IN order to forecast next year's income tax receipts it is necessary to determine the income of the current year upon which the tax is based. Since the latest data tabulated from income tax returns and published in Statistics of Income apply to income reported two years ago, the gap between recorded data and the present must be bridged by an estimate of last year's income. Analysis of the data published in Statistics of Income has resulted in a tentative method of forecasting the income that will be returned and published in the next two volumes of the Statistics of Income.' The various stages in the development of this method as applied to forecasting corporation income, both before and after the data in Table i (this REVIEW, vol. XI, pp. i8o-i) were derived, may profitably be described. At the outset, attempts were made to use independently such items as gross sales, miscellaneous income, cost of goods sold, and miscellaneous expenses. The ensuing analysis revealed many variations which could not be explained by economic changes, and which were traced and found to be due to changes in laws, methods of tabulation, and accounting practice. Therefore, items of expense or deductions, that might have been interchanged on account of any one or all of these causes and consequently show an inverse relationship, were combined into one group, and items displaying a similarity of movement were combined into another. The dominant characteristics of these resulting groups were those of fixed and variable (supplementary and prime) costs, and it appeared desirable to pursue the study according to such a classification although considerable overlapping existed due to the nature and number of separate items tabulated. A corresponding separation of income into fixed income and variable income cannot be satisfactorily obtained from the data. However, the separation of deductions into two such parts becomes a function of, or dependent on, the separation of income into corresponding parts, as well as on the weight or influence attributed to other unknown factors. Since complete separation of deductions into fixed and variable parts was not possible by direct use of the tabulated data, resort was had to trial and error experimentation to obtain an approximate division. It is fully appreciated that the results are tentative, that publishing detailed results without expressing all of the limitations and mental reservations which the authors have attached to the data and to the method is hazardous, and that as more data become available many of the present conclusions may be discarded. The method in its present stage of development consists of (i) forecasting gross income,2 (2) deducting therefrom an estimate of variable costs, (3) deducting similarly an estimate of fixed costs, and (4) obtaining a resultant estimate of net income. There follows a description of the analysis at two previous stages of development and an explanation of the more important intervening trials and errors. The first stage of development was reached before the data for I926 became available in I928. After discarding the separate items of income and deductions as rather meaningless, and recognizing the limitations of gross income data as published, but without attempting to make the necessary adjustments, an index of gross income was constructed and substituted in place of the tabulated gross income data. This index of gross income was constructed by combining several published index numbers and using constant *Continued from November, I929, issue. 1 Forecasts of prior year loss deductions from net income and of credit for size of net income also must be made in estimating net income subject to tax. Furthermore, the net income as tabulated in the Statistics of Income is not exactly the same as the income on which taxes are actually collected during a given period. For example, during the calendar year I927, some current taxes were collected on incomes tabulated in the I925, I926, and I927 Statistics of Income, owing to the method of tabulating fiscal year returns, and the legal provision for payment of taxes thereupon in quarterly installments, the first of which is due two and one-half months after the close of the fiscal year for which the return is made. 2 Gross income as adjusted and explained in the preceding article.

NYSE and Nasdaq trades increasingly cluster on multiples of 500, 1,000, and 5,000 shares. Such clustering varies over time and across stocks, and tends to increase with the level of trading activity. Furthermore, rounded trades tend to have more persistence both in occurrence and in trade initiation. Finally, medium-sized rounded trades tend to have greater relative price impact than large rounded trades. From these observations we surmise that trade-size clustering is consistent, at least in part, with the actions of stealth traders who tend to use medium-sized rounded transactions in an attempt to disguise their trades.