Because not all employers responded to our requests for information in the survey, we statistically inferred some employment data to replace that which otherwise would be missing. This allows us to estimate the total amount and type of employment, which provides the basis for other estimates of economic impact.
In general, InterVISTAS’ approach bases these inferred estimates on information provided by responding firms for each business type and validated against information from other publicly available sources of data. This approach is conservative in that we assumed that the non-responding firms are smaller than responding firms.
The employment data in this report was compiled from a combination of two sources:
Employment reported by employers on surveys submitted to InterVISTAS.
Employment inferred for employers who did not provide a survey response. Inferred employment was based on employment information from those firms in each business type that did respond to the survey. The mean employment of respondents in each business type was calculated, excluding outliers, and then conservatively adjusted downwards. For instance, those firms with especially large employment levels were excluded from the "mean without outliers" to obtain conservative results. This "adjusted mean" employment for each business type was then applied to those firms who did not respond to the survey.
- 1 Our approach is broadly consistent with the donor imputation model typically used in survey work by government census and statistical agencies in Canada, the US and the European Commission.