Time and Toyota have taught us a lot about how to run a successful automobile company. For years prior to the mid-1970s, Peter Drucker, Alfred Chandler, and the many writers influenced by them had heaped praise on the General Motors Corporation for its choice of corporate structure and for its managerial techniques. This structure and these techniques were said to have been responsible for GM's success. GM's weak performance in recent years raises the suspicion that either they aren't managing as well as they used to or else managerial technique never was the key to their success. In fact, GM's drive to dominance in the automobile industry during the 1920s cannot easily be attributed to its managerial techniques; newly discovered material in the Ford archives reveals that during these years the Ford Motor Company was in most ways better managed than GM. GM became the leading car company in the late 1920s, not because it was better managed than Ford, but because it built better cars.
Ford has good claim to being the first modern manufacturing corporation. This is true in two respects. First, Ford pioneered in establishing a multidivisional, decentralized system of assembling, distributing, and selling automobiles. In a very influential discussion, Chandler has identified the introduction of this type of corporate "structure" at E.I. du Pont de Nemours & Co. and General Motors with the "making of the modern corporation." Yet Ford had begun implementing a decentralized structure as early as 1909, well before either du Pont or GM. Second, Ford also pioneered in developing means of controlling inventory, receiving short-run feedback from dealers, forecasting sales, and keeping production scheduling in line with sales. Once again, Chandler credits to GM innovations that were actually Ford's.
The account of Ford's early history presented here is drastically different from that to be found in Chandler's work or anywhere else in the literature. The following comparison of the performances of Ford and GM in the years up through 1939 reveals the points at which the traditional view has gone astray.
During the 1910-11 downturn several prominent automobile companies suffered severe reversals. The United States Motor Company, which had been producing twenty-eight models under seven names, was placed in the hands of receivers in September 1911. At that point it had in inventory close to the extraordinary total of 100,000 completed automobiles (total industry production for 1911 amounted to only 210,000 automobiles) and what was only described in the petition to receivers as "a large amount of supplies and materials." During 1910 General Motors was reorganized, with its founder, William C. Durant, temporarily forced out of active control of the company. The problem automobile companies faced during this time was an inability to control inventories. Two industry practices made the companies and their dealers prone to inventory problems. Ordinarily, dealers were required to pay for automobiles on delivery, while producers had 30 to 90 days to pay suppliers. In addition, dealers were often obliged to accept shipments they would rather not have taken or risk losing their franchises. In combination, these practices made producers less sensitive to fluctuations in final demand than they would otherwise have been.
The conventional view is that Ford weathered the problems of 1910- 11 without serious difficulty simply because the Model T was already proving to be a very popular car and Ford experienced no declines in sales during 1910. In fact, newly discovered monthly sales figures indicate that Ford did in fact suffer a significant decline in sales in the summer of 1910. While sales in the first six months of 1910 had run well above the levels of 1909, sales for July, August, and September were 47 percent below the year before. Ford avoided inventory problems because it was unique among automobile companies in having an efficient system that gathered detailed data from dealers on their sales, their orders on hand, their orders filled, and so forth. The surviving records show that this information was gathered at least as frequently as each month, beginning at least as early as October 1908. Later records indicate that Ford had begun collecting this information at ten-day intervals by 1920, whether it was collecting the information at ten-day intervals as early as 1908 is not known.
This detailed information was put to good use. For the period from October 1909 to June 1914 a simple regression of deviations of production from trend on deviations of sales from trend and on deviations of unfilled orders from trend, lagged one period, yields the following results (omitting the constant term and with t-statistics in parenthesis):
Production = 1.06Sales + 0.12Unfilled Orders-1
(13.46) (4.16)
Adjusted R2 = .86 F = 169.10 D-W = 1.95
These results indicate that at Ford production of Model T's was tied
closely to sales, which allowed the company to avoid the inventory
problems that sank less well managed companies.
Ford's pioneering use of the moving assembly line opened up the possibly for greatly reducing inventory holdings. The reliance on firm production cycles and the precise timing of production that the moving assembly made possible, meant that there could be greater coordination in the delivery to the plant of raw materials and purchased components. Although the moving assembly line had been fully developed at Highland Park by the end of 1914, the possibility of setting up tightly timed schedules to reduce inventory holdings were not fully exploited until Ernest Kanzler was placed in charge of scheduling during 1920. Kanzler success was such that average inventory holdings at Highland Park appear to have dropped by about two-thirds.
Henry Ford is quoted in an interview with the Detroit News published in July 1921 as having described the new system as follows:
Formerly we bought in vast bulk lots, using up stock as we needed it. But that would not do under our changed conditions. We have worked out a new system which, I believe, is not duplicated anywhere. There are 8,000 parts to the Ford car. Each one of those parts is given a number-symbol. Once each month we take a schedule of the exact number of cars we will make the next month. Then we figure out the exact amount of stock needed to make just the number of parts to fill that schedule and buy that amount of stock and no more.
Perhaps the most remarkable aspect of these procedures is that not only did they precede those for which General Motors is wrongly given credit for pioneering, but they also anticipated by decades the celebrated "just- in-time" inventory procedures associated with Toyota.
The key to the reforms were new production and inventory controls. Du Pont and Sloan were particularly concerned that the division managers had been allowed to purchase whatever materials they felt were necessary without higher authorization, and that excess production was often forced on reluctant dealers. The first step was to rein in the division managers by stripping them of their authority to purchase materials on their own accounts. Each general manager also was required to prepare a written estimate of the sales expected by the division over the following four months and of the materials and labor necessary for the indicated production. When these estimates had been agreed to by the general office, authority would be given for the purchase of one months' worth of materials.
Ford's performance during the 1920-21 downturn makes up one of the more colorful episodes in existing accounts of the company's history. The usual story is that Ford bulled its way through by forcing price reductions on its suppliers and forcing finished automobiles on its dealers. In fact, the key to Ford's ability to pull through this period was its reliance on the inventory and production controls which it had by this time had in place for more than a decade.
Table 1 provides the crucial statistics on production, sales, and changes in dealer holdings of inventories during the key months from September 1920 through September 1921. The first column gives production of motors at the Highland Park plant. The second column gives assembly of finished automobiles at Ford's thirty-four domestic branch assembly plants. The third column gives sales of automobiles to dealers and the fourth column gives retail sales of automobiles by Ford dealers. Production and sales clearly moved in close tandem. This dispels the myth that Ford bulled ahead with production, disregarding sales. As in 1910, the feedback from dealer operations guided production scheduling.
The table also provides for the first time reliable figures on the famous episode in which production of motors at the Highland Park plant slowed in January and then stopped altogether in February. The result was that while only 38,012 motors were produced during these two months, 65,188 automobiles were assembled. The difference between these two numbers represents automobiles that were assembled from motors that had previously been produced and other parts that had previously been produced or purchased. In legend, many of these automobiles were than shipped to unwilling dealers who were forced to accept them for fear of having their franchises revoked and who then bore the financial burden of financing this presumably unsold inventory.
Dealers were required during January 1921 to accept the full quota of cars for which they had previously contracted. Undoubtedly, many dealers would rather not have accepted these automobiles. However, for Ford dealers as a whole these shipments would not appear to have been too burdensome. The last column of the table indicates that dealer holdings fell in each of the first four months of 1921. The total decline was 80,315 automobiles.
Not surprisingly, previous accounts of this episode have had difficulty keeping the relevant numbers straight Obtaining reliable monthly statistics on the number of automobiles produced and sold by Ford during this period has been difficult. In addition, it is easy to err in interpreting the numbers. As Table 1 reveals there are four different sets of numbers to keep straight: production of motors and other components at the Highland Park plant, assembly of finished automobiles at the thirty- four Ford assembly plants spread across the country, sales of finished automobiles to dealers, and sales by dealers to the public. Even within the records of the Ford company these fours sets of numbers were not always consistently labeled. "Sales" in these documents generally refers to sales by dealers to the public, but occasionally sales to dealers from the assembly plants is meant. Similarly, "production" sometimes corresponds to assembly of motors at Highland Park and sometimes to assembly of finished automobiles at the assembly plants. Only through cross-checking and comparing different sources is it possible to sort out and properly label the statistics.
Another reason why accounts of Ford's performance during this period have been faulty is that the lack of statistics has resulted in a reliance on reminiscences of Ford employees. Nevins and Hill, for instance, make great use of the reminiscences of William C. Klann, who was the assistant superintendent of production at Highland Park, E.G. Liebold, Henry Ford's personal secretary, and others who were struck by how unusual it had been to shut down production of new components, lay off most of the work force, and assemble existing parts into finished automobiles to be shipped immediately to dealers. However, these observers were not generally privy to the statistics on the relationship between sales by dealers on the one hand, and production at the assembly plants and sales to dealers on the other hand. When these statistics are examined, it appears that however unusual the procedures employed during the winter of 1921 might have been, they do not seem to have resulted in extreme hardship for Ford dealers.
Donaldson Brown, at that time head of GM's financial staff, later analyzed the problem: "We did not know the rate [at] which our product was actually passing into the hands of the ultimate consumers, nor did we know what the stocks were in the hands of our dealers...." Note that Ford had been collecting this information since at least 1908.
GM dealer stocks of new automobiles, which had stood at about 77,000 on April 30, 1923, had ballooned to about 181,000 on April 30, 1924. The situation at Chevrolet was particularly bad. On April 1, Chevrolet dealers had 119,000 cars on their lots while a further 32,000 completed cars were on hand at the factory. These total are extraordinary given that Chevrolet's sales for the entire year were only 295,456 automobiles.
Brown's solution was to organize a system whereby "each car division receives from its dealers every 10 days a report giving the number of new orders taken, total orders on hand, and the number of new and used cars on hand." This system was identical to the one that had long been in place at Ford.
What accounts for this exceptional record [of paying dividends throughout the 1930s] in a period when many durable-goods producers failed or came close to bankruptcy? ... I think that the story I have told shows that we had simply learned how to react quickly. This was perhaps the greatest payoff of our system of financial and operating controls.
In histories of Ford, there is usually a strong emphasis put on the very able men who had left the company by the early 1920s. The most prominent of these were: C. Harold Wills, who had played a key role in designing the Model T and in developing the moving assembly line; James Couzens, who among other things had set up the dealership network and the system of branch assembly plants; Ernest Kanzler, who had been responsible for the efficient production scheduling system; and William Knudson, whose production skills were evident from his successes at Chevrolet after leaving Ford. Although these were certainly talented men, it is inaccurate to pin the difficulties Ford suffered during the Great Depression on the supposed administrative shortcomings of their successors.
Newly discovered data on Ford dealer holdings of finished automobiles allows a comparison of the relative successes of production scheduling and inventory control at Ford and General Motors. For the period from January 1929 to July 1939 GM dealers held in inventory 2.0 automobiles for every one sold, while Ford dealers held 1.0 automobiles for every one sold. The difference is statistically significant (t = 10.396). In addition, inventories held by GM's dealers were much more volatile. A fair inference would seem to be that during this period Ford did a better job of scheduling production, and hence avoided the degree of volatility in dealer holdings of automobiles experienced by GM.
Data are available on a yearly basis from 1923 to 1938 on total dollar inventories and on total production of passenger cars, commercial cars, and trucks for both Ford and GM. These show that in 1923, just prior to the second phase of GM's inventory reforms, GM's ratio of inventories to production was more than 300 percent higher than Ford's. GM's performance improved markedly over the next several years, so that by 1927, as production at Ford was shut down to allow for the changeover from the Model T to the Model A, GM's inventory-to-production ratio was about 15 percent lower than Ford's. However, during 1929 GM's inventory-to-production ratio rose to exceed Ford's by more than 60 percent and the ratios did not approach parity again until 1932. This result can be made a bit more formal by regressing the ratio of GM's inventory-to-production ratio on Ford's inventory-to-production ratio on a constant, a quadratic time trend, and a dummy variable for the Depression (defined as 1929-1938). The results are:
(GM Inventory/GM Production)/(Ford Inventory/Ford Production) =
4.93C - 0.89Time + 0.04Time2 + 1.33Dummy
(10.17) (5.66) (5.35) (2.50)
adjusted R2 = .73 D-W stat = 1.51 N = 16
Since the mean of the dependent variable is 1.91, the size of the estimated
coefficient on the dummy indicates that GM's performance in controlling
inventories during the Great Depression was markedly worse than Ford's.
Having only a single model to sell made Ford much more vulnerable than GM to a recession. Between 1929 and 1931 sales of all GM models declined by 46 percent, while sales of Ford Model A's declined by 64 percent. This larger sales decline is the explanation for Ford's much poorer financial performance.
With hindsight Ford's decision to concentrate on producing a single model seems very ill-advised. Certainly a stroll through the refrigerator and washing machine section of an appliance store or the television and VCR section of a department store reinforces just how commonplace the GM strategy of a diversified product line with different models of varying quality and price has become in the consumer durable market. But the world was very different in the 1920s. Nothing like the automobile had appeared before. It was certainly conceivable that the market would remain with the high end divided up among several companies each producing a relatively few cars for the wealthy and the low end dominated by whoever could best reap the scale economies to be found in concentrating on a single model. In fact, Alfred Sloan reveals in his autobiography that GM made the fateful decision to reorganize their product line into a hierarchy of models largely because there was little faith that the Chevrolet would ever be able seriously to challenge the Model T. In fact, there were many at GM, apparently including Pierre du Pont, who felt that the company's only hope of competing successfully with Ford was to pour massive resources into the development of a new Chevrolet that would be better and cheaper than the Model T. Sloan's reflections on the new GM policy represents the situation in the automobile market in the early 1920s as it appeared to contemporaries:
In the perspective of so many years gone by, the idea of this policy seems pretty simple, like a shoe manufacturer proposing to sell shoes in more than one size. But it certainly did not seem simple at the time, when Ford had more than half the market.... For all we knew then, our policy might not have worked best. If the industry had thought it would work, the others would have adopted it at the time.
Sloan should receive full credit for devising what was ultimately a winning strategy, but Ford's failure to adopt a similar strategy in the early 1920s was not as foolish as is usually made out.
Change in Motors
Autos Sales to Retail Dealer Holdings of
Month Produced Assembled Dealers Sales Finished
Autosa
Sept. 1920 96,485 88,692 85,816 77,456 8,360
Oct. 1920 100,155 93,827 96,901 82,084 14,817
Nov. 1920 91,386 85,324 84,552 66,338 18,214
Dec. 1920 80,492 67,731 71,823 41,631 30,192
Jan. 1921 38,012 29,883 31,669 57,208 -25,539
Feb. 1921 0 35,305 36,575 63,603 -27,028
Mar. 1921 74,183 61,896 61,330 87,221 -25,891
Apr. 1921 96,891 91,358 90,295 92,152 - 1,857
May 1921 100,500 101,476 101,488 95,776
5,712
June 1921 106,528 107,132 108,141 102,871 5,270
July 1921 108,602 107,149 105,720 91,286
14,434
Aug. 1921 104,410 109,187 109,739 80,190
29,549
Sept. 1921 120,271 90,189 90,010 96,717 -
6,707
a The numbers in this column are sales of automobiles to dealers minus
retail sales of automobiles by dealers.
Source: Motors produced calculated from a list of the motor numbers of
the first motor assembled each month in Accession 33, Box 41 in the Ford
Archives. Sales to dealers are from Accession 940, Box 25. Automobiles
assembled and retail sales from Accession 572, Box 13.