accounting objectives. For example, many people might agree that more timely, more complete, and more detailed information on time-use within the household is needed. But does this mean establishing huge longitudinal surveys with minute-by-minute electronic diaries covering every household member? Probably not. Something less ambitious and costly may do. But how much less ambitious? How much is needed over and above the time-use data that the Bureau of Labor Statistics (BLS) is already collecting?
Data requirements all crucially depend on assumptions embedded in the conceptual model. For example, conventional national accounting implicitly assumes that the value of government-supplied education equals its cost. For its nonmarket environmental accounting, the U.N. implicitly assumes that the value of environmental degradation equals the costs of restoring the environment to a pristine state. Without some theoretical underpinnings, it is not possible to determine whether these assumptions are "good enough."
The biggest factor affecting the need for new data is simply the availability of existing data. Even among accounts that use essentially the same theoretical framework, the data requirements (that is, the amount and type of additional data that must be assembled) may vary widely because of vastly different levels of current data and resources available for data collection.
The panel tentatively plans to determine data requirements in the following manner: (1) establish an accounting framework that meets useful nonmarket accounting objectives; (2) specify the general sorts of data consistent with realistic implementation of the framework; (3) survey existing data that may be available; (4) identify gaps and assess them as to their relative importance in precluding the accounting objectives, and (5) specify meaningful ways that the important gaps could be filled, either through new data collection or by developing suitable proxies based on theoretical considerations.
It is, at this time, appropriate to offer a few words about the BLS’s soon-to-be-released time use data. Consensus is fairly widespread that the single most important information required for nonmarket accounts is data on how the population spends its time. Like its market analogue, the most pervasive input in nonmarket production is often time—both market and nonmarket.2 This is particularly true in such areas as education, human capital development more generally, and home production. Data on market inputs, such as home and materials, exist, though the underlying production function indicating how these inputs are combined is not well understood; the big missing piece of data is for time, a not-always-marketed input.
The American Time Use Survey (ATUS), conducted by the BLS in 2003, will represent a huge step forward in assessing time devoted to household production and will make it possible to provide estimates of the value of household production similar to those currently provided by the Australian Bureau of Statistics and Statistics Canada. ATUS began in January, 2003, collecting time diaries every month from roughly 2,000 unrelated individuals who had completed their tenure in the Current Population Survey in the previous month. Each year this survey will generate the largest number of diaries of any time-budget study in the world. Most important, by putting time-budget surveys on a