Chapter Listing | Return to 1997 Topical Index


MEASURES

Dependent Variable. Based on the subsequent phone interviews conducted in 1991 and 1992 to verify status of the 2,653 respondents to the 1986 survey, we coded firms that were still in operation with a "1" and those that had disappeared as "0". Seventy-two percent of the 1986 new firms were found to still be operating in 1991. As we treated the firms as a dichotomy?survival or failure?logit regression analysis was deemed appropriate.

Independent Variables. Four independent variables were included in the study: the sales growth performance  of  the  firm,  location  of  the  new  firm, managerial problems of the new firm, and industry of the new firm.

Managerial Problems.  Owners of firms in the sample responded to eight types of managerial problems: (1) coping with government regulations, (2) developing good relations with unions, (3) finding qualified people, (4) selecting lawyers and accountants, (5) coordinating, motivating, and compensating personnel, (6) minimizing startup team conflicts, (7) finding qualified executives, and (8) establishing goals and strategic plans. Respondents were asked "for each type of problem, please indicate the severity of the problem at start-up". The response categories were "never occurred", "minor problem", and "major problem". They were then asked if currently the problem was "fully solved", "partially solved", or "not solved yet".
 
     The responses to these questions allowed us to construct three scenarios to managerial problems that may or may not have been encountered at the start-up of the firm. It also enabled us to determine if the problem was solved or still existed at the time the questionnaire was administered. Three scenarios were created based on responses to the questions: (1) did not encounter a problem at start-up, (2) encountered a problem at start-up, but have now solved the problem, and (3) encountered a problem at start-up, but the problem still exists. Distinctions were not made between problems that were minor or major, nor whether the problem was partially or fully solved. The structure of the questionnaire did not allow for the respondent to indicate that a problem did not exist at start-up, but existed at the time the questionnaire was administered. However, our interest is gauging problems at firm start-up and the ability to solve those problems as a precursor for understanding firm performance and survival.

Sales Growth Performance. The questionnaire asked each respondent to provide information on sales revenue for each year they were in business. Where the respondent did not have precise records, they were asked to provide an estimate. The collection of this information allows for the computation sales revenue growth by examining the increase or decrease from year to year.

    Sales performance was of the new firm was measured two ways: (1) sales revenue in the first year of the firm; and (2) percentage increase in sales from one year to the next. Sales revenue performance for the first year was classified as two types. High or low sales revenue performance. Firms were clustered by industry type to determine sales performance within each industry. Hence, retail firms were compared with retail firms and not with manufacturing firms in terms of high or low sales. Based on the distribution, those firms above the median in sales revenue were classified as having a "high start". Those below the median were classified as having a "low start" based on first year of sales. This was done for each of the five industries represented in the sample. We then measured the percentage increase in sales performance from one year to the next. Again, we used the median within each industry to identify those firms that are "high growth" and those that are "low growth" in percentage change in revenue sales performance. The average annual growth rate was used regardless if the firm was two years old or four years old. Firms that were one year old in the sample were excluded because of a lack of sales history. Firms that were over four years of age were excluded as well. The reasoning behind this latter exclusion was concerns that firms with averages longer than four years would be abnormally low because of their age. It was also felt that problem solving would have the most immediate impact on performance in the early stages of the new firms life cycle than in later, or older, stages.

    One additional problem concerning sales revenue performance was addressed: the effects of inflation. Some firms had their first sales in 1982 while others had their first sales in 1985. Inflation can be a serious problem in comparing sales performance of firms in different time periods. While a 3-4 year period may not seem to be great, the mid-1980s did experience high levels of inflation. For instance, the "real" value of the U.S. dollar declined by more than half over the 1977-1992 period (U.S. Bureau of the Census, 1994; pg. 487). Thus, younger firms would have an upward bias in the classification of "fast" or "slow" start than older firms. For this reason, dollar figures for sales or financial support were adjusted, using the Consumer Price Index.

    Four types of firms were created based on the forms of performance they recorded in sales revenue: (1) fast start, high growth; (2) slow start, high growth, (3) fast start, slow growth; and (4) slow start, slow growth. For instance, a firm that was classified as "fast start, high growth" would be one that had sales performance above the median in the first year based on its industry classification and was above the median in sales revenue growth over subsequent years. Firms were coded as "1" or "0" depending on which one of the four sales performance classifications matched.

Industry.  Industry was entered as a control variable. Firms were classified into industry categories that resulted from aggregating several SIC codes assigned by Dun and Bradstreet. Overall eight industry classification categories were identified.  For the present study firms from mining and agricultural services were dropped as were firms from health, education, and social services.

    Industry sectors were combined into five industry entities: (1) manufacturing, (2) distribution, (3) retail, (4) services, and (5) construction. Manufacturing, distribution, and retail are each an industry sector in the SIC codes. Services is a combination of SIC codes for consumer services and product services. These two industry sectors combine establishments engaged in utilities, finance, insurance, real estate, social, and consumer services. Each firm was assigned a code of "1" if they resided in the industry and a code of "0" if they did not reside in the industry.

    The sample size in the analysis was 391 firms.

Data Analysis. Logit regression analysis was conducted using SPSS-PC to predict the relationship between new firm survival and the measures of managerial problems, sales performance growth, and industry. The logit regression procedure is useful for locating explanations between variables when the variables are categorical. The logit model assesses variables on new firm survival to determine significance in the relationship. Step-wise techniques were used to determine how factors of variables classified by managerial problems, sales growth performance, and industry would contribute to the analysis. This step-wise procedure resulted in four models being developed. Each variable had to achieve a significance level of less than .10 to be able to enter the model. Since previous research has shown that survival of new ventures varies by industry, we include industry as a control variable. In the first model we entered industry. We then entered the eight managerial measures to test the problem solving effectiveness hypothesis (model 2). We then entered the measures associated with sales growth (model 3). Finally, we added multiplicative variables to test the two-way interactions: sales performance growth by managerial problems (model 4).

 To test the hypotheses we evaluated differences in the levels of the independent variables by using deviation contrasts.  This approach compares each level, or factor, to the pooled effect of the variables levels.  For example, if the pooled effect of sales growth performance is statistically significant as indicated by the Wald value, the effect of fast start/high growth can be evaluated in comparison to the average effect of all four sales performance factors (i.e., fast start/low growth, slow start/high growth, slow start/low growth).

Top of page | Chapter Listing | Return to 1997 Topical Index


1997 Babson College All Rights Reserved
Last Updated 03/30/98