Lg^ Y 4 ''\ Division of Agricultural Scienc hk&fS and VOLUME PRICING rKUltDUKtj in California D. A. Clarke, Jr. CALIFORNIA AGRICULTURAL EXPERIMENT STATION BULLETIN 757 The system of uniform pricing in milk distribution, used for many years in California, encourages hidden rebates the growth of captive creameries inefficiencies of distribution, such as the serving of a single customer by several distributors. By lading to effect real economies through appropriate price re- ductions, the system has not provided an incentive for con- sumers to choose more efficient distribution methods. Furthermore, the system of flat pricing is inadequate and inequit- able. Possible alternatives: 1 . A system of service charges coupled with reduced levels of flat prices could reflect actual costs and margins quite accu- rately. 2. A system of volume discounts, while less accurate, would be superior to the flat price plan. Results of the research reported here may be applied . . . 1 . To provide a basis upon which administered prices could more closely approximate perfect market prices. 2. To suggest the nature of the savings that can be expected with a more "realistic" system of pricing. GENERAL discussion of fluid milk pricing will be found on pages 3 to 18. SPECIFIC details of the development of the analysis appear on pages 18 to 66. THE AUTHOR: D. A. Clarke, Jr., is Associate Agricultural Economist in the Experiment Station and on the Giannini Foundation, Berkeley. DECEMBER, 1956 MILK DELIVERY COSTS and VOLUME PRICING PROCEDURES in CALIFORNIA D. A. CLARKE, JR. Introduction -LLUID MILK DISTRIBUTION consists of a series of operations performed in getting milk from the farm to the ultimate con- sumer. The "costs" of these marketing operations are often measured by use of average price spreads, such as those shown in table 1. The term "price spread" (or simply "spread"), as used in this bulletin, refers to the difference in prices at various levels in the market- ing system. Most of the available data on prices, however, quote the price paid to farmers f.o.b. city plant. Therefore, the costs of country assembly and trans- portation to the city are excluded from these particular measures of marketing charges, and the figures refer directly to the charges added for plant processing and bottling, delivery, and store margins where the milk is sold through such out- lets. As the table shows, many of the 137 cities have higher prices and wider spreads than do the California markets, while many other cities have lower prices and narrower spreads. Such comparisons are often used to measure the relative efficiency of differ- ent markets. It must be recognized, how- ever, that at best they are poor indices of efficiency in terms of either the output per unit of input or the extent to which prices reflect "necessary" costs as distinct from monopoly returns to producers, labor, or distributors. While it is true 1 Submitted for publication April 3, 1956. 2 Italicized numbers in parentheses refer to "Literature Cited," page 76. that higher marketing charges in one area than in another may mean that the second market is the more efficient, wide spreads may also result from differences in relative incomes of the two areas as reflected by wages for plant and delivery- men. Furthermore, variation in spreads may represent substantial differences in quality, in sanitary controls, and in added services and convenience to the customer. Differences of these types can be enumerated and, to an extent, meas- ured, but it is extremely difficult to devise a measure of their value (or "utility") to consumers. (For a discussion of the difficulties involved in the use of prices and margins as a measure of efficiency, see 1; 13). 2 If, for example, utility did not differ, then it could be said that the distribution of milk through stores is more "efficient" than the system of retail home deliveries, because the "store dif- ferential" (difference between consumer prices at stores and home-delivered) in 137 cities averaged 0.7 cent per quart, and in California markets, 0.6 cent. Yet some customers do choose to pay for home delivery. In fact, in a survey made in Connecticut in 1946, 44 per cent of the consumers interviewed in Willimantic and 63 per cent of those in Hartford in- dicated they would be unwilling to give up home-delivery services for less than a 5-cent store differential (2), thus indi- cating a substantial difference in the utility, to them, of the alternative prod- uct-service combinations. [3] Table 1. Price Spread in Milk Distribution, 137 United States Cities and 37 California Marketing Areas, January, 1956 33 c V M O IS O ot co '0 "o hi 0) CO hi 00 "«tf " M p. OT CD A (S bo > ft 1 02 CO * Averages calculated from original data and may not reconcile exactly due to rounding. Sources: United States data from: U. S. Agricultural Marketing Service, "Fluid Milk and Cream Report" (Washington: Govt. Print. Off., January, 1956). California data from: California Crop and Livestock Reporting Service, "Dairy Information Bulletin" (Sacramento, January, 1956), vol. XII, no. 10. (Data for October, 1955.) California raw product cost based on 3.6 per cent butterfat. "3 co 4> V P «)10(«50)MIOHHOON CO CO tH CO *. fi 0) cr3 to a 3 ^05^05^05^*05^05^ OlOOiOOiOOlOOlOO n "3 a co CO "o CO .2 '0 "o CO £1 a 3 1 =0 CO hi 00 ©OOCOCOCOCOCNT-ltHOOOOOOO© tH © tH "3 co •• •a ii j5ot NniOt-MN^^iO(00)»}^NOOOrl tH tH tH 3 4J CO *> ft 3 05^05^ , O5"^*05 , «J'05^'05^J < 05' , 5j , 05^ 1 O5^' l>QOo6oio!c)dHH(N(NMM^^ldlOei tHtHtHtHtHtHtHtHtHi-ltHi-ltH loowowoiqoiooiopwpinoioo N«Jo6d«ddHHN(NMM^T)iiOlOci co -3 a> s. CO 4) M OT CO .s '0 *o u CO a 1 «« ■3 '3 co u OOOrH05t>0000 tH tH CO "3 co P DOT t-iO00«O^(OiOtNt- q.3 O c Oi^OS^OS^OS^OS^OS^OS^OS^OS ioL>o6oddo)d6iHHcid co co iftOiOOOOOpiqpwpiqpiqoiq lodtdN^ooodoidddriHNNMn CO "3 co 0) s. CO "S CO .2 '3 *o u V x> a 1 co 31 00 OOOtHtNOO^fOOCOtHOOOOOOOO CO tH •3 co co a> •si DOT 0©0"odo6oioJddHH*(NNMM^tjiiou5d lOOiOowoooooubpiqpiqpiqp ^MooddddrlHNNMcott^ioifJti The major differences between costs of wholesale and retail milk distribution occur in the items of processing and de- livery. Processing costs are higher for wholesale, reflecting primarily the dif- ference in container expense. Practically all milk sold at wholesale in California is in fiber containers; retail sales are made in glass. The difference in expense amounts to about 1.1 cents per quart more for fiber containers than for re- turnable glass bottles. Lower delivery expense for wholesale, on the other hand, results from the larger volume of output per man on wholesale routes. This in turn is influenced by the lighter and more compact fiber containers, especially where large-volume customers are con- cerned. Relative importance of wholesale and retail milk sales. The proportions of total milk sales through wholesale channels — grocery stores, restaurants, and institutions — vary considerably by areas in the United States. Table 2 shows the amount of this variation and the rela- tive importance of sales through outlets of this type as of January, 1950 (the latest date for which nationwide figures are available). With the exception of New York City, sales at wholesale ap- pear to be a more important factor on the West Coast than on the East Coast. West Coast markets typically average 50 per cent or more sales through wholesale channels, whereas very few of the eastern markets report more than 35 per cent of total sales through these outlets. Structural characteristics of the industry. Fluid milk processing and dis- tributing are subject to economies of large-scale operation. In general, plants with large volumes have lower unit costs than those with small volumes even when both are operating at the same degree of capacity. On the other hand, because of particular advantages, such as special abilities that result in high efficiencies in plant and/or delivery operations, the availability of unpaid family labor, a desire to be "on their own," or for other reasons, there are a large number of small distributors. Thus, most milk markets contain a small number of large distributors and a large number of rela- tively small operators. Typically, the sales of the three largest distributors account for approximately half of the total sales within the local market. For example, in 1954, the three largest handlers in Sacramento made 66 Table 2. Per Cent of Total Milk Sales Classified as Wholesale* Wholesale Market as per cent of total sales Portland, Maine 30 Manchester-Nashua, New Hampshire 30 Portsmouth, New Hampshire 20 Burlington, Vermont 40 Boston, Massachusetts 35 Worcester, Massachusetts . . . 35 Lowell-Lawrence, Massa- chusetts 40 Springfield, Massachusetts. . 25 New Haven, Connecticut 25 Hartford, Connecticut 25 Providence, Rhode Island . . . 30 New York, New York 70 Plainfield, New Jersey 34 Newark, New Jersey 34 Durham, North Carolina 50 Winston-Salem, North Carolina 50 Lewiston, Montana 6 Portland, Oregon 50 Klamath Falls, Oregon 75 California : Average for state 52 Bay area 56 San Joaquin Valley 60 Southern California 49 * All data except those for California areas are based on estimates received from milk control offi- cials or industry representatives in reply to mail surveys, collected and compiled by R. G. Bressler, Jr., and refer to January, 1950; California data from California Crop and Livestock Reporting Service, "Dairy Information Bulletin" (Sacramento, Janu- ary, 1956), vol. XII, no. 10. (Datafor October, 1955.) [5] per cent of the fluid milk sales. In each of the following markets, the three firms with greatest volumes accounted for these percentages of total sales: Fresno, 59; San Francisco, 59; San Diego, 44; Alameda-Contra Costa, 47; and Los Angeles, 35. (That this degree of con- centration is neither unique to California nor a result of a tendency toward "big business" in recent years is borne out by similar percentages for the three leading distributors in other areas during the 1930's: Boston, 64; St. Louis, 69; and Phoenix, 84.) On the other hand, there are nearly 500 licensed distributors in California, nearly half of which do busi- ness in the above-named markets of the state. In spite of scale economies and the fact that some corporations operate on a regional or nationwide basis, fluid milk distribution is primarily local in nature. This results partly from the bulk and perishability of fluid milk — transporta- tion over long distances is expensive rela- tive to value — and partly from artificial trade barriers. Early in the development of commercial milk distribution, health officials became alarmed over the number of epidemics which could be directly traced to bacteria in milk. As a result, local ordinances and state regulations were passed in an effort to control the quality of the milk supply. Many of these, although reasonable at the time, are no longer important in safeguarding the health of local consumers. Instead, they have developed into trade barriers which protect the financial health of the local industry. Many communities have ordi- nances which require pasteurization within the city limits. A full coverage of the types of restrictions to the free flow of milk between areas is beyond the scope of this study, and has been discussed by others (see 7, 8, 11, and 14). The point to be made here is that these devices have served to "protect" the local nature of fluid milk markets. As previously stated, the typical or- ganization of the industry within a mar- ket involves a few dominant distributors with the major share of the volume and a relatively large number of small dis- tributors which handle the balance. While in the business sense the fluid milk in- dustry tends to be intensely competitive ( that is, firms expend great efforts to win customers from their rivals), the struc- ture and organization of the market dif- fer significantly from the economist's conception of "pure" competition. In this industry, therefore, "free enter- prise" does not have the same implica- tions as "free competition." The former tends toward expensive advertising out- lays designed primarily to attract busi- ness away from rivals, and the provision of costly services which, in themselves, are not necessarily desired by consumers at the price which must be paid for them. In most cases, consumers do not have the opportunity to make a rational choice among service combinations since the ad- ditional costs of new services may be temporarily absorbed by the firms which have initiated them. To the extent that a device is effective in attracting cus- tomers, all firms are forced to follow suit, and such services become a part of the "normal" operations. The additional costs must sooner or later be recouped from the consumer if the industry is to continue to operate in the long run. At times, free enterprise in an indus- try organization such as this takes the form of "price wars." While generally welcomed by consumers, who may obtain short-time price benefits, price wars are usually precipitated to force weaker firms out of business and thus provide greater monopoly power for the remaining firms. Viewed from this standpoint, price con- trols administered in the public interest could do much toward improving the ef- ficiency and lowering the costs of indus- try operation. [6] Pricing Procedures and Policies of the Bureau of Milk Control Beginning in 1935, with the passage of the Young Act, and supplemented in 1937 by the Desmond Act, prices for fluid milk in California have been estab- lished by an agency of the state govern- ment. The historical origins of govern- ment price control have been covered elsewhere and will not be repeated here (see 5 and 12) . At present, 37 separate marketing areas are established in California. Vir- tually all of the Class 1 sales (which in California include both fluid milk and fluid cream) are covered under the fluid milk pricing provisions of the control act. The provisions establish producer prices, prices for interdistributor sales (includ- ing the sales of processed and bottled products to subdistributors), and whole- sale and retail minimum resale prices. The latter involve a schedule by type (such as standard grades, premium grades, and cream in areas where a cream plan is in effect) and by container sizes. The provisions of the law, as they ap- pear in the Agricultural Code, specify the types of prices to be established, and generally outline the procedures to be followed in determining the level of those prices. In arriving at minimum resale prices, a factor to be considered is that, in addition to the raw product cost (the price paid producers — also established by the state), such prices "will tend to maintain in the business of distributing fluid milk and fluid cream, or both, such number of reasonably efficient retail stores and distributors of fluid milk or cream, or both, in such marketing area as the director finds is necessary to in- sure to consumers in such marketing area sufficient distribution facilities of the several types or methods commonly used by consumers" (Art. 10, Ch. 17, Div. 6) . At the time that the control program was enacted, it was recognized that the pressures within the industry which had previously broken out in the form of price wars would tend to take other forms. For this reason, the legislation providing for minimum resale prices also specifies a number of unfair trade prac- tices, the use of which is punishable under civil law. These unfair practices include price rebates or unearned dis- counts, and the giving away of any prod- ucts, services, or articles of any kind for the purpose of securing the fluid milk or fluid cream business of any customer. In this manner, both direct and indirect price reductions are prohibited. (For a list of unfair trade practices, see Ap- pendix C, p. 76.) The Bureau of Milk Control— the agency administering price control in California — makes periodic audits of dis- tributor records to determine distribu- tion costs. These, in turn, are used as guides in establishing resale prices. Cost studies of sample plants are carried out at relatively frequent intervals. In addi- tion to these regular studies, revisions are frequently made on the basis of ma- jor cost changes (such as a new labor contract or adjustments made in the price schedules for containers and other sup- plies ) . Most of these analyses are made by Bureau of Milk Control personnel through an audit of the financial records of individual distributors. Under the law, the distributing firms are required to turn over to the Bureau accountants any and all materials necessary to ascer- tain accurately the costs experienced and the volumes of sales. Furthermore, in making studies of costs, the Bureau per- sonnel frequently apply "efficiency fac- [7] tors" in order to arrive at the so-called ''reasonable costs" which are used for pricing decisions. For example, an effi- ciency factor might be a standard rate of performance or a "reasonable" size of delivery route volume. For plants not meeting this rate of performance, costs are calculated as they would have been if this rate were achieved. Other adjust- ments are made in these computations, such as the prorating of lump-sum ex- penses — for example, the purchase, at one time, of a large supply of bottle caps — over the period of time under study. Relation of Costs and Prices Under conditions of a perfect market (see 3 and 9) , the allocation of resources, the volumes of the various goods and services produced, and the product and factor prices are all determined simul- taneously. These solutions are achieved through the operation of free choice on the part of the consumer and a freely competitive productive system — reflect- ing the basic or ultimate resource and technological limitations. Under this sys- tem, prices would equal costs (in the long-run sense) , including normal profits and the inherent determination of ap- propriate rents and other factor prices. 3 This perfect market system would be economically efficient in that no changes in organization could be made which would increase the welfare of all individ- uals in the society. Economic efficiency, however, is only one of the factors con- tributing to the social welfare. Thus, Congress may adopt measures designed ' This would be true only in circumstances which approach those of perfect competition — that is, perfect knowledge on the part of both buyers and sellers, homogeneity of product, and an atomistic organization of the industry. Where scale economies which are incompatible with an atomistic structure prevail, a rule that entrepreneurs equate marginal cost to price (rather than to marginal revenue) is required to meet the condition of price equals cost. While this poses some perplexing problems anent the organization of the industry and its regulation, the point to be made here is that such an "ideal" system would result in a set of prices consistent with an economically efficient allocation of resources. to increase farm income at the cost of the income of the nonfarm segment (see 4). In the same way, there may be an instinctive reaction to protect the small businessman and "preserve the American way of life." Society may, in some cases, choose to provide a subsidy rather than let this form of economic life die out. Such subsidies may take the form of some type of "protection" from lower cost competitors. When such protection exists, it takes the form of "income trans- fers" (taking from one group and giving to others). Note that such income trans- fers are not necessarily incompatible with an efficient allocation of resources, but that often the "means" selected to effect such transfers are those which work through the pricing system, with conse- quent distortions in resource use. In this instance, as in any other, the sacrifices involved in such decisions — that is, the corresponding loss in eco- nomic efficiency — must be recognized if choices are to be made intelligently so that costs are weighed in relation to gains. Determining milk pricing systems that adequately reflect the nature of costs will ( 1 ) provide a basis upon which milk prices could reflect cost differentials and so approximate perfect market prices, and (2) indicate the costs of subsidiz- ing certain elements in milk distribution by suggesting the nature of the alterna- tive and economically efficient organiza- tions and the savings which would accrue with reorganization. [8] Impact of Flat Pricing Under Conditions of Differential Costs It has long been recognized that it costs less per unit to deliver a large order of milk than a small one. This applies both to wholesale and retail milk dis- tribution, and is convincingly brought out through a comparison of the results of the processing and distribution cost studies made by the Bureau of Milk Con- trol. A survey in the Los Angeles market showed that total distributor's expense for wholesale sales of milk in fiber con- tainers had a range of over 2% cents per quart ($0.026522) between the high- est- and lowest-cost firms. This survey covered the period August through No- vember, 1952, the latest date for which the firm serving the very large volume customers was included in the study. (More recent cost studies have been made of other firms in the Los Angeles area.) Further inspection indicates that the largest part of the expense difference ($0.013789) occurs in the single item of delivery cost. The high-cost firm spends nearly four times as much per quart in getting the milk from the plant to the customers as does the low-cost firm. Cus- 4 Some recent exceptions have been made to the uniform pricing policy. Late in 1952, a service-charge system was established for retail deliveries in the Alameda-Contra Costa mar- keting area. At the same time, a % cent per quart discount on home-delivered sales in ex- cess of 60 units per month was adopted in sev- eral southern California markets. A 5 per cent discount was also allowed on sales to schools and other public institutions where sales of $250 or more per month were made. In January, 1956, a new order was issued for the Fresno area, under which sales of milk in excess of a value of $5 per delivery were subject to an 8 per cent discount. At that time, in that market, a service-charge system on retail home-delivered sales similar to that in the Alameda-Contra Costa area was established. At the time of writing, it is contemplated that similar orders will be issued for other markets in the state in the relatively near future. tomers of the high-cost firm average about 60 units per delivery; those of the low-cost firm average nearly 2,400 units. While this example is extreme, the rec- ords of all firms studied indicate a defi- nite tendency for delivery costs per unit to decrease as the average volume per stop increases. The Bureau of Milk Control for many years has set minimum resale prices which have been uniform among cus- tomers. It made no provision for dis- counts or other allowances for the differential costs involved in serving different types or sizes of delivery/ As in almost every instance of arbitrary price controls, pressures have resulted which are directly attributable to this flat price policy. The impact of these pressures was no doubt postponed in the early years of milk control because of the war situation and the accompanying restrictions and shortages. Such pres- sures were present, however, and have become increasingly important since the end of the war. Although stiff penalties have accompanied the relatively few vio- lations that have been uncovered, it is highly probable that secret rebates, dis- counts, and discriminatory price advan- tages on nonpriced items have been granted to preferred customers. The incentive for efforts to by-pass the provisions of the flat price schedules can readily be seen. Assume that the price set by the Bureau of Milk Control takes into consideration the delivery cost of the average-sized customer in the market. This average-sized customer represents many customers which are smaller than average and many which are larger. Under conditions where it costs less per unit to deliver a large volume than a small volume, the flat price set by the Bureau is inadequate to cover the costs [9] involved in stops smaller than average, and is more than adequate for larger than average stops. (The term "stop" will be used throughout to denote a single deliv- ery to a customer.) It might be argued, from the stand- point of the distributors' welfare (omit- ting for the time the effect upon consumers), that this is an equitable procedure, provided each distributor has the same proportionate share of large and small deliveries. The margin estab- lished by the Bureau should be sufficient to cover the cost of a reasonably efficient distributor in serving stops of average size; thus, all distributors would be treated fairly. If all distributors did not have the same size distribution of deliv- eries, then any individual firm with a greater than proportionate share of small stops — even if "reasonably efficient" in all other operations and in the organiza- tion of all factors under its direct con- trol — would find it difficult to operate within the price spread based on costs for the average size of delivery in the market. Flat pricing under circumstances of differential costs raises a serious equity problem between different customers. When a procedure is used which "aver- ages out" the differential costs associated with varying volume categories, the small-volume customers are obviously being subsidized by the large-volume stops. The uniform price system in Califor- nia markets has had two major effects. In the first place, it has decreased the efficiency of the milk distribution system. Since volume-cost differences are not reflected in price, individual customers find no monetary incentive to place their total order with a single dairy firm. Rather, so long as supplies are plentiful, the customer — particularly the grocery store — is under some pressure to divide his trade among several distributors, thus taking advantage of consumer brand preferences. These split stops — also ob- served in home deliveries — of necessity reduce the average volume of delivery in the market and so increase the dis- tribution cost. The second effect of flat pricing has been the rather elaborate procedures by which wholesale customers become at- tached to particular milk distributors. Among the more extreme devices which may be substituted for a price incentive in order to keep a set of particularly desirable wholesale outlets are those fa- miliarly known as the "captive creamery" and the "captive supermarket." A captive creamery is a milk processing and dis- tributing concern whose capital stock is held by the owners of supermarkets to whom its products are sold. A captive supermarket involves the reverse owner- ship pattern — in this case, milk distribu- tors hold stock in supermarket corpora- tions. The idea, essentially the same in both instances, is to form an attachment between the milk company and the more attractive of the wholesale customers in such a way that the profits resulting from the economies involved in sales to these customers can be "captured." The development of the captive cream- ery stems from the realization, on the part of large-volume, low-cost outlets (mainly supermarkets), that substantial savings are involved in making large de- liveries. Savings of this same type exist in many other commodities, and explain, in large part, the success of the super- market movement. By buying in large quantities and by selling large volumes with a rapid turnover, the per unit cost of distribution is reduced. With milk in California, however, these large-volume outlets are forbidden, by the legally es- tablished uniform prices, to participate in these benefits. As customers, these out- lets cannot share in the profits to the distributor for which they were, in fact, responsible. As owners — stockholders in the milk processing corporation — they can participate in the profits. The opera- tion of captive creameries, their eco- [10] nomic justification, and their importance are well illustrated in the following ar- ticle quoted from the Milk News Weekly (October 15, 1954) : Supermarkets Offered Stock in Creamery San Francisco, Oct. 11 [1954] — An in- dependent wholesale milk distributor has offered to sell up to 49% of its corporate stock to a dozen supermarkets in this area for the purpose of doubling the creamery's sales and permitting the markets to partici- pate in the savings made possible by vol- ume distribution, it was reliably reported here today. Details of the plan require 75% of the market's dairy products purchases to be from the distributor. In return, the market would be allowed to purchase stock up to % of their monthly billings. It is estimated that the dividends would clear the cost of the stock in a few months, and thereafter, dividends would amount to between 1$ and 1%# per quart of milk. The distributor's capitalization would permit participation of about 18 supers taking 40 cases of fibre quarts per day, it is estimated. The action seeks to match the economies enjoyed by chain store captive creamery delivery trucks which set off high volume loads, which have led to the claim "we could sell milk to consumers for less if permitted by state law." The creamery's proposal also envisions the addition of ice cream and bread lines in order to meet chain store consumer prices which are consistently lower. Market participation in creamery mar- gins are [sic] already in operation in Los Angeles, with the possibility of another offer under consideration in Oakland, it was reported. It is small wonder that, under circum- stances of this sort and especially where a sufficient number of large supermarkets exist to make the operation feasible, the captive creamery has become a factor of increasing importance. Annual reports of the California Department of Agricul- ture show that in Los Angeles, for exam- ple, one milk company, a wholly owned subsidiary of a large grocery chain, be- gan operations in 1929. A similar organi- zation controlled by another grocery chain was first licensed to process and distribute market milk and market cream in 1932. These concerns supplied the stores within their chains exclusively and so were the original captive creameries — in the sense that the ownership of the milk distribution and store enterprises was identical. No other milk distributor could hope to acquire the patronage of the retail stores within these chains unless consumer preference for the particular brands of "outside" distributors should force the chains to handle other than the "sponsored" brand. Two similar developments have oc- curred more recently in the Los Angeles market. In 1941 and 1948, two other milk processing and distributing firms were established, in both of which a substan- tial part of the capital stock was owned by individuals who also operated super- markets, even though the markets them- selves were owned independently. Both firms specialized in the sale of milk and dairy products to large-volume stores for resale to consumers. The growing impor- tance of the sales of these four specialized distributors — the two chain-operated and the two independent, supermarket-oper- ated — is shown in table 3. In a move to counteract the tendency for large-volume, wholesale customers to become affiliated with captive creamery operations, several of the larger inde- pendent milk concerns are reported to have made substantial investments in supermarkets in an effort to gain the power to influence the purchase policies of these establishments. 5 In addition to the captive creamery, therefore, there has developed the captive supermarket. The importance of its effect on the milk in- dustry is much more difficult to deter- mine than is that of the grocer-owned milk concerns, because independent milk 5 The following news item appeared under a Los Angeles date line of August 15, 1950, in a dairy trade journal: Joint announcement was made at a press luncheon today by Blair Holdings Corp. and Golden State Co., Ltd. (a milk distrib- utor) that the recently-formed subsidiary, Western Industries, Inc., has entered into [ii] Table 3. Estimated Total Sales of Market Milk Through Stores in Los Angeles, and Sales of Four Specialized Milk Companies, 1944-1955 Year Estimated sales through stores* Sales of four "captive" companies f Four companies' percentage of total sales 1944 thousand gallons of market milk 53,992 58,316 59,837 59,010 59,265 57,929 57,600 61,732 66,845 74,718 73,196 77,127 7,617 8,215 9,478 9,231 10,971 10,974 11,830 13,160 14,769 15,988 16,531 17,382 14.1 14.1 15.8 15.6 18.5 18.9 20.5 21.3 22.1 21.4 22.6 22.5 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 * Estimates for 1944-1948 and for 1951-1954 are from "Sales of Fluid Milk by Type of Trade and by Size and Type of Container," annual reports of the California Crop and Livestock Reporting Service covering sales for the month of October each year. These estimates were derived from the wholesale sales of milk in quart and half-gallon fiber containers. As no records are available for sales by container size for the period 1948-1951, estimates of store sales in 1949 and 1950 were made on the basis of what percentage of total wholesale sales during 1948 was store sales. panies before 1948. distributors are not required by law to keep records of sales to individual cus- tomers. In addition, the distributors so involved would be loath to release infor- mation of this type for fear of retaliation from rival distributors. The additional sales acquired by cus- tomer-owned companies have been made up almost entirely of customers handling large volumes. This seems reasonable in view of the fact that such outlets can be served at low cost — the main incentive for this type of organization. Further- an agreement to acquire the Jim Dandy super market chain in Los Angeles. The five super markets in the chain with an annual gross sales in 1949 totaling $7 million, will become the first asset acquired by the Blair-Golden State subsidiary. (From The California Milk News (Los Ange- les: August 18, 1950), vol. XXXVI, no. 21. The name of this publication has since been changed to the Milk News Weekly.) more, in reflecting cost savings back to the owner-outlets, such distribution firms have been able to make additional sav- ings by reducing delivery services. For example, deliveries may be made only to the store loading dock; the milk prod- ucts are then carried into the store by store employees whose wage rate is lower than that of the route drivers. As a mat- ter of fact, these operations often may be carried on without increasing store per- sonnel. Both outgrowths of the flat pricing procedure — the tendency for a greater proportion of split stops and the grow- ing importance of the captive creamery — have reduced the average size of delivery by the independent distributors. It is interesting to note that the effect of the captive creamery has actually been to improve the efficiency of distribution so [12] far as these captive firms and their affili- ated outlets are concerned. As distribu- tion firms of this type have expanded into the large-volume stops, the average delivery per customer has fallen off for the remaining firms. Furthermore, these low-cost outlets are kept from taking full advantage of their cost position by the establishment of out-of -store prices which are also maintained on a flat price basis. Alternative Price Plans Flat pricing, as discussed above, does not reflect differential costs adequately. Furthermore, it is inequitable, discrimi- natory, and fails to provide an incentive for more efficient and economical dis- tribution. It also encourages certain prac- tices, including secret rebates and the growth of captives. Two systems are con- sidered as alternatives to this type of pricing procedure. Service charges. This system in- volves two-part payment for delivered milk — first, a fixed charge per delivery, and second, a uniform price per quart delivered. Naturally, since a part of the payment is in the form of this fixed, or service, charge, the uniform price under this system is lower than that under the flat price procedure. The volume pricing effect, of course, stems from the alloca- tion — or "spreading" — of this fixed charge per delivery over the number of units delivered. Under current California cost condi- tions, a service charge system involving a 35-cent charge per delivery plus a uni- form price of 17.1 cents per quart for volumes less than 221 labor units, coupled with a charge of SI. 43 per de- livery and a uniform price of 16.6 cents per quart for deliveries of more than 221 units, would closely reflect costs for wholesale sales. This schedule would be consistent with returns under the present flat price system of 17.5 cents per quart. (A more complete discussion of how these schedules were derived, and of the relation of these prices to costs is pre- sented beginning on page 63.) In a similar manner, a retail service price schedule of 10.5 cents per delivery, with a uniform price of 18 cents per quart, would closely approximate costs of home-delivered milk. This schedule would provide returns comparable with a flat price system of 21.5 cents per quart. (The development leading to this pro- posed retail schedule is discussed begin- ning on page 63.) Volume discounts. While necessarily less accurate than a service charge sys- tem, volume discounts can be developed which are greatly superior to flat pricing in reflecting differential costs. An almost infinite number of discount schedules can be devised. As an example, one schedule for wholesale sales has been arbitrarily chosen to illustrate the mechanics of vol- ume discount arrangements, and is later compared with the cost-volume relation- ship. This particular schedule involves four classes and a "base" price of 19 cents. At each class — or "bracket" — a progressively increasing percentage dis- count (ranging from 8 to 12) applies to the total milk purchases. The pricing agency has been adopting a somewhat different volume discount procedure in those California markets where volume pricing is being initiated. Here, a single discount figure applies to all sales in excess of a stated volume. For example, the schedule currently in effect in Fresno involves a base price of 17.5 cents per quart with a discount of 8 per cent on all milk sales in excess of $5 value per delivery. (The effects of both types of volume discount schedules are considered more fully beginning on page 64.) [13] Differential Pricing Systems in Other Areas It is difficult to discover the extent and nature of similar pricing plans in mar- kets outside California. Most major mar- kets do not have resale price control, and prices — especially to large and attractive wholesale customers — result from nego- tiations between buyer and seller. That discounts are given to large-volume, wholesale customers in uncontrolled markets is evidenced by the fact that such discounts were common in the Los An- geles market before resale price control. Spencer {10), in a study of this market in the summer of 1930, made the follow- ing statement: "A considerable propor- tion of the wholesale milk, both bottled and bulk, is sold for less than the quoted prices. Wholesale customers who take a large volume of milk and cream most often receive a discount or rebate. . . . This does not apply to groups of chain stores, restaurants, or confectioners, the individual units of which may not take any larger volume than independent stores or restaurants which receive no discount." He goes on to say that it is impossible to eliminate such discounts. "Buyers that take a large volume of goods are such attractive customers that even if all established distributors agree to eliminate discounts, others would come in to take advantage of the situation." Some available information on the ex- tent of open and established discount pricing in retail deliveries appears in monthly market information reports is- sued by the United States Department of Agriculture. According to the January, 1956, Fluid Milk and Cream Report, dis- counts were given to large-volume retail customers in 39 out of 137 markets in the United States reported for that month. The delivery volumes at which these discounts become effective differ between markets, and the discounts amount to either 1 or 2 cents per quart. Although this report provides no infor- mation which would indicate the exist- ence of discounts on wholesale deliveries, it is well known that such discounts do exist in many markets. One of the differential retail pricing systems in current use is the so-called "Elwell Plan" (6). Named for the man who headed the distributing firm which initiated it, the plan has been in operation in the Minneapolis market since 1938. As it operates at present (January, 1956), its essential feature is a service charge of 4 cents per delivery, reflected in a price of 23 cents for the delivery of one quart plus a price of 19 cents per quart for all additional quarts in the same delivery. (If costs in Minneapolis are consistent with those determined in California, however, the 4-cent service charge is too low to reflect cost differ- ences accurately.) According to popular reports, this plan has been well received by both consumers and distributors. Potential Savings Through Volume Pricing Pricing plans such as those discussed above could not be introduced without impact on the industry. Lower net prices would encourage wholesale customers to consolidate their orders and to limit the number of dealers from whom they pur- chase milk. Retail customers, on the other hand, would be discouraged from de- f 14] veloping "split stops" — that is, obtaining every-day delivery by getting milk from each of two distributors on alternate days. Furthermore, customers receiving home delivery while supplementing their milk purchases at groceries would be en- couraged to take all of their milk from one source or the other. Such consolida- tions would mean a real increase in effi- ciency for the market as a whole, which would be reflected in lower average costs of distribution and in lower average gross incomes for distributors. Just how these price changes would affect individual distributors would de- pend on what changes occurred in the number and average size of their cus- tomers. Immediate route reorganization would be required, permitting route con- solidations to take advantage of the in- creased load sizes possible with larger customers. Although the exact nature of the changes cannot be predicted without actual experience, the potential savings resulting from consolidation of the entire milk market can be estimated. Savings in wholesale distribution. The duplication involved when a whole- sale customer is served by more than one distributor can be reduced either by ad- ministrative order (as during World War II) or by a pricing system in which the economies of large-scale deliveries en- courage purchases from a single dis- tributor. Wholesale customers of the independent distributors included in the Los Angeles sample study of 1950 (see p. 19) received dairy products from an average of 1.71 distributors. The aver- age volume per delivery per company amounted to 77 labor units. Where each customer is supplied by a single distribu- tor, the average size of delivery would thus be increased to 132 units. On the basis of the developed cost relationships, this would mean that route volumes could be increased 26 per cent, resulting in a 19 per cent saving in unit delivery costs. This 26 per cent increase would permit a 20 per cent decrease in number of routes, holding total market volume con- stant at the current level. Since duplica- tion exists primarily in grocery stores rather than in the other types of outlets, the savings in terms of stops of this type would be greater than the over-all aver- ages presented here. The savings to be expected in other markets depend, of course, on the exist- ing degree of duplication and the present average size of deliveries. Within the city limits of Fresno, the average number of distributors per grocery is 3.14, and the average volume per stop amounts to ap- proximately 45 labor units. The average volume per grocery for this market, therefore, is over 140 labor units. A com- plete reorganization of the market, with each customer receiving from only one distributor, would permit an increase in load sizes of more than 65 per cent. This increase in route volumes could allow a reduction of 40 per cent in the number of wholesale routes, and about the same in the unit costs of delivery operations. Savings in retail distribution. No current data are available that would in- dicate the probable effects of a volume pricing system on the average volume per stop and, in turn, on the costs of retail delivery. Assuming, however, that the average volume per customer was in- creased by 1 labor unit per serve ( an increase from about 3 units to 4 units), certain conclusions can be drawn. (The assumption is based on the existence of an incentive for very small customers to switch to store purchases, large- volume customers otherwise purchasing from stores to switch to home delivery, and for "split stops" to be consolidated.) Such an increase in average volume per serve would allow an increase of 25 per cent in route volumes at a corresponding reduction of about 20 per cent in the number of routes operated and in the unit costs of retail delivery. In this case, under current cost levels, the saving in delivery cost would amount to nearly 1 cent per quart. [15] Potential Savings Through Reduced Services In addition to savings which can be made through consolidation of deliveries, economies in the distribution of milk could also be brought about by reducing the number of services provided. During World War II, special deliveries and special call backs for collection were eliminated, and retail delivery was put on an alternate-day basis. Furthermore, on wholesale milk routes, the order had to be specified in advance. Some of these wartime practices, such as every-other- day home delivery, were continued after the emergency, but other services were restored when supplies of milk and man- power became more plentiful. Table 11 (p. 35) shows the direct de- livery, time-volume relationships for al- ternative types of services and points out the reduction in time per customer, for any given volume, that would accompany the elimination of certain operations. Thus, delivery type "A" represents the full set of operations typically performed by wholesale route drivers at present, type "B" stipulates the elimination of the "secure order" and "collect" operations, and type "C" specifies a set of minimum operations, including a set-off delivery at the sidewalk or platform. Many other types of reduced service combinations can be visualized and could be handled equally well. The effect of the reduced direct deliv- ery time on permissible route organiza- tion has been calculated for these two additional types of deliveries. The result- ing route volumes compatible with an eight-hour route day under the alterna- tive delivery types are shown and com- pared with the full set of delivery services in table 19 (p. 66) . In addition, the table also shows the corresponding influence on unit costs under current cost rates. As can be seen, substantial gross mone- tary savings in truck and delivery labor costs could be accomplished through a reduction in the amounts of services pro- vided. For the average sized stop, a 37 per cent reduction in costs would result from a situation in which the order was known in advance by the route man, and collection was made by mail. This saving is increased to 52 per cent if, in addition, the products are set off at the sidewalk. The net savings would, of course, depend upon the additional costs involved when the customer phones in his order, on the extra bookkeeping and clerical work re- quired for handling the mail collections, and on the increase in store personnel required to bring the dairy products into the store from the sidewalk. Possibly more important than the monetary savings, at least in times of critical manpower shortage, are the labor savings that may be achieved under these alternative systems. In addition to the possible savings in manpower through consolidation of deliveries, the produc- tivity per route driver could be increased by 69 per cent and by 138 per cent, re- spectively, under the limited service types of delivery combinations described (p. 35). The magnitude of the cost differences, plus the fact that cost functions applica- ble to these reduced services can be de- termined, suggests that a pricing system could be devised which would reflect these differential services. Under a serv- ice-charge system of pricing, the costs (delivery costs plus other, nondelivery costs) of type "B" services could be ade- quately reflected for delivery volumes under 200 units by a charge of 18 cents plus 16.75 cents per quart. For type "C" services, for the same range of volumes, [16] a charge of 12 cents plus 16.65 cents would be sufficient. These compare with similar charges for type "A" services of 35 cents per delivery plus 17.1 cents for each quart delivered. It should again be noted that these figures do not allow for added costs — such as bookkeeping — in- curred by the distributor nor do they in- clude estimates of the added cost of services provided by store personnel. Differential Pricing at the Consumer Level All of the preceding discussion of pric- ing wholesale sales has been in terms of prices paid by customers — or the "in- store" price. The question then arises as to how such savings will be passed on to the ultimate consumers under circum- stances where, as in California, resale prices are established by public author- ity. The direct answer to this obviously depends on the attitude and action taken by the control authority. At least three policies are possible. First, the Bureau of Milk Control, having established a dif- ferential pricing system reflecting deliv- ery costs at the in-store level, could decide to do nothing about the consumer price and could continue to set uniform out-of -store prices on the basis of average costs for the market plus a flat store mar- gin. (The latter is the difference between the consumer price and the price paid by the wholesale customer.) Second, an in- tensive analysis could be made of the unit costs of store operation, and the con- sumer price could be based on a variable in-store price and variable store margin for different types and sizes of stores. Third, since the Bureau is directed to establish only minimum prices, a "stop- loss" price could be established, based on the most efficient combination of de- livery and store operations. Under any procedure, consumer prices in higher cost outlets would be free (so far as the Bureau is concerned) to depart from the minimums in an upward direction, and thus reflect the added costs of the extra services and conveniences. Carried to its logical extreme, the latter procedure would eliminate the "fixing" of store margins, but would possibly keep the wholesale customer from selling to con- sumers at prices less than the in-store de- livered cost. Uniform store prices. Under this policy, consumers would gain somewhat over a period of time. To the extent that consolidations take place and the dupli- cation of delivery services is eliminated, the efficiencies so gained would be re- flected in average delivery costs. In turn, these savings could eventually be re- flected in lower established prices. Fur- thermore, society as a whole would gain through resource use conservation since milk delivery would require fewer men and less truck resources when the route reorganization possibilities were fully exploited. This alternative, however, has obvious disadvantages. First, it would provide inequitable margins among stores. Large-volume outlets would re- ceive high margins and small-volume stores would receive small (and conceiv- ably even negative) margins. (This pre- sumes that small stores do not increase out-of-store prices above the minimums established, which, of course, they are legally free to do.) Note that, from the standpoint of the storekeeper, this is an exaggeration of one of the limitations of the flat pricing system. Second, the strongest argument against this proce- dure, from a social standpoint, is that it does not provide the ultimate consumer with an incentive to select the most ef- ficient system of satisfying his desires — that is, the combination of services which [17] yields the greatest amount of satisfaction relative to cost. Differential store prices. This pro- cedure is theoretically possible. Further- more, it is entirely logical that margins charged by large-volume stores reflect the lower unit costs of doing business through outlets of this size. However, it would be almost impossible to make such a study which would adequately repre- sent the "true" level and nature of such costs. Not only would it involve a highly complex problem of multiple products and multiservices, requiring allocation of joint costs among hundreds of items; it would also require consideration of such factors as the drawing power of certain items. For example, how much drawing power do fresh milk and bread have in attracting customers to a store where, once inside, they may buy additional items? In spite of these limitations, it is quite possible that information on the direct costs of handling milk in these out- lets could be obtained. This information would prove helpful in determining the variations in direct costs associated with different types and sizes of retail stores. Studies of this type have been carried out by the Bureau of Milk Control in the past and have been used in determining the prices established at the out-of-store level. Furthermore, this type of pricing system would resemble the procedure used by the OPA during World War II in setting retail price ceilings by class of stores — which generally reflected differences in sales size. Uniform store prices based on ef- ficient levels of operation. This pol- icy of establishing a price "floor" based on the most efficient combination of de- livery and store operations appears to be a practical procedure. Of course, it represents a near approach to complete decontrol of prices at this level. Through this system, the "dangers" of price wars resulting from aggressive store merchan- dising policy are minimized since the burden of the price wars could not be shifted back either to producers or dis- tributors as it is assumed that prices at these levels would continue to be estab- lished and enforced by state agency. Furthermore, the additional costs of the added convenience and services of the small-volume grocery would be allowed to find their own level, and consumers — with the exercise of free choice — could then be effective in determining the sys- tem that provides them the greatest net gains. Scope and Procedure of the Analysis The marketing of milk consists of a series of services. Although the physical composition of the product is not basi- cally altered in most cases, the great bulk of fluid milk must be collected or assem- bled at country points, transported to urban consuming centers, processed to some extent (pasteurized, standardized, and bottled), and delivered from the processing plants to stores and homes. Since several services are involved in milk marketing, and especially since all of these services need not be performed by the same corporate entity, it is con- venient to deal with the various aspects separately. Following this procedure, country assembly and transportation may be studied. When this is complete, anal- ysis may be made of the operation of processing plants. From this then follows an investigation of delivery costs. When all of these aspects have been studied, they can then be brought together into an analysis of the whole system — since each of these operations is essentially in- dependent, and the system as a whole [18] consists of a combination of the separate parts. 6 Operating on this premise, this study is undertaken as a detailed and intensive analysis of milk delivery operations. Plant operations and the services per- formed between the wholesale customer and the ultimate consumer are not in- cluded. Intimate details of the functional relationships of a single phase of milk marketing are provided in such a way that they will be useful not only in deter- mining means of improving efficiency in this phase, but also in providing meas- ures which can be included in an analysis of the system as a whole when more com- plete information on other phases be- comes available. The present analysis has three objec- tives: (1) to determine the costs of de- livering milk and milk products and to determine and measure the influence of the major cost-affecting variables; (2) to contrast these results with various pricing plans and so determine the type and nature of pricing systems which will adequately reflect costs; and (3) to de- termine the nature and effect of savings that might be realized through the reor- ganization of milk deliveries. Further- more, the specific application of all phases of the analysis will be confined to the major California markets although tests will be applied to determine the gen- eral applicability of the results to other areas. Milk customers — in addition to home- delivery ones — include grocery stores, 6 Such interdependencies as exist in the ag- gregate stem from the fact that the long-run average costs for the whole system reflect the compensating influence of the simultaneous existence of economies of scale in one of the parts (for example, processing plant opera- tion), and diseconomies within another (for example, through increased transportation re- quirements) . The separate aspects are inde- pendent, however, in the sense that the rela- tionships that determine physical inputs and costs are unaffected by the realized volume in any of the other components. restaurants, drugstore fountains and soda bars, schools, institutions, and com- missaries. Conditions of delivery vary widely within these classifications. Gro- ceries include large, modern supermar- kets handling great volumes of all commodities, including milk, as well as the neighborhood store where the quan- tity of milk purchased from one distribu- tor each day may more closely resemble the volume of a single retail delivery. While not of the same magnitude, similar discepancies exist in the volumes received per delivery within the other categories. One of the major aspects of this study is a consideration of the influences of variations in the volume per stop on de- livery costs. Furthermore, differences exist in the number of services provided for the cus- tomer, particularly by the wholesale route driver. Since these, in turn, influ- ence the costs involved in making the delivery, the effect of differential types of wholesale delivery services is also con- sidered. As has been frequently observed, one of the "wastes" involved in our present system results from the duplication of marketing services. The majority of wholesale customers receive dairy prod- ucts from more than one distributor. Studies made by the California Bureau of Milk Control during 1950 show that the average number of distributors serv- ing groceries in Los Angeles was 2.10; in Stockton, 2.47; and in San Francisco, 3.53. A 1954 study in Fresno showed that groceries were being served by an aver- age of 3.14 distributors. The effect of such duplication will be measured by making estimates of the possible savings (in terms of money, manpower, and equipment) through a consolidation in which each customer is served by a single distributor. Although brand names and advertising are used to increase the de- mand for specific brands, fluid milk and dairy products tend to be closely stand- ardized and undifferentiated in physical [19] composition. Certain sanitary regulations limit the bacteriological content of the products, and legal standards with re- spect to minimum milk fat and nonfat solids content for the most part also de- termine the maximum amounts which will be sold per unit of product. Actually, very little difference can now be found among the products sold by individual distributing firms. In spite of physical homogeneity of the products, however, it is possible that consumers' attitudes may be such that they would be willing to pay for the opportunity to make a choice among various brands or pack- ages. Data on the cost of duplication, therefore, will provide a basis for intel- ligent decisions relative to the "value" of being able to choose among products of different firms. Theoretical Framework of Analysis Determination of costs. The two basic costs involved in the delivery of milk are for truck operation and the pay- ment for labor. The magnitude of these costs depends on: (1) the levels of in- puts of truck resources (such as gallons of gasoline and quarts of oil) and of labor resources; and (2) the cost rates appropriate for these various inputs dur- ing the time period under consideration. The physical inputs of these resources are not fixed on a route-day or a unit basis, but vary according to the condi- tions of route organization. Thus, it does not cost so much to operate a small deliv- ery truck as it does a large trailer van. Fewer inputs are required to operate a route 10 miles long than to operate an identical type of route 100 miles long. Furthermore, the various labor require- ments in the over-all delivery operations are not independent. When the route driver works for a specified time period, the amount of time he spends in driving between customers and to and from the plant will affect the amount of time re- maining for delivery and, in turn, the number of customers that can be served during the day. It is therefore necessary to know the nature of these relationships. The theoretical model in a study of costs of this type is drawn from pro- duction economics. The production of delivery services requires the use of productive factors — labor and the vari- ous resources involved in truck opera- tion. When rationally applied, varying amounts of inputs of these labor and truck resources will result in different levels of outputs of delivery services. These outputs of delivery services are in themselves multiple products, however, which combine the amounts of the prod- ucts delivered, distance, and the associ- ated direct services. Furthermore, these direct services are multidimensional in that they involve alternative combina- tions of operations. The question of the cost of delivering a quart of milk, then, must be answered in terms of these at- tributes of the service output. Thus, in- stead of treating output as a function of inputs, inputs (such as labor) are ex- pressed as a function of the output char- acteristics or dimensions. In this way daily route labor (time) will be a com- plex function of route distance, route volume, the number of customers served, volume per customer, and the type of service. This can be represented by the general function: r = ad, v,c, v., ■ ■ -,x) where T represents the route labor time; D, the route distance; V, the total route volume; C, the number of customers served; V s , the volume per delivery; and X, the type of service — or the number of operations performed for the customer. Some of these "independent" variables are closely interrelated. For example, C multiplied by V s equals V. [20] Conventional problems of production economics deal with intensification upon a fixed factor and the technical condi- tions which give rise to diminishing re- turns. Thus, a machine may be operated at various rates of speed in order to ar- rive at an "optimum" rate. The present problem is not primarily concerned with changes in rates (that is, to make the man work faster) but with determining typical, or normal, performance rates. The volume of products delivered (just one of many dimensions of delivery serv- ice output) will be increased by changing the conditions of delivery or by working longer hours. This is reflected in the generalized time function shown above. Note that this is not a problem of di- minishing returns in the usual sense since there is no intensification in the rates of input and output flows. Since X — the type of service — is not a continuous variable, it will be appropri- ate to determine a separate time function for each type of service that will be con- sidered. Each of these will then specify the functional relationship between time inputs and the route characteristics when the type of service is held constant and will take the form : T = f(D, V, C, Vs) The influence of the independent vari- ables expressed in this equation is fur- ther complicated by the fact that, while D enters into time requirements only for driving time, C — the number of custo- mers — may influence both the time in- volved in driving and also the time spent directly servicing customers. Thus, total time may be broken down, for analytical purposes, into relatively homogeneous elements. This might take the form: T t = T 1 + T, + T, + T, where T t refers to total route time, T 1 to time spent driving, T 2 to time spent directly servicing customers, T s to time spent at the plant, and T 4 to miscellane- ous and personal time. Then each of these elements can be analyzed separately rela- tive to the appropriate variables with the results: T l = a + bC + cD T 2 = d + eV s T z = /+ gC + hV T A = i Summing these coefficients arithmeti- cally, they form the total time expression : T = (a + d + / + i) + {bC + gC)+cD + hV + eV, a more specific form of the function. In these expressions, a limit to output is provided by the number of hours that can be worked during a given period. Since virtually all milk routes in Califor- nia operate on an eight-hour workday, eight hours can be used as a practical limit to daily capacity and therefore a maximum value — or cutoff point — for the time function. Expressing time in minutes, and substituting this maximum in the total function, result in the fol- lowing : 480 = f(D, V, C, V s ) In this form the function can be used to determine the effect of the interrela- tionships among the "independent' 1 variables. Remembering that V — CV S , analyses can be made to determine the interrelationships with D. Knowing the extent of all of these interrelationships, it will then be possible to show how changes in any one of the components (such as V 8 ) will permit changes in the others (such as V) . With total route time fixed at 480 minutes, time per unit of volume (or per customer served) can be calculated by dividing 480 by V (or C) . [21] Labor costs per unit of volume will then be the number of minutes required per unit, multiplied by the appropriate labor cost per minute or, as an equivalent, the labor cost per day divided by the total units delivered (V). Truck costs can be determined in a similar manner. Although labor costs per eight-hour day in most California markets are constant — that is, they are not influenced by route volume, distance, or number of customers — some positive relation between truck costs and volume will be expected because larger-sized loads require larger trucks. The influence of volume and/or number of customers and volume per stop can be considered in arriving at the total truck cost per day. Truck costs per unit are, then, the daily truck costs (which may be a function of volume and other of the attributes) di- vided by the daily volume. General Procedures The general procedure and form of the analysis have been specified in the previ- ous section. Certain types of information are required in the quantification of the input functions. For example, it will be necessary to know not only that distance enters into the determination of route time requirements, but also how it enters. Data will be required on route character- istics, such as the mileages between vari- ous points, the number of interruptions and delays involved, and the resulting time requirements. Furthermore, details will be needed upon the number of dif- ferent operations performed, the volumes delivered, and other variables which in- fluence the amount of time spent servic- ing customers. In many cases it will be impossible to establish an "ideal" formulation of a specific relationship. Some variables, although logically related to the opera- tion under consideration, are extremely difficult to measure in quantifiable terms. An example of this might be the particu- lar skill and ambition of an individual. Limitations of time and available funds are also important — both from the stand- point of obtaining, at one time, all the detailed information desired and also of going back to obtain further informa- tion on details which may have been originally overlooked. Such situations require either that further simplification be made in the model, thus justifying the use of alternative or related variables, or that the differences in these elusive vari- ables be ignored, and the results of the analysis interpreted to specify average or typical levels of such variables in the sample study. When the data have been analyzed and the significance of the results tested, de- tails of the relationships between input requirements and the corresponding out- puts of the various aspects of delivery services — such as the influence of dis- tance on driving time and of volume per delivery on direct service time — are known. These can then be combined — or synthesized — into an expression of the input requirements of the route as a whole. This provides the "hard core" of the analysis and, in fact, the end prod- ucts of the empirical analyses. These re- sults can then be applied to numerous practical problems which may arise. In this report, the application of these re- sults will be limited to two interrelated problems, both of which have immediate and pressing importance. One of these concerns the economic pressures which arise from the administration of an arbi- trary flat pricing system in California markets. The other concerns the poten- tial savings in vital resources through changes in the organization of milk de- livery systems. [22] Analysis of Time Study Data Description of Data Data were obtained on milk delivery route operations during the period 1950- 1954. Wholesale routes were studied in all of the major markets in the state, in- cluding Los Angeles, San Francisco, Ala- meda-Contra Costa, Sacramento, San Diego, Fresno, and Stockton. Under the supervision of the Giannini Foundation, Bureau of Milk Control accountants — trained and experienced in making and recording time studies — rode and ob- served a total of 272 wholesale routes operated by 48 processing plants in these seven markets. In addition, de- tailed time studies were made on 113 retail routes operating in the Los An- geles and Fresno markets. A total of more than 8,000 individual customer de- liveries was recorded for wholesale op- 7 The "sample" of routes covered, and upon which time observations were recorded, is not representative in the sense that the number of observations in each of the various size groups and types of outlets served which appears in the sample is proportional to the numbers in the universe as a whole. This is particularly true of the wholesale routes studied. The ma- jority of wholesale milk deliveries are of rela- tively small volumes, which means that the frequency distribution of delivery sizes is not only skewed to the left but also "tails off" to the right, with only a small percentage of the total number of observations having large- volume deliveries. With a proportional sample, therefore, the fewness of cases in this high- volume range will result in substantially less reliability for this range for any given size of sample than will a sample which is dispropor- tionately weighted by large-volume observa- tions. For this reason, an effort was made to obtain as many large-volume delivery observa- tions as possible. Each company studied was asked to supply a given number of routes to be observed and was further requested to provide those routes with the largest number of large- volume customers. The number of routes re- quested from each company roughly reflected differences in the total volume for each market handled per firm. erations, while more than 18,000 custo- mer deliveries were observed for retail home-delivered sales. 7 For each individual delivery, all of the operations performed by the route driver were noted and the amount of elapsed time was recorded. In addition, the minute and hour were noted when the route man reported for work, left the plant, arrived at his first delivery point, completed the last delivery, returned to the plant, and made his returns for the day to the plant cashier. The time the day's work was completed was also re- corded. The mileage registered on the truck speedometer was obtained at the beginning of the day, at the first de- livery, at the last delivery, and again when the truck returned to the plant. Wholesale delivery observations. At a typical wholesale customer service, the driver performs a series of separate operations. Quite frequently, he first walks into the customer's establishment and determines the amount of products to be delivered. He may simply look over the milk and dairy products supplies re- maining from previous deliveries or he may obtain the order direct from the manager. In some instances, of course, the driver may do both. Knowing the amount to be delivered, the driver re- turns to the truck and assembles the order. The next operation involves the actual delivery — usually to a point con- veniently located by the storage box or refrigerated showcase. The method of delivery is determined by the quantities of products to be delivered. For small volumes, where two cases or less are re- quired, the driver ordinarily carries the cases by hand. For larger deliveries, most wholesale milk delivery trucks carry a small, four-wheel dolly on which five or six cases can be piled and pushed directly into the customer's premises. Where very [23] large volumes are delivered at a single outlet, a larger dolly, hand truck, or me- chanical fork-lift truck may be made available by the store management. In some instances, a system of conveyors is utilized. After the products have been de- livered, the driver makes the financial settlement. This may consist of either col- lecting cash or obtaining the signature of the store manager, or some other au- thorized person, on the sales slip. The driver then assembles empty containers, returns and loads them on the truck, and is ready to proceed to his next stop. Each of the above-mentioned opera- tions was timed separately. In addition, any time spent in waiting or talking while serving the customer was observed. The distance involved in making deliveries was recorded for each customer, and a notation was made of special services — such as arranging or putting the products away in the display space. Quantities of products delivered, by type and size of container, were recorded for each set of observations. Notation was also made of the number of distributors who normally serviced each individual outlet. (See Ap- pendix A, p. 68, for instructions and data forms given to the timers.) Retail delivery observations. No effort was made to break down the indi- vidual retail delivery into its component parts. Here the gross time spent in di- rectly serving each customer was noted — including all time, from the moment the truck was securely parked at the delivery location until the driver was ready to proceed to the next delivery point. In ad- dition, however, any operation which was not normally a part of a routine de- livery — such as time spent waiting or talking, time spent making collections, and any other nonrecurrent interrup- tion — was timed in order that the fre- quency of these interruptions and their effect upon time requirements could be accounted for in the analysis. The num- ber of customers served per delivery trip, the volume delivered per customer, and the walking distance required per de- livery trip were also recorded. Analysis of Time Requirements During the course of a day's work, the route driver performs several quite dif- ferent types of duties. Time requirements for these various tasks are related to dif- ferent types of variables. For this reason, the total route time in this analysis was broken down into driving time, time spent in directly serving customers, and miscellaneous time. Wholesale Route Driving Time The amount of time spent in driving the delivery truck depends on the distance to be traveled and the rate of speed which can be maintained. The rate of speed, in turn, depends upon such factors as traf- fic and road conditions and the number of stops — either for traffic or to service customers — which are involved. It was not possible to include, in the analysis, the influence of road and traffic condi- tions, however, as these are difficult to measure objectively. Nor was an attempt made to obtain information on the num- ber of traffic stops or the corresponding time requirements. The results, therefore, represent average traffic conditions for the routes studied. Presumably, average rates of driving speed would vary with the different con- ditions met in driving from plant to route, between customer stops while on the route, and from route to plant. For this reason, data were obtained on the distance traveled between these points and on the driving time required. In ad- dition to the distance factor, the number of customer stops made by the driver was available, and was used in multiple [24 correlation analyses of the time spent in driving between the first and last custo- mer stop. The results of these analyses are summarized in table 4. Driving time was closely related to route distances as indicated by the fact that, in all instances, this single factor explained more than 65 per cent of the total variance in time requirements for this function. If the slight variations in rate of travel on the various subdivisions of the wholesale routes are ignored, and if the separate coefficients are combined on the basis of weighting by the appro- priate average distances traveled, the fol- lowing expression of total driving time requirements for wholesale routes is obtained. Td = 27.54 + 1.11M + 0.28C (1) where Td represents total route driving time in minutes, M the total daily miles, and C the number of customers served. Note that this expression of driving time results from an analysis of many routes operated under different conditions. For this reason, the regression coefficients — particularly that relating to distance traveled — should not be considered as representing the "marginal" speed of wholesale milk trucks. Many of the dif- ferences in miles traveled occurred be- tween routes operating in densely popu- lated areas and those serving outlying areas separated by relatively sparse re- gions. Routes serving the former were over heavily trafficked city streets, while part of the latter driving was on high- speed highways. Thus, the 1.11 minutes per mile expressed above does not mean that wholesale milk trucks in California travel at a rate of 54 miles per hour! Direct Wholesale Delivery Time Requirements The operations under consideration in this section are those performed at the customer's establishment, by the route driver in providing delivery services. Time observations for each set of services began when the truck arrived at the buyer's premises, and ended when the driver re-entered the truck cab, ready to continue his route. Individual recordings were made of the amount of time spent in each of the following categories: (1) securing order; (2) putting up order; (3) delivering; (4) waiting; (5) talk- ing; (6) securing signature or (7) se- curing cash; and (8) completing de- livery. Provision was further made for any functions not specifically noted Table 4. Effect of Distance Traveled and Number of Customers Served on Time Spent in Driving, Wholesale Milk Delivery Routes, All California Markets Combined Area covered Total distance Time per day (fixed) Time per mile Time per customer Corrected correlation coefficient Plant to first stop per cent 6.9 83.7 9.4 minutes 2.36 21.10 4.08 minutes 1.26 1.08 1.26 minutes 0.28 0.886 First to last stop 0.833 Last stop to plant 0.830 Total 100.0 27.54 1.11* 0.28 * Obtained by weighting the average mileage by the proportion of the mileage traveled in each phase of the total route coverage. [25] above so that the sum of the time spent on all functions or operations accounted for the total time spent at the customer's premises. Each of the deliveries was classified into one of the following types: (1) gro- cery stores where delivery did not in- clude placing or arranging products in display case; (2) grocery stores where the delivery did include display case services; (3) restaurants where the prin- cipal item of delivery (usually half pints of milk) was in paper containers; (4) restaurants where the principal item of delivery was in glass containers; (5) drugstore fountains and dairy bars; and (6) schools, institutions, and commis- saries. Table 5. Cumulative Frequency Distributions of Number of Customers and Volume, by Volume Groups, Groceries, and Markets,* Los Angeles, 1950 Volume per stop Cumulative percentage of number of customers Cumulative percentage of volume delivered labor units 0- 14 11.3 25.0 50.6 66.7 76.7 81.9 85.5 88.3 89.8 91.2 92.4 93.0 93.9 94.9 95.8 96.5 97.1 100.0 1.1 3.7 12.8 22.3 30.7 36.2 41.0 45.4 48.1 50.9 53.5 55.1 57.5 60.6 64.1 66.8 69.3 100.0 15-24 25- 49 50- 74 75- 99 100-124 125-149 150-174 175-199 200-224 225-249 250-274 275-299 300-349 350-399 400-449 450-499 500 and over * Not including those receiving special display case services. For these types of customer outlet, the data were broken down into relatively small groups according to the volume of delivery. Table 5 presents the cumulative frequency distributions of the percent- ages of customer numbers and of total volume for the classification of grocery stores where no special display case serv- ices were provided. While these data re- late to the Los Angeles market only, they typify those found in all other Cali- fornia markets. Note that half of the cus- tomers received less than 50 labor units of milk products per delivery, but that these customers accounted for only 13 per cent of total volume. At the other ex- treme, less than 3 per cent of the custom- ers received 500 or more labor units, and these large customers accounted for more than 30 per cent of total volume. (A labor unit refers to a system of weights, de- veloped by the Bureau of Milk Control, by which the various products and con- tainer sizes handled on milk routes can be aggregated. In theory, each labor unit is equivalent, in terms of its labor re- quirement, to a quart of milk. These labor unit modifiers for the various items are given in Appendix B, tables B-l and B-2, pp. 74 and 75.) With the exception of the very small customers (receiving 24 labor units or less per delivery), most wholesale outlets received milk products from more than one distributor. The number of distribu- tors serving each customer who made purchases from independent distributors, again for the Los Angeles market only, is shown in table 6. This means, of course, that the aver- age size of customer is considerably larger than the average volume per de- livery and that the economies of large- volume deliveries are not currently being realized to full advantage. On the basis of an intensive analysis of a relatively few of the total number of routes for which data were available, it was decided to use the single, inde- pendent variable of volume (measured [26] Table 6. Percentage of Distributors Delivering to Customers in Each Delivery Volume Group, Los Angeles, 1950 Size of stop Per cent of total stops Number of distributors per customer 1 2 3 4 5 labor units 0-24 25-48 49- 72 26.9 26.0 16.2 10.0 13.7 7.2 per cent 64.8 43.2 41.6 41.0 41.0 37.9 per cent 24.5 40.9 42.8 41.0 38.0 37.9 per cent 8.3 13.2 13.6 14.3 18.2 20.0 per cent 2.1 2.2 2.0 3.7 2.5 3.1 per cent 0.3 0.5 73-96 97-192 193 and over 0.3 1.1 in terms of labor units) to explain varia- tion in the amount of time spent in direct delivery. These trial analyses further brought out the fact that the functions were not linear and that the relatively simple curve types (such as second- degree parabolas) that could be fitted mathematically were inappropriate for the entire range of the data. These curve types are inappropriate in that they con- sistently overestimate time requirements in the low-volume ranges. One of the great advantages of the use of regressions which have been fitted mathematically lies in the fact that inter- pretations can be made with respect to various companion statistics concerning the significance of the regression coeffi- cients and the reliability of the regres- sion itself. These measures (such as t- ratios and standard errors of estimate) are based on certain characteristics of the dispersion of normal distributions. While the frequency distributions of data to which these measures are applied are often not "normal," the error so intro- duced is overlooked in favor, of the con- venience and general acceptability of the method. A further condition for the ap- plication and usual interpretation of the standard error of estimate 8 is that the 8 The standard deviation of the residuals of the actual observations around the regression. variances of the original observations be equal to all points along the regres- sion. This is to say that the variance is not functionally related to the magnitude of the independent variable. A test was made to determine whether the dispersions of the data being ana- lyzed were homoscedastic. (Arrays in which the variances are equal are said to be homoscedastic; in the contrary case, heteroscedastic.) This investigation made it clear that the variances of the time requirements are not equal, but are an increasing function of the volume of milk and dairy products delivered. For this reason, and because none of the rela- tively simple curves appeared to be ap- propriate, it was concluded that use of the more orthodox forms of mathe- matical regression and correlation tech- niques could not be justified in the analysis of the present problem. Rather, the decision was made to classify the data into relatively small groups according to the volume de- livered. The time-volume relationships then could be determined on the basis of the means of these groups. The vari- of the individual observations ance around these means could be calculated, and the differences between the variance of the groups could be considered in order to arrive at meaningful statements [27] with respect to the significance and re- liability of the basic relationship. Since the standard error of estimate is, in fact, the standard deviation of the residuals, these standard deviations can be de- termined for each group, thus providing a measure of the dispersion of observa- tions comparable to the standard error of estimate but taking into consideration the inequality of variances in the data. Furthermore, estimates of the standard error by size groups can be made of the time-volume relationship. This will then provide a basis for describing the relia- bility of the relationship in terms of probability statements of expectations of the range in which similar time-volume means (from other samples from the same universe) will fall. It is recognized that independent consideration of the several volume classes will somewhat overestimate the standard error range of the regression, but this appears to be the most appropriate method that could be devised. Labor requirements for customers receiving full service. When the data were segregated into volume groups, the average time per stop was calculated for each of the operations performed in pro- viding delivery services. The same calcu- lations were carried out for each type of Table 7. Average Direct Time Per Stop for Specified Delivery Operations, Groceries and Markets, Los Angeles, 1950 Volume per stop Secure order Put up order Deliver* Wait and talk Collect Complete delivery Total full service labor units time in minutes pe r stop 0- 14 15-24 25- 49 1.28 1.54 1.85 2.16 2.32 2.70 3.34 3.02 4.06 3.19 3.80 4.38 4.86 5.24 4.66 5.21 5.82 5.03 3.73 5.42 7.10 5.15 9.57 0.72 1.22 1.73 2.37 3.06 3.55 4.49 4.48 5.60 5.36 5.25 6.15 6.93 7.84 6.99 8.60 8.17 10.99 10.76 9.68 11.06 13.60 14.19 0.50 0.59 0.70 0.88 1.00 1.51 1.85 2.19 2.63 2.07 2.76 3.95 3.65 4.35 4.51 5.22 4.46 4.59 7.61 7.44 10.51 5.60 12.82 0.14 0.31 0.36 0.73 0.72 0.70 0.80 1.46 0.72 0.61 0.65 4.68 0.23 2.38 0.40 1.38 0.38 2.41 1.39 0.33 2.39 0.43 6.90 0.77 0.54 0.73 0.94 1.61 1.57 2.00 2.10 1.68 2.05 2.14 2.08 2.81 2.80 2.88 2.49 2.47 3.52 3.57 4.84 4.39 4.12 3.43 0.53 0.63 0.78 1.09 1.24 1.46 1.70 1.96 2.15 2.34 2.86 2.94 2.09 3.52 3.63 2.89 2.97 4.62 3.95 5.88 5.73 2.37 7.47 3.94 4.83 6.15 50-74 75-99 8.17 9.95 100-124 11.49 125-149 14.18 150-174 15.21 175-199 16.84 200-224 15.62 225-249 17.46 250-274 24.18 275-299 20.57 300-349 26.13 350-399 23.07 400-449 25.79 450-499 14.27 500-599 31.16 600-699 31.01 700-799 33.59 800-899 41.18 900-999 31.27 1,000 and overf . . . 54.28 * Excluding stops where products were arranged or placed in display case. These display case services were provided in less than 20 per cent of all grocery stops in Los Angeles. t Average volume per stop of "1,000 and over" approximately 1,800 labor units. [28] >s ^ k. a> > a> x^**"^ ts w. X x^^^^ X a> 30 Q. V) Q> X Z3 C % ^r *^^ X E 20 ^ x Vy^o x.°( io JQ O • x o X Groceries and markets _ JO £>'x o* Restaurants, fountains u * and bars a> i_ '5 ) 200 400 600 800 Volume per delivery (in labor units) Fig. 1. Effect of volume per delivery on delivery labor— wholesale milk routes. 1000 stop classification previously mentioned. These averages relating to deliveries to groceries and markets not receiving spe- cial display case services are presented in table 7. The way in which the total direct time requirements per stop increase with increases in volume per stop is indicated in figure 1 where the "total full service" column of the table is presented by x's. (The analysis of total direct delivery time has been made by summing the average time requirements for each op- eration in order to eliminate the influ- ence of variation in the amounts of serv- ice provided by the driver.) Since no consistent differences could be found be- tween similar sums of these averages calculated for the two restaurant classi- fications and fountains and bars, these groups were combined, with the results shown in figure 1 by o's. As shown by the curve drawn through them, these points follow a relatively uniform pat- tern. This curve, used to describe the time-volume relationship for wholesale milk deliveries to outlets of these types, smooths out the random variation in the time observations present in the sam- pling distribution. The same type of average time require- ment calculations made for schools, in- stitutions, and commissaries differs sub- stantially from those for the grocery and restaurant classification shown in the figure. A comparison of the amount of time required to serve equivalent vol- umes to groceries and to the institutional type of outlet is shown in table 8. These differences exist in spite of the fact that allowances have already been made for differences in the types of product and container sizes delivered and for varia- tions in the number of services provided. Since the time-volume relationship in figure 1 was fitted graphically, the stand- ard error of estimate was calculated from the deviations between actual and esti- mated values for individual observa- tions. 9 Correlation coefficients were then calculated from these standard errors of estimate and the standard deviations -4 2& [29 of the originals, and are presented in table 9. 10 As noted before, however, the procedure used in calculating these sta- tistics — based on independent considera- tion of the individual volume classes — will tend to overstate the standard errors and thus underestimate the correlation coefficient. While the figures do give some indication of the reliability of the relationships presented, they must, how- ever, be interpreted somewhat loosely. Similar analyses were made of whole- sale time requirements for the six other >F3 markets studied. With the exception of one component operation — that of "deliver" — results closely approximated those for the Los Angeles market. This reflects the fact that the basic operations for these component elements, such as "secure order," "put up order," "col- lect," and others, were performed essen- tially in the same manner and to the same degree in all of the markets when routes were analyzed. Subsequent investi- gations to determine why time require- ments for "deliver" differed in the Los Angeles market showed that the opera- tion itself was essentially different. In all markets except Los Angeles, the route Table 8. Comparison of Average Direct Time Requirements for Full- Service Deliveries to Groceries, Markets, Schools, Institutions, and Commissaries, Los Angeles, 1950 Volume per stop Direct time per stop: groceries and markets Direct time per stop: schools, institutions, and commissaries labor units 0-14 minutes 3.94 4.83 6.15 8.17 9.95 11.49 14.18 15.21 16.84 15.62 17.46 24.18 20.57 26.13 23.07 25.79 24.27 31.16 31.01 33.59 41.18 31.27 54.38 minutes 4.69 4.52 5.33 7.31 8.84 6.16 8.02 8.05 9.98 6.58 15.42 12.15 20.78 7.66 15.05 11.13 27.66 16.68 28.17 13.53 15-24 25-49 50-74 75- 99 . 100-124 . 125-149 150-174 175-199 200-224 225-249 250-274 275-299 300-349 . 350-399 400-449 450-499 . 500-599 600-699 700-799 800-899 900-999 1,000 and over [30] Table 9. Correlation Between Direct Time per Delivery and Volume per Delivery for Specified Types of Wholesale Outlets, Los Angeles, 1950 Type of outlet Uncorrected correlation coefficient* Proportion of variance in time which can be explained in terms of volume delivered Groceries without display case services 0.864 0.885 0.650 0.504 0.661 per cent 75 Groceries with display case services 78 Restaurants receiving products in paper 42 Restaurants receiving products in glass 25 Fountains and bars 44 Total of all observations 0.869 76 * Correction factors were not applied because of the indeterminacy of the number of degrees of freedom present in a function of this type. The number of observations was sufficiently large, however, so that the amount of correction required will be quite small. t (18.0) 14 12 10 2 - o Time requir x Time requir six other ements for Los Angeles ements for > : X X X X X X X x x X o o x x X - X X f/oo °° ° o o o o o o o 100 200 300 Volume per delivery (in labor units) Fig. 2. Comparison of time for "Deliver" element, Los Angeles, San Francisco, Contra Costa, Sacramento, Stockton, Fresno, and San Diego markets— wholesale milk routes. [31] 400 driver is expected to place the day's de- liveries in the refrigerated display case, while in Los Angeles no such services are normally provided. The differences in the time requirements for the single function of "deliver" are shown in figure 2. The average direct delivery time re- quirements for full service operations for the six markets other than Los Ange- les are shown in table 10. A "full service operation" is defined as a delivery in which all of the individual elements are performed. This includes "secure order," "put up order," "deliver," "collect," and "complete delivery." In other words, if the driver knew the order in advance and, therefore, was not required to per- form the "secure order" element, it would not be considered as a "full service" de- livery, and the observation would be eliminated from this part of the analysis. The averages are presented, by volume classes, for each market as well as for all markets combined. In addition, for each market group there is a column headed "standard deviations." It should be noted that these are not standard de- viations around the individual means in the usual sense. Rather, the statistics have been calculated from the individual observations in each group. They are based on the sum of squared residuals of actual time from estimates of time which were derived from a relationship (smoothed by freehand methods, to the means of all markets combined) . In gen- eral, this table illustrates that the vari- ances of the groups increase with in- creases in size of volume classification. On the other hand, there do not seem to be significant differences between mar- kets as judged by the group-by-group magnitude of these standard deviations. No one market shows a definite pattern of consistently larger or smaller standard deviations than the others. Since such bias was not observed, it was taken as partial justification for combining data from the different markets. The fact that no significant differences were found among time requirements in these different geographical markets is not surprising. Since individual opera- tions or elements of delivery are the same, regardless of the market in which they occur, the time requirements might be expected to be similar. In other words, if a route driver has to go through a series of physical operations to make a delivery, it makes no difference whether he does so in San Francisco or in Fresno. On the basis of this reasoning, the results of these studies should be applicable to areas other than California, provided the basic elements or components of delivery operations are similar. Figure 3 shows the effect of the added time requirements — display case services commonly provided by the wholesale milk driver in California markets other than Los Angeles. Here the smooth rela- tionship previously presented in figure 1, representing the total "full service" time requirements for the Los Angeles market, is reproduced and compared with a relationship similarly constructed which portrays the "full service," time- volume relationship for the six other markets. Since the difference between these two relationships reflects primarily the different degrees of service rendered (the delivery operation in the six mar- kets includes filling and arranging the display case, whereas in Los Angeles it does not), these different relationships are labeled "delivery with display case service" and "delivery without display case service," respectively, in the dia- gram. Labor requirements for customers receiving less than full service. Up to this point, the discussion of labor requirements has been appropriate for deliveries in which the driver performs the full set of services — that is, secure order, put up order, deliver, wait and talk, collect, and complete delivery. In many instances, not all of these oper- [32] Table 10. Average Direct Time Requirements and Standard Deviations,* Wholesale Milk Deliveries, Groceries, Restaurants, and Fountains, Six California Markets, Full Service Operations 03 M u ti s < II cox) a o |H 4) Pi 01 43 d d a ot>Mtoo>McocoqcnHWNMNcnHqMH HTHNNNWMM^eOIO^lOIOtOt-C'OOHrl CO CN d Ja-2 IS COT) ^wi«qqq^t>;^oqcq^cioq , «3jr> t-j NNNNMHNMiOrirli^riNd^ ! Oi II >** < (NtOHMtOt>l>l>«OOSNM^M(CH -^ idid^oococnHriMNdio^todri ' o »Hi-li-lrH,-l,-liHi-ICNCN . CN . . OS id o d 4) B s u a c3 CO X) d Js-2 COX) hnnnmn^^'mm^n^io : d CO i> ^ 00 ID I 1 coddi>ciddeoNNdioi>N ; 06 cn «d *-< 1-ltHi-lT-lrHT-li-lT-ICN . H CO (N CO CN CN CO d o o o CO x) d 53-2 «1 ca > +* 4> COx! ^QNONWI>N^^CDNCOTHNt;NM CO HHNcoMco'^'i'od^^io^ridddd ' i> 1-1 »H 8.1 tONeoMNt^HiocotONcn^WNflqoqw • t> Modt-flioJHdN^dd^oidoi^w "o rlHHHHHHHHNCOM . -tf o d 03 4) U x> d COX) i^qNiOrjNqcqcoMNqqqqHinco'j' T-J CO s 3 wqqwqiqqqMqnqqqN^wcnto • w^dNco^drlwdwiow^Noi^N^ HHHHrlrlHNHNNN od eg 03 o o e3 is d o O es x) V 1 3 •o d ■§1 I? COX! MqqeowHcqqwHiocoooMNHNN^q 43 4) .5 < ^N^^Niooq^qqoqrjqHqc-qcoqq MTjiidcdo6aJ©^eornidc^t>o6coidi>i>t>o6 HHHHHHHHrlNNNININ -t— q O u 03 '3 d £ d CO xi d Je 2 «i -M 4) COT) qq^wi>oiot.ooHMoqeorjH • co i-h o co H H rl OS 4) S3 S > e3 m d i'S.2 »■— o d d h O Xi e3 HNCO^lOlOt-WOJHMWt-cneOt-HlOOJ HHHHHNNCOCOCO oooooooooooooooooooo HNM^iOtOt-COOJO(NTji«COO^CON«0 HHHHHNNNMW o o 200 400 600 800 Volume per delivery (in labor units) 3. Effect of volume per delivery on direct delivery labor with and without display case service— wholesale milk routes. 1000 ations are carried out. For example, the daily requirements for some customers are known by the driver prior to his arrival, and the operation, "secure or- der," is eliminated. Furthermore, various forms of streamlining the set of opera- tions performed by the driver may exist. Depending upon the nature of the com- pany policy in these cases, therefore, only certain combinations of operations and their respective time requirements are appropriate for specific types of de- liveries. It will be recalled that the time-volume relationship discussed under direct time requirements was based on the summa- tion of the averages in columns 1 through 6 in table 7. The average time require- ments for each operation were calculated on the basis of only those stops where the operation was, in fact, performed. (This statement is not true for the aver- ages included as "wait and talk." These figures represent the average amount of time spent in this manner for all stops, including those where no waiting or talking was done. This procedure seems justified by the chance or random nature of these time categories.) It was possible to determine the time requirements for stops receiving less than full service by including in the total only the amounts of time required to perform those opera- tions which were actually carried out. Two types of streamlined services ap- pear to be particularly interesting: (1) that in which "secure order" and "col- lection" are eliminated; and (2) that in which, in addition to the above, the op- eration of "delivery" is eliminated by placing the products on the sidewalk or platform. The first of these involves a set of operations similar to the services normally performed in a retail delivery operation where the order typically is known in advance and collection is made on a weekly or monthly basis. The sec- ond is applicable to a set of minimum [34] services characterized by sidewalk de- livery. The total time requirements for stops with these differential types of serv- ices are shown in table 11 where service type "A" refers to the full set of services discussed in previous sections; type "B" refers to stops where the order is known in advance and collection is made other than by the route driver; and type "C" is appropriate for a minimum set of op- erations with sidewalk delivery. Smoothed regressions have been de- termined for time requirements based on these types of service differentiation and appear with the time-volume relationship for full-service stops in figure 4. The savings in direct delivery time that can be made with these streamlining opera- tions can be readily seen. In the same amount of time, a driver can make a de- livery of 90 labor units under type "A," approximately 165 labor units with type "B," and over 250 labor units when type "C" services are performed. Reliability of the time-volume re- lationship. As indicated earlier, the variances associated with the average time requirements are not constant but tend to increase with increases in the vol- ume delivered. The standard errors of Table 1 1. Comparison of Time Requirements for Alternative Types of Delivery Services Volume per stop Type of service* "A" "B" "C" labor units 0- 14 15- 24 minutes 3.94 4.83 6.15 8.17 9.95 11.49 14.18 15.21 16.84 15.62 17.46 24.18 20.57 26.13 23.07 25.79 14.27 31.16 31.01 33.59 41.18 31.27 54.38 minutes 1.89 2.75 3.57 5.07 6.02 7.22 8.84 10.09 11.10 10.38 11.52 17.72 12.90 18.09 15.53 18.09 15.98 22.61 23.71 23.33 29.69 22.00 41.38 minutes 1.39 2.16 25-49 2.87 50-74 4.19 75- 99.. 5.02 100-124 5.71 125-149 6.99 150-174 7.90 175-199 8.47 200-224 8.31 225-249 . . 8.76 250-274 13.77 275-299 9.25 300-349 13.74 350-399 11.02 400-449 450-499 12.87 11.52 500-599 18.02 600-699 16.10 700-799 15.89 800-899 19.18 900-999 16.40 1,000 and over 28.56 * Service type "A" includes secure order, put up order, deliver, waiting and talking, collecting, and com- pleting delivery. Service type "B" excludes secure order and collect from above. Service type "C" involves sidewalk delivery and eliminates the delivery time included in type "B." In all cases, these figures are based on deliveries which do not include display case services. [35] 60 50 Q. 10 40 £ 30 20 10 Fig. 4. ^^^ Type ^ "" i^-Type "B" _____ Type "c"^ - "■"y""" > 200 400 600 800 Volume per delivery (in labor units) Effect of volume per delivery on direct delivery labor time with alternative amounts of service performed— wholesale milk routes. 1000 D 200 400 600 800 Volume per delivery (in labor units) Fig. 5. Standard deviations of residuals of actual values from estimated values- wholesale milk routes, Los Angeles, 1950. [36] 1000 estimate for the various size and stop classifications for full-service deliveries are presented in figure 5. A smooth rela- tionship, passing through the average of the standard deviations of residuals," was fitted by freehand methods to these ob- servations and is represented on the diagram. This linear function was then applied to the smoothed time-volume re- lationship (shown in fig. 1) to determine a band around the regression as shown in figure 6. This band, which represents one standard deviation of residuals above and below the regression, includes approxi- mately two thirds of the individual obser- vations of the sample. It can be seen that the dispersion of the individual time observations around the mean relation- ship is considerably greater for large- volume than for small-volume customers. It should be noted, however, that the distribution of time observations for any 11 Note that these are the standard errors of estimate but that the above term will be used in the discussion in order to reduce confusion in terminology. volume group tends to be somewhat skewed, with the modal time displaced toward the lower end of the distribution. This means that, on the average, the dis- tance between the lower limit to the band and the regression will include somewhat more than one third of the cases, and con- versely, that the distance between the regression and one standard deviation of residuals above the regression at any point will contain slightly less than one third of the observations. The linear relationship of the standard deviations of residuals has also been uti- lized to determine the smoothed approxi- mation to the standard error of the regression. By dividing the value of the standard deviations of residuals (as in- dicated by this relationship) by the square root of the number of observa- tions found in the sample for each vol- ume group, estimates of these standard errors were obtained. These appear in figure 7 where irregularities have been smoothed out by a continuous curve to eliminate the effects of chance variations in the volume frequency distribution. >, bO 0) > 5 50 2.4 a; -o i_ 2.0 to If these interconnections are substituted in the above equations, we obtain: T = 96.74 + 3.61 ^ + 0.08556F r V s - 0.0000806F r F s + l.UD (6b) T = 96.74 + 15.04 V s + 0.024875W + l.UD (7b) To simplify these equations further, we may ask what interconnections there are between distance and the other fac- tors since changes in either route volume, number of customers, or volume per de- livery might well be reflected in changes in route mileage. Analysis of the data for the routes studied, however, indicated that none of these factors — independently or in combination — had a significant ef- fect on distance. Multiple correlation re- ] suits indicated that neither the correla- tion nor the regression coefficients were of significant magnitudes. For this rea- son, the following computations have been based on a constant route distance of 36.3 miles per day, the average for all routes studied in California markets. 1 ' Inserting this in the above equations yields the equations below, where time is expressed simply in terms of the two factors, route volume and volume per customer: T = 137.03 + 3.61 ^ + 0.08556 V r V s 0.0000806 V r Vs T = 137.03 + 15.04 + 0.024875 V r (6c) Vr v 8 (7c) Note that it would be appropriate to sub- stitute the mileage figures for the dif- ferent markets in determining specific relationships for individual areas. We have now specified total route time as a function of route volume and volume per customer. Moreover, by selecting an appropriate length for the workday, these functions will enable us to represent route volume as a function of volume per delivery — the essential problem in determining the effect of volume per stop on delivery costs. While any length of workday could be selected, labor con- tracts in California markets specify the basic day as eight hours (exclusive of lunch periods), and most routes operate very close to this limit. The results of substituting T = 480 minutes in the above equations are shown in figure 10. Volume 13 The average route distances in the different markets ranged from about 17 miles per route in San Francisco to over 46 miles per route in Fresno. 8000 200 400 600 800 Volume per delivery (in labor units) 1000 Fig. 10. Effect of volume per delivery on total route volume and number of customers served— wholesales milk routes. [47] per delivery is found to exert a most pro- nounced effect on route volumes, with total volume increasing at a decreasing rate to more than 8,000 units as average volume per stop increases up to 1,000 units. Figure 10 also shows the effect of vol- ume per customer on the number of cus- tomers served, obtained simply by dividing total volume by the volume per customer. As would be expected, this curve falls off rapidly in the low-volume- per stop ranges, and then more gradually as volume per stop exceeds 200 units. Both of these curves have been given as smooth, continuous functions. It is clear, however, that fractional customers are impossible and that both curves should therefore be discontinuous to reflect the discontinuous change in the number of customers. The discontinuities would be very small in the low-volume ranges, however, while in the higher ranges the tendencies for the curves to flatten out would reduce their importance. For these reasons, plus the fact that adjustments to whole customers may be made by ob- vious methods, the smooth curves are used here. Effect of Volume per Stop on Total Delivery Time per Stop The determination of total route vol- umes and customer numbers, as discussed in the previous section, simultaneously determines the allocation of total labor on a per stop basis. Heretofore, the only part of the route driver's time that could be allocated to an individual stop was that spent at the customer's premises (treated in the section concerned with direct delivery time). Driving time and miscellaneous route time, "indirect" from the standpoint of the contribution of any specific stop or stops, have been discussed only in terms of route totals. The total time per stop — direct and indirect — has been determined by divid- ing the 480 minutes in an eight-hour day by the number of customers served per route (on the basis of the relationship discussed in fig. 10), and appears in figure 11. Since the number of customers served per route is a decreasing function of the average volume per customer, and the amount of indirect (driving and mis- cellaneous) time is constant per route, the resulting amount of indirect time per "otal delivery time in minutes per deliver o o o O o c - - / 200 400 600 800 Volume per delivery (in labor units) Fig. 1 1. Effect of volume per delivery on total delivery labor— wholesale milk routes. 1000 [48] stop is not constant, but is an increasing function of the size of stop. This pro- cedure is the equivalent of using the ratio of the amount of indirect time per route (137.03 minutes) to the number of cus- tomers, and combining it with the direct time relationship shown in figure 1 (p. 29). Labor Costs per Stop- Wholesale Time requirements per stop have been converted to labor costs by using the current union contract for the Los An- geles market and assuming a 6-day, 48- hour route week. Regular drivers average 44 hours per week and work 11 days out of each two-week period. Overtime is paid at the rate of time-and-a-half for work in excess of 40 hours per week, and 4 hours of overtime are guaranteed. Re- lief drivers are also paid on the same basis, but at a slightly higher hourly rate. Other labor costs include vacations, sick leave, health and welfare program con- tributions, and unemployment and com- pensation insurance. Details of these costs are given in table 15. Using $24.89 per day for route labor cost, the total delivery labor cost per stop appears in figure 12. Since wages and allowances paid to drivers for an eight- hour day under the Los Angeles contract are not influenced by size of stop or total volume, the direct determinant of labor cost per stop is the number of stops made — customers served (fig. 10). Dividing labor cost per day by the route volume, of course, would determine the labor cost per unit delivered. It should be emphasized that the labor cost per stop relationship presented in figure 12 merely represents the physical time requirement data which appeared (dollars) b o >» a> ■2 2.00 a> •o QJ a. — labor cost b o >s V- > O 200 400 600 800 Volume per delivery (in labor units) Fig. 12. Effect of volume per delivery on delivery labor cost per delivery, based on Los Angeles wage rates, 1955— wholesale milk routes. 1000 [49] Table 15. Calculation of Whole- sale Delivery Labor Costs, Los Angeles Wage Rates, 1955 Factor Rates per week Regular driver $ 104.75 21.63 5.53 10.56 5.23 1.66 Relief driver Foreman Holidays, vacation, health and welfare Unemployment taxes and com- pensation (3.67 per cent) .... Social security Total labor cost per route per week $ 149.36 Labor cost per day (six-day week) $ 24.89 in figure 11, converted to the currently appropriate labor cost per minute (about 5.2 cents). (This is, of course, appro- priate only when wage rates are on a flat daily basis and no commissions or bonuses are involved which affect the labor cost per minute.) It is thus easily possible to make adjustments in labor costs for changes in the wage rates. Such changes will be reflected in a different labor cost per minute which, when ap- plied to the basic physical input require- ments, will provide an appropriate delivery labor cost relationship. Further- more, wage rates prevailing in other mar- kets may be applied to the physical relationship in order to determine ap- proximate delivery labor costs in areas where delivery conditions and techniques are similar to those in Los Angeles. Truck-Operating Expenses- Wholesale The present study did not include an analysis of truck costs based on original records, but summary data of such costs were available in sufficient detail to pro- vide the necessary information on the nature of these costs. The Bureau of Milk Control made an intensive study of proc- essing and distribution costs for firms operating in several California markets during the last quarter of 1953. Included in these cost analyses is a breakdown of the elements of truck expense. These summaries contain both direct operating costs — such as gasoline and oil, tires, re- pairs, depreciation, taxes and licenses, et cetera — and indirect operating expenses, such as ice, supplies, laundry, and the general overhead allowances for the oper- ation and maintenance of garage and storage facilities. The total expenses in these categories applicable to wholesale milk delivery trucks, and the number of wholesale milk route days operated were made available by the Bureau for each of the firms studied. In addition, where possible, the total volume (in labor units) of milk products distributed during the three- month period was also provided. From this information, it was possible to cal- culate the average truck-operating cost per route day ,and the average volume per route for each of these companies. These averages have then been weighted by the number of routes operated, and are shown in figure 13. Since some com- panies operated larger trucks, carrying larger total volumes, than did other com- panies, these weighted averages indicate that the daily truck costs increase as the [50] o ■D £d.UU 24.00 20.00 16.00 12.00 Weighted average of smaller load sizes -v . ^^*^"^ ** Weighted average of larger load sizes 8.00 4.00 - 2000 4000 6000 8000 IQOOO Daily route volume (in labor units) Fig. 13. Effect of route volume on daily truck-operating costs— wholesale milk routes. ~ 3 to o o 2 CD Q. O a 3 00 00 - - 200 400 600 800 1000 Volume per delivery (in labor units) Fig. 14. Effect of volume per delivery on truck cost per delivery, Los Angeles, 1955— wholesale milk routes. [51] average size of load increases. The de- tails of why these costs per day increase cannot be determined on the basis of the accounting records available. It would be logical to assume, however, that such factors as larger trucks being required to handle larger loads would have some influence. The larger truck, of course, would involve heavier depreciation charges, greater rates of gasoline, oil, and tire consumption, et cetera. If two trips are made with a smaller truck in order to handle the additional volume, then total route mileage would be in- creased and again truck costs per route day would increase. A straight-line relationship has been determined on the basis of the weighted averages of the four smaller load sizes and of the two larger load sizes. The equation describing this relationship is: TC = $7.41 + 0.0012226F r (11) where TC represents truck-operating ex- penses per route day, and V r represents the total volume handled per route, ex- pressed in labor units. Estimates of the total truck costs per day for different sized routes can be calculated from this relationship. Knowing the route volume and the number of customers served (per eight-hour route day) that are associated with various sizes of customer stops (from fig. 10), calculations can be made of truck-operating cost per stop by meth- ods already described. These costs are shown in figure 14. As was the case in the consideration of labor costs, the basic determinants of truck costs are the physical input require- ments which were developed from the time study analyses. So long as technol- ogy and the conditions of delivery in the market remain unchanged, these in- put requirements will be unaffected. Changes in costs resulting from changes in the level of prices and cost rates can then be adequately accounted for if the new cost rates are known. Total Delivery Costs-Wholesale Labor costs and truck costs can now be combined to show the effect of volume per stop on total delivery costs. Figure 15 represents the cost per stop for pro- viding the full set of service operations to customers receiving different volumes. Although this relationship is curvilinear, it may be well approximated by the use of the two linear functions shown in the figure. The equations describing these linear approximations are: C = $0.35 .00844 V s (12) which is applicable for volumes per stop of less than 221 labor units: and, .43 + $0.00355 V s (13) [52 which will approximate the delivery cost of stops where more than 221 labor units are delivered. In both instances, C repre- sents the total delivery cost per stop and V s the volume of milk products per stop measured in terms of total units. As can be seen from this diagram, total delivery costs increase with increases in the size of delivery, but at a less than proportionate rate. This means, for ex- ample, that it does not cost as much to make one 400-unit stop as it does to make four 100-unit deliveries. Furthermore, there is a certain amount of fixed cost in- volved in making a customer stop which is independent of the volume being de- livered — which at present cost rates amounts to about 35 cents. When con- verted to terms of the delivery cost per ] 200 400 600 800 Volume per delivery (in labor units Fig. 15. Effect of volume per delivery on total delivery costs per delivery, based on Los Angeles cost rates, 1955— wholesale milk routes. 1000 unit (fig. 16), this relationship points out that delivery costs are high for small- volume deliveries and that these costs fall rapidly as the volume per stop in- creases until a volume of about 200 labor units is reached, and then decline much more slowly for deliveries larger than this amount. 1 ^^" «• 3 240 ai °- 200 Q) > 120 80 40 - ^y^ - Daily rou volume V te / Number custome - - routes studied (132.5 feet) for F and further substituting the average retail route mileage (33.7 miles), the equation simplifies to: T = 120.53 + 1.4464 — + 0.1178F r Vs (10c) which expresses total route time in terms of the two variables, the volume per stop and the total route volume. Since retail routes in California are also normally organized on an 8-hour basis, we sub- stitute 480 minutes in the above equation to determine the effect of variations in volume per stop on total route volumes and number of customers served per route day. The results of these interrela- tionships are shown in figure 17. Volume per stop has a definite influence on daily route volumes where total load sizes in- 1400 1200^ "E 1000° 800- 6 01 Q. W o ° 2 4 6 8 Volume per delivery (in labor units) 10 Fig. 18. Effect of volume per delivery on total delivery labor, Los Angeles and Fresno markets combined, 1954— retail milk routes. crease from less than 250 units with an average of 1 unit per delivery to nearly 1,300 units when 10 units are served per customer. The decrease in number of customers per route with increased de- livery sizes reflects, of course, the fact (shown in fig. 9, p. 44) that total direct time requirements are greater for large- than for small-volume customers. Effect of Volume per Stop on Total Delivery Time per Stop The influence of delivery volume on delivery time per stop is shown in figure 18. This has been obtained by dividing the 480 minutes per day by the number of customers per route as determined in figure 17. As mentioned in the wholesale route study discussion, this is the equiva- lent of assigning the "indirect" time — such as driving time and miscellaneous route time — in ratio to the number of customers served per route (a decreasing function of delivery size) and adding this allocation of indirect time to the direct time requirements presented in figure 9 (p. 44). Labor Costs per Stop-Retail The time requirements per stop have been converted to labor costs by applying appropriate wage rates. In all California markets, except Los Angeles, wages paid retail route drivers are on a straight salary basis. In other words, the total wage cost per day is constant, regardless of route volumes. In Los Angeles, how- ever, a change was made in wage pay- ments which incorporates a guaranteed base plus a commission on all deliveries in excess of a basic quantity. Since this basic volume is lower (423 units) than average load sizes in the market (473 [55] Table 16. Calculation of Retail Delivery Labor Costs, 1955 Wage Rates* Factor Rates per week Regular driver $ 86.60 35.82 4.52 13.83 5.17 1.94 Relief driver Foreman Holidays, vacation, health and welfare Unemployment taxes and com- pensation (3.67 per cent) .... Social security Total labor cost per route per week $ 147.89 Labor cost per day (seven-day week) $ 21.13 * These wage rates do not apply to any specific market, but are based on levels appropriate to the current Los Angeles rates when calculated on the basis of the average commissions. units when studied in 1954) , most routes involve some commission payments. The type of wage payment, of course, influences the labor cost per minute. Under a straight wage payment, the per minute cost is constant and may be ob- tained simply by dividing the total daily labor cost by the minutes worked per day. On the other hand, where a commis- sion arrangement is involved, the total daily labor cost is a function of total route volume, and thus the cost per minute is variable with different load sizes. In view of the prevalence of the former system in most California markets, the following discussion assumes a flat wage- payment system. Table 16 presents the computations entering into the determi- nation of daily retail labor costs. Based on $21.13 per day, the total re- tail delivery labor cost per stop appear? in figure 19. It is again emphasized that this figure merely represents the basic physical time requirements shown in figure 18, converted to dollar costs by applying the cost per minute (4.4 cents) applicable to present labor costs. Adjust- ments can be made to account for dif- ferent labor costs in other markets, or for changes in wage levels, merely by applying the appropriate labor costs (in cents per minute) to the basic physical time requirements from figure 18. It should again be noted that the specific nature of this labor-cost relationship de- pends upon the type of wage agreement in effect. The costs shown in figure 19 are based on conditions where a flat wage is paid — that is, where the cost per minute is constant for any route volume. Where wages are based in part or in whole on commissions, the cost per minute increases as a function of route volume. This in turn is directly related to volume per stop, and the cost per minute thus increases with increases in delivery volume. For this reason, the slope of the labor cost-volume per stop relationship will be steeper under com- mission than under flat salary arrange- ments. Note should be made of the effect of a combination of a guaranteed base and commission, such as currently exists in Los Angeles, on the nature of this cost- volume relationship (fig. 20) . Until route volumes appropriate for the level of the base volume are reached (consistent with average deliveries of less than 3 units), the labor costs per day are constant, and the slope of the relation is comparable with that in figure 19. For route volumes in excess of the base volumes, the com- mission becomes effective, and the slope of the cost-volume relation increases. Therefore, in total, this curve is "kinked" in an upward direction as illustrated in figure 19. [56] Fig. 2 4 6 8 Volume per delivery (in labor units) 19. Effect of volume per delivery on delivery labor cost per stop, Los Angeles and Fresno markets combined, 1954— retail milk routes. 60 50 - 40 30 s 20 ^^^^ 01 23456789 Volume per delivery (in labor units) Fig. 20. Effect of volume per delivery on labor cost per delivery, Los Angeles, 1955- retail milk routes. 10 [57] Truck-Operating Expenses-Retail The Bureau of Milk Control studies were drawn upon again to determine the level of retail delivery truck-operating expenses. These studies were in sufficient detail to provide accounting-type data on such fixed cost elements as depreciation, licenses and taxes, and garage overhead, as well as such operating expenses as gas- oline, oil, tires, repairs, and other rele- vant items. The period for which these studies were made was the last quarter of 1953. An analysis of the records of eight plants operating a total of 915 retail routes in the Los Angeles and Fresno markets indicated an average daily truck- operating cost of $6.56. Although some variation in volumes per route existed, these differences did not involve substan- tial differences in type or size of truck, and no relationship between average route volumes and average truck-operat- ing costs was found. It is recognized that load sizes consistent with the larger aver- age delivery volumes — as high as 1,350 units, for example — would undoubtedly involve larger trucks than are currently used, with correspondingly higher oper- ating costs. In the absence of data on the operating costs of these larger trucks, it is arbitrarily assumed that the daily truck expenses for a route operating with double the present average volume will be higher than the present costs by 25 per cent. This is assumed for the follow- ing reasons: First, it is logical to expect that, while operating expenses increase with load sizes, they increase substan- c o 6 ' c >- Q) > -