CHAPTER 6 CONCLUSION This study has attempted to draw as accurate a picture of the beef industry as possible from existing data and from independent statistical analysis. Admittedly, the picture is incomplete. Very little is really known about the impact of "boxed beef"; about the effect of high market concentration, especially at the local level; or about the relative size of the margins of small and large members of the industry None of these relationships can be adequately analyzed until more accurate data is gathered. The FTC line-of-business reporting would be a good start, especially in providing information on some of the larger and more influential firms. Another possible reform might be the collection of more extensive price data-for instance, live prices from direct purchases rather than from terminal and auction markets. Such data, coupled with an intensive analysis of market structure at the wholesale and retail level, could help to unravel the discrepancies between the market model and the demonstrated price relationships in the industry. Although no significant anti-competitive effects are yet apparent in the industry, this should not be an excuse for complacency. Some disturbing trends are still apparent; most significantly the high fourfirm concentration figures in local or regional markets. Another trend worth monitoring is the shift toward "boxed beef", with its resulting effect on the structure of the industry. All of those factors should be components of a continuing watchdog function performed by the FTC, the SEC, the USDA, and Congress. CHAPTER 7 STATISTICAL PROBLEMS The rise of daily price and slaughter figures exacerbated a problem common to most time series analysis, serial correlation of data. Most methods of statistical analysis require the starting assumption that all observations are random. However, with time series data one finds that an observation is almost always influenced by the previous observation. For example, when a meat packer opens for business he will probably not roll dice to determine opening prices; he will most likely start the day's trading at the previous day's closing prices. To test for the presence of serial correlation, statisticians have developed the Durbin-Watson statistic. Numerous approaches to the analysis of serially correlated data have also been developed. Among these is the Cochrane-Orcutt iterative method which controls for the effect of one period's observations on the next period's observations though it does not control for the effects of one set of observations on those two or more periods later. The Cochrane-Orcutt technique was used for all analysis in this study. It mitigated the serial correlation problem sufficiently in all cases except the analysis of the relationship between carcass prices and slaughter figures for 1974. The fact that the Durbin-Watson statistic failed to reach the confidence level in this relationship may indicate that a substantial amount of two or more day serial correlation is present in the data. A. RESULTS AND CONCLUSIONS OF THE STUDY OF RELATIONSHIPS BETWEEN PRICE AND SLAUGHTER NOTE. Given the nature of the data, a slope that is "not statistically different from zero," is one that is not high or low enough, to state with any reliability whether or not a relationship is positive or negative. In statisticians' terms, such a slope is one where the statistic falls below the 0.5 confidence level. 1. RELATIONSHIP BETWEEN LIVE PRICES AND SLAUGHTER A. January 1970 through June 1971-daily Regression coefficients: Approximately 0.5. Interpretation: Though the regression coefficients are statistically significant, they are not large enough to make reliable conclusions about the relationship. B. January 1974 through December 1974-daily Regression coefficients: Approximately .66. Slope: Consistently negative. Interpretation. An inverse relationship exists between live prices and slaughter. Conclusions. For the 1970-71 period no conclusive relationship was found while for the 1974 period there was a clear cut inverse relationship. The slopes of same-day to five-day lag correlations for 1974 ranged from 64 to .90. Apparently the market was working in accordance with the normal market model during the 1974 period. The normal market model would lead one to expect an inverse relation.ship between raw material cost and production. The fact that the expected inverse correlation was not apparent in 1970-71 data is surprising. Some factor that was not accounted for must be responsible for this apparent contradiction of the market model. 2. RELATIONSHIP BETWEEN CARCASS PRICES AND SLAUGHTER A. January 1970 through June 1971-daily Regression coefficient: Approximately 0.75. Interpretation.-A large increase or decrease in carcass prices does not significantly change the number of cattle slaughtered and a large increase or decrease in the number of cattle slaughtered does not significantly change the carcass price. B. January 1974 through December 1974-daily Conclusion. For both time periods the slopes of the regression lines were not statistically different from zero. Both regression coefficients were high; the coefficient for the 1974 period was higher. These results indicate that on a day-to-day basis carcass prices were not a significant determinant of slaughter and slaughter was not a significant determinant of prices for either of the two periods studied. However, the reliability of this conclusion is questionable since the Durkin-Watson statistic for the 1974 carcass price correlations was approximately 1.2, significantly below confidence levels that indicate elimination of the serial correlation problem. B. RESULTS AND CONCLUSIONS OF THE STUDY OF THE RELATIONSHIP BETWEEN GRAIN PRICES AND LIVE CATTLE PRICES (SEE APPENDIX XIII) A. January 1970 through June 1971-weekly Regression coefficients for same week, one week lag, and two week lag: Approximately 0.92. Slope: Not statistically different from zero. Interpretation. For short term lags, a large increase or decrease in grain prices does not significantly change live cattle prices and a large increase or decrease in live cattle prices does not significantly change grain prices. B. January 1974 through June 1974—weekly Regression coefficient for same week and one week lag: approximately 0.8. Slope: Not statistically different from zero. |