The survey on which Problems and Priorities is based was conducted from February through mid-April of 2024 across a randomly drawn sample of 40,000 members of the National Federation of Independent Business (NFIB). Separate samples of 1,500 each were drawn for CA, OH, NY, and TX to create state specific rankings. Sampled small business owner members received a four-page mail questionnaire and up to two follow-ups. They provided 2,873 usable responses by the April cut-off date.
Appendix Table 1 provides a comparison between NFIB members and the overall small business population by employee size of business and industry, the two most important variables distinguishing respondents in the survey. Note that NFIB member respondents have marginally larger businesses than the population. But the distributions are reasonably similar and reflect the large skew toward the smallest firms. Also, NFIB member respondents contain 14 percent non-employers. The population sample contains non-employers for the week in which the data were collected. However, those nonemployers did have employees at some time during the preceding year, information not available for NFIB non-employers. Totals will therefore marginally reduce concern over employee-related problems and somewhat overstate concerns with certain regulatory issues.
The industry comparison between NFIB member respondents and the population is not as close, in part due to the detail of the NAICS codes. The major discrepancy is that respondents more frequently have businesses in traditional industries, such as manufacturing and construction, and less frequently in rapidly growing newer services industries. Agriculture represents the most pronounced difference because official statistics do not include farmers and ranchers in the population and NFIB does. The result of these differences is that the concerns of the services will be muted in totals, though far from unrepresented, while those from production will be louder than its population’s share. Still, significan’t numbers of respondents reported from every major industrial sector and industry differences are revealed in the industry break-out. Overall, the roughly 300,000 NFIB member firms are subject to the same policies and economic conditions as non-member firms. Consequently, it is likely that their responses accurately reflect the views of owners in the larger small business sector.
The sampling frame could lead to modest biases, but they are likely minimal. Certainly they will not alter the relative position of any problem by more than a rank or two. Problems that are of great concern would remain problems of great concern (even with a weighted result) while problems in the middle would remain in the middle and those at the bottom would remain at the bottom.
Despite being only four pages, the questionnaire could easily become tedious for the respondent. To avoid possible bias brought about by respondent fatigue, half of the sample received version A of the questionnaire and half received version B. The two versions are identical except that version B is inverted. The fi rst question on version A is the last question on version B, and so on. The purpose is to ensure that should fatigue set in, it would not affect one half of the questionnaire’s responses any more than the other. The data collected from version B was inverted prior to tabulation to produce a unified data set.
NFIB member respondents evaluated each of the 75 potential problems presented to them on a scale of “1” to “7.” The former represents a “Critical Problem.” The latter represents “Not a Problem.” The numbers between represent varying degrees of problem difficulty within the 1 – 7 extremes. An average for each problem was calculated and it served as the basis for ranking or rank-ordering problems. There are two associated issues. Non-response could be treated as non-interest, effectively relegating it to the lowest rating (“7”), or it could be treated as indecision or oversight, effectively giving the problem average score. The latter was selected because non-response seemed to generate no pattern across problems. The second issue is the rank of those problems with the identical average score. Ties led to the arbitrary decision to give precedence to those with a higher standard deviation.
