Forecasting unemployment: a small business-survey model; Small business plans and expectations provide accurate and timely forecasts of national employment

Business Economics, Oct, 2004 by William C. Dunkelberg, Jonathan A. Scott, William J. Dennis, Jr.

(1) The Census data are available on the SBA website, www.sba.gov/advo.> (2) Note that small business is also credited with producing the bulk of net new jobs created in the U.S. economy. See www.sba.gov/advo/stats for the latest statistics on the contributions of small business to the American economy and Popkin (2003) for a description of methodologies used.

(3) Some argue that the owners of small firms with flatter organizations might sense changes in economic conditions more quickly than their counterparts in large firms, making indicators of the economic health of small firms relatively more responsive to changes in economic conditions.

(4) A description of the origin and content of NFIB's economic survey can be found in Dennis and Dunkelberg (2000).

(5) Early in the history of the surveys, NFIB experimented with industry weighted and employment weighted measures but found no significant improvement in the time series qualities of the measures.

(6) The survey question for HIREPLN data reads: "In the next three months, do you expect to increase or decrease the total number of people working for you?"

(7) The survey question for JOBOPEN data reads: "Do you have any job openings that you are not able to fill right now?" (8) The net percent of firms planning to increase total employment will be smaller than the number of firms hiring new employees, since many firms will be replacing workers--leaving total employment unchanged--or might hire workers but terminate even more workers--reducing total employment. In recent surveys, between 45 percent and 50 percent of the owners report looking for at least one employee each month.

(8) The net percent of firms planning to increase total employment will be smaller than the number of firms hiring new employees, since many firms will be replacing workers--leaving total employment unchanged--or might hire workers but terminate even more workers--reducing total employment. In recent surveys, between 45 percent and 50 percent of the owners report looking for at least one employee each month.

(9) The survey question for ERSALE data reads: "Overall, what do you expect to happen to the real volume (number of units) of goods and/or services that you will sell during the next three months?"

(10) A one-quarter lag of one quarter means that the survey in the first month of a quarter is used to predict the value of the dependent variable in the following quarter. Thus, the second quarter April value of the dependent variable would depend on the January level of the NFIB variables. The lapse of time between the NFIB forecast and the release of preliminary data by the Bureau of Labor Statistics could be as long as five or six months.

(11) While regression coefficients cannot be compared to determine which is the most important predictor in an equation, standardized regression coefficients can be compared. Standardized coefficients can be computed as follows: beta coefficient = (regression coefficient X standard deviation of X)/standard deviation of Y. If all variables in a regression are converted to standard normal variables, then the resulting coefficients are beta coefficients.