from afl year, calculate the numberof temporary Workers who shouldbe hired fora forecast of 1,200 shipments (2) Which one of the two regressions appears tobe beter? Explain. (@) Explain the circumstances under which this regression should be used in planing for temporary workers. (4) Explain how the regression might be improved. (ICMA adapted) C32 Cost Behavior Analysis; Correlation Analysis; Standard Error of the Estimate, ‘Albama Company's Cost Department has compiled weekly records of production volume (Gn units), electric power used, and direct labor hours employed. The range of output for which the following statistics were computed is from $00 to 2,000 units per work: tectrle Power: 2 1,000 + 4x, where yi electic powpr and x's units of production Standard enor ofthe estimate: 100 Gootllnt of correlation: 45. Direct labor: Y= 100 + 1.2, where yi drect labor hours and xis units of production ‘Siandard ertor of th estima: 300 Coefficient of correlation: 70 Required: (1) Compute the best estimate of the additional number of required direct labor hours, if ‘production for the next period is 500 units greater than production inthis period. (2) Comment on the reliability of the above equations for estimating electric power and direct labor requirements, together with the necessary assumptions if the estimating ‘equations are to be used to predict future requirements. An interpretation of the coef ficient of correlation and the standard error ofthe estimate should be included. ‘CGA-Canada (adapted). Reprint with permission.Chapter 3 > Cost Behavior Analysis 39 C33 Cost Behavior Analysis; Alternative Models. Motorco Corporation plans to acquire sev- eral retail antomotive pars stores as part of its expansion program. Motorco carries out extensive review of possible acquisitions prior to making any decision to approach a spe- cific company. Projections of future financial performance are one of the aspect of such 4 review. One form of projection relies heavily on using past performance (normally 10 prio years) to estimate future performance. Currently, Motoreo is conducting a preacquistion review of Alpha Auto Parts, a regional chain of retail automotive pars stores. Among the financial data to be projected for Alpha are the future rental costs for its stores. The following schedule presents the rent and revenues (in millions of dollars) fr the past 10 years: ‘Annual Rent Havens Expense. $1.00 118: 140 110 155 128 185, 180 180 195 BSSBaR SSeS sesaessesh ‘The following thre alternative methods of estimating future ental expense are being considered: ‘Alternative A: A liner regression was performed using times asthe independent vari- able. The resulting formula i as follows: Rental expense = 63 + .0896y ‘308, ‘Standard eror of he ostinato 150 where ys equal 10 (actual year ~ 208); for 20) = 10 ‘Alternative B: Toe annual rental expense was related to annual revenues through linear regression. The formula for predicting rental expense in this case is s follows: ntl expense = 5507 + 02219) 978 ‘Standard oor ofthe estimate = 070 here y is equal a (Revenues + 1,000,000); e.g. for 204 is 60. ‘Alternative C: The third alternative is to calculate rental expense as a percent of rev- cnues using the arithmetical average forthe 10-year period of 20201 inclusive. The for- ‘mula for predicting rental expense in tis case is as follows: ental expense = GE + SRy= (1448 + 300yy say where SE is equal to the sum of the rental expenses for the 10-year period, SR is equal to the sum of the revenues forthe 10-year period, and y is equal to revenue ia te predic- tion year; e.g. y for 201 i 60.30, 3s Part 1 > Costs: Concepts and Objectives Required: (1) Discuss the advantages and disadvantages of each of the three alt_mative methods for estimating the rental expense for Alpha Auto Parts. (2) Identify one method from alternatives A, B, or C that Motorco should use to estimate rental expense and explain why that alternative is preferred. (@) Explain whether a statistical technique is an appropriate method in this situation for estimating rental expense. (ICMA adapted) Multiple Regression Analysis. Multiple regression is a procedure used to measure the relationship of one variable with two or more other variables. Regression provides a rela- tional statement rather than a causal statement with regard to the relationship. The basic formula for a multiple regression equation is: Yo=as brs cz,+6, For a regression equation to provide meaningful information, it should comply with the basic criteria of goodness of fit and specification analysis. Specification analysis is determined by examining the data and the relationships of the variables for (a) linearity within a relevant range, (b) constant variance of error terms (homoscedastcity),(c) inde- pendence of observations (serial corelation), (4) normality, and (e) multicollinearty. Required: (1) Explain what is meant by the following: “Regression provides a relational staement rather than a causal statement.” (2) Explain the meaning of each of the symbols that appears inthe basic formula of the ‘multiple regression equation just stated. (3) Identify the statistical factors used to testa regression equation for goodness of fit and, for each item identified, indicate whether a high value ora low value describes a good fit. (4) Explain what each ofthe folowing terms means with respect to regression analysis: () Linearity within a relevant range (b) Constant variance (homoscedasicty) (6) Serial coreation (@) Normality (€) Multicottinearity (ICMA adapted) ‘Multiple Regression and Correlation Analysis. John Wood, a financial analyst for a ‘major automobile corporation, has been monitoring the funds used in advertising cam- ‘paigns and the funds used for automobile factory rebates. Financial data have been accu ‘mulated forthe lst 24 months along with customer sales, thats, automobiles sold. Wood contends that there may be a relationship between the level of automobile sales and funds ‘expended on advertising and/or factory rebates. If such a relationship can be determined, the company may be able to estimate sales demand based on various levels of funding ‘commitments for one or both types of expenditures. Regression equations and supporting statistical values developed forthe various rela- tionships between variables areas follows:Chapter 3 > Cost Behavior Analysis 3at Standard Error of the Estimate =e Equation 1 1.325 Equation 2 1436 Equation 3 2.202 Equation $ D=-.1844 186A +.497R ‘22 where the notations used in the equations are as follows: ‘A= advertising funds in $100,000 R= funds for factory rebates in $1,000,000 = customer sales demand (automobiles sold) in 10,000 units ‘The appropriate ¢ values for use in determining confidence intervals are as follows: 50% confidence. 069 95% confidence 207 Required: (J) Assuming the corporation is projecting advertising expenditures amounting to $1,500,000 and factory rebate expenditures amounting to $12,000,000 for the next time period, calculate expected customer demand in units using: (@) Equation 1 (b) Equation 2 (2) Assume that equation 4 is employed to estimate a total customer demand of 104,160 automobiles fora time period. Using a 50% confidence level, determine the range of automobile sales that could occur during the time period. (3) Select the regression equation that would be most advantageous to predict customer sales demand and explain why itis the best. (4) Explain the significance of equation 3 and how it may be used in evaluating equation 4. (ICMA adapted)