tion may be extremely high for such occupations. Such occupations should not be ignored, especially not those that are thought to be emerging or fast-growing occupations, but more directed efforts at data collection that are less costly should be considered. For example, it might be possible to sample from locations where the occupations are known to exist in some numbers and to use matrix sampling of occupational attributes (rather than collecting data on all attributes).

Making predictions about the future is hazardous in any field, and the future of occupations is no exception. However, the existence of an accurate, comprehensive, current occupational database, especially if it is linked to labor transactions as mentioned above, should provide an empirical base for making statistical, algorithmic projections as well as a solid footing for subjective estimates by occupational experts of all stripes.