To assist users in ascertaining the reliability of the Occupational Requirements Survey (ORS) estimates, standard errors are available with the estimates released through the public data query tools and complete dataset (XLSX). Standard errors provide users a tool to judge the quality of an estimate to ensure that it is within an acceptable range for their intended purpose. Collected and imputed data are included in the standard error calculation. BLS will continue refining estimation processes including evaluating the impact of sampling and nonsampling errors on the ORS estimates.
The ORS estimates are derived from a sample and thus, are subject to sampling errors. Sampling errors are differences that occur between the results computed from a sample of observations and those computed from all observations in a population. Estimates derived from different samples selected using the same sample design may differ from each other. In the case of the ORS, the population of an estimate is an occupation or occupational group within the civilian ownership, which includes private industry and state and local government workers.
The standard error is a measure of the variation among these differing estimates. It can be used to measure the precision with which an estimate from a particular sample approximates the expected result of all possible samples. Standard errors can be used to define a range or level of confidence (confidence interval) around an estimate. For instance, the 90 percent confidence level means that if all possible samples were selected and an estimate of a value and its sampling error were computed for each, then for approximately 90 percent of the samples, the intervals from 1.645 standard errors below the estimate to 1.645 standard errors above the estimate would include the "true" population figure. For example, the 90 percent confidence interval for a percentage estimate of 20.1 percent with a standard error of 0.8 percentage points would be 20.1 percent plus or minus 1.3 percentage points (1.645 standard errors multiplied by 0.8 percentage points) or 18.8 to 21.4 percent;
The chances are about 68 out of 100 percent that an estimate from the survey differs from the true population figure within one standard error. The chances are about 90 out of 100 percent that this difference would be within 1.645 standard errors. This means that in the example above, the chances are 90 out of 100 percent that the estimate is between 18.8 and 21.4 percent.
The ORS uses balanced repeated replication (BRR) to estimate the standard errors as explained in the Handbook of Methods. In some instances, the standard error is unavailable as the rounded value is less than 0.05 (footnote 5) or 0.5 (footnote 6) depending on the level of calculated precision. When categorical estimates (for example the requirement is present or not present) are equal to 0 or 100 percent, the standard error is not available and denoted by footnote 15.
The same calculation can be used for each estimate type, see examples below.
20.1 percent with a standard error of 0.8
20.1 +/- (1.645 x 0.8) = 20.1 +/- 1.3 = [18.8, 21.4] percent
Percent of day:
61.0 percent of day with a standard error of 1.5
61.0 +/- (1.645 x 1.5) = 61.0 +/- 2.5 = [58.5, 63.5] percent of day
43.33 lbs. with a standard error of 2.5
43.33 lbs. +/- (1.645 x 2.5) = 43.33 +/- 4.1 = [39.22, 47.44] pounds
2.15 hours with a standard error of 0.44
2.15 hours +/- (1.645 x 0.44) = 2.15 +/- 0.72 = [1.43, 2.87] hours
1,095.00 days with a standard error of 100
1,095.00 days +/- (1.645 x 100) = 1,095.00 +/- 164.5 = [930.50, 1,259.50] days