Regression estimates for accounting purposes;

              REGRESSION ESTIMATES FOR ACCOUNTING PURPOSES
by Maurice S. Newman Partner, Executive Office
Presented before the New York Chapter, American Statistical Association— November 1972
There are many places in financial accounting where estimates can be used and should be used in place of the tedious 100 percent counts or calculations that are the general rule today. Some of the more obvious possibilities are estimates of inventories, accounts receivable, cost of sales, unearned discounts and various kinds of reserves. Accountants need to be convinced that satisfactory estimates can be obtained for these purposes, and that very large amounts of time can be saved through the use of statistical estimates.
In many companies today, the detailed accounting records are maintained on a computer. This means that there is an existing record that could produce a list of the book values associated with all the components of the asset, liability, revenue, or expense account to be estimated. To be more specific, let us talk in terms of an inventory of 2,500 different parts, maintained on a perpetual inventory record, by computer. Associated with each part number would be a quantity on hand for that part and a unit cost. An extension of the quantity on hand times cost for each part number would provide the book value of the inventory.
The purpose of this talk is to show how a satisfactory estimate of the physical value of this inventory may be obtained and, by extension, how estimates of other such accounting records may be obtained with a minimum of time and effort.
FEATURE OF SAMPLING INTEREST
The question at issue is whether the perpetual inventory record is a reliable indicator of the exact condition of the physical inventory maintained in the stock room or warehouse. If we could be sure that there was an exact correspondence, the perpetual record would be the best estimate of the inventory value. In a few situations, such as inventories of precious metals, the control procedures may be sufficiently good that such an exact