Lightcast's job counts are shown in Analyst as whole numbers, but behind the scenes, job counts (and many other metrics) are fractional. These fractions are produced by the hundreds of processes through which the raw data is run on its way to becoming final Lightcast data.
For example, Lightcast job counts are unsuppressed by means of bi-proportional algorithms. In other words, initial estimate values are inserted into cells where no data was provided by the source, and these initial estimates are then adjusted until all cells sum to disclosed totals. Often the adjustment process allows us to get very close to disclosed totals, but there might be several residual remainder jobs. In these cases, the marginal remainder jobs are evenly distributed across all categories, resulting in fractional job counts (or earnings or population counts). Additionally, since we average job counts across 12 months of QCEW data to create annual averages, fractional jobs are created even for disclosed data points.
To give another example, Lightcast occupation job counts are generated by taking final industry job counts and applying staffing patterns. Staffing patterns are percentage maps that show how the jobs in a particular industry are broken out into the occupations that staff the industry. This process nearly always results in fractional occupation job counts (e.g. breaking out 15,377 jobs by percentages will not result in whole numbers).
Many processes such as the one outlined above can and do result in fractional job counts, earnings, population counts, and other metrics. For ease of use, the Analyst and Developer interfaces round these figures to the nearest whole number. Users who wish to use the additional detail can export reports to CSV where possible to show precise values: