"The economy has changed and our data-collection efforts have not kept pace with it"
Zvi Gilische ("Productivity, R&D, and the Data Constraint", 1994)

Definition and Immeasurability

The simplest and perhaps most common definition of productivity is that of a "black-box": it is total output divided by total input. Despite being easy to define, however, it is notoriously difficult to measure. Output value cannot be determined solely from the quantity of "widgets coming out of a factory", but requires the consideration of various other intangibles, ranging from quality, timeliness, customization, and variety. Similarly, input can no longer be counted by labor hours. The amount of capital equipment used, material consumed, worker training and education, or even the "organizational capital" encompassed by the business relationships and investments into new practices must all be factored in. (Brynjolfsson, 1998).

New Innovation and Fixed Costs

Diewert's study found several instances where the measurements of inputs and outputs were blurred. The development of new products or technologies necessarily exacts a certain amount of fixed costs, including basic R&D, the development of new capital, the retraining of workers, etc. Should the growth of new products exceed that of sales, the resulting productivity measurement is affected adversely by the product's initial fixed cost, a bias that is eliminated over time as the product undergoes regular consumption. The theory of this claim is borne out through the realization that product diversity is increasing, thus "instead of demanding more from a fixed menu of available commodities, consumers are demanding smaller portions from an expanding menu of consumption possibilities", exacerbating the effects of fixed costs. Triplett, whose paper "The Solow productivity paradox: what do computers do to productivity?" also examines the inadequacy of productivity measurements, provides a counterargument, however, claiming that economists are overestimating the affect of new innovation, since new innovations are being counted on an arithmetic scale, when it should be logarithmic.

Consumption Disguised as Business Expenditure

"Business travel and entertainment expenses, along with company gyms, centers, cars and home loans are examples of former consumption expenditures which would not show up in final aggregate GDP." Diewert leaves this as a hypothesis. Surely the corporate world's version of pork barreling exists, but the extent to which it should it affect final productivity measurements is unclear. In any case, the ethics of the matter are something separate from data verifiability.

Mismeasured Outputs

Margins: Often, statistical agencies only collect data on total sales, whereas it would be more appropriate to take the margin in arbitrage transactions. This results in a positive bias when measuring productivity.
The Treatment of Interest: "National income accounting follows financial accounting in treating interest as a distribution from profits or operating surplus rather than as a cost of production."(Diewert)
The Treatment of Risk: Measures are quite inaccurate for industries whose profits are probabilistic (i.e. lacked a estimable real price), including insurance, gambling, and options trading.
Measuring Multifactor productivity in Service Sectors: The complexity of measuring all total contributing inputs for large, hierarchical service industries, (i.e. medical, sports, educational, telecom) makes them virtually immeasurable.