METHODS

Sample

Data were taken from the New Business in America data-base, which reports upon data from surveys administered to 2994 entrepreneurs tracked over a three-year period. The surveys were modified each period, to adapt to the information available to entrepreneurs at each point in time. The initial sample covered firms which had been established for 17 months or less, with an average age of 11 months. The sample is broadly representative of new businesses in America, representing all industries and all geographical areas. Due to the information of interest, the present analysis is based on a subsample drawn from the year 2 questionnaire, with performance assessed in the follow-up questionnaire a year later. The total sample size for the second wave of the NBA data-base was reduced to 1190 entrepreneurs, mostly due to nonresponse. After selecting for ventures with at least one full-time employee in addition to the founder, information on the strategy and human resource management items of interest was available for a subsample of 857 firms. This initial sample of entrepreneurs, then, represented businesses that had been active for less than two and a half years, in a variety of industries, and with an average of around 6 full-time employees, besides the owner.

 

Measures

Performance was measured with both an objective measure and a perceptual measure relative to a priori expectations (Ramanujam, Venkatraman, and Camillus, 1986). Cash taken out of the business by the entrepreneur (salary, dividents, draw, etc.), during the 12 months preceding the administration of the second year questionnaire, was used as an indicator of the financial performance of the start-ups (TAKEOUT). Previous published studies using this measure support the use of this variable as a good proxy for performance. A second, subjective measure of performance was constructed by comparing the entrepreneurs satisfaction at the beginning and at the end of the period under study. In each period, satisfaction is a complex measure combining the scores of three items, which capture, respectively, the entrepreneur's satisfaction with sales, profits, and overall, with respect to a-priory expectations. All three items are 3-point ordinal scales ('less than expected' to 'more than expected'). Principal components factor analysis of these three items on the year 2 questionnaire confirmed the unidimensionality of the scale. After varimax rotation, the three indicators loaded strongly on the same factor, which explained 71.8 % of the variance. Further, reliability analysis performed on this scale confirmed again the contributions of the three items as indicators of an overall construct of performance. Their standardized Cronbach- coefficient was 0.80. Similar results were found as a result of the analysis of this measure for the year 3 data, which shows a standardized Cronbach- coefficient of 0.82. The final subjective measure of performance employed in this study is the difference between both periods measures of satisfaction, or the increase (or decrease) in the entrepreneur's level of satisfaction (INCPERF).

Strategy was measured through a battery of nine indicators, asking the entrepreneur's assessment of the percentage importance of each indicator as part of his business strategy. The approach to the management of human resources (HRM) by the entrepreneur was measured through a battery of seven 5-point Likert type indicators, asking the entrepreneur's assessment of the importance of each factor for the motivation of employees. Finally, a measure of size of the venture was also used as a control variable in the final part of the analysis. The proxy for size employed is the number of full-time-equivalent employees, computed as the number of full-time employees, plus the number of half-time employees divided by two.

Data Analysis

HRM and Strategy indicators were both factor analyzed using principal components with varimax rotation, and factor scores were computed. Next, entrepreneurs were cluster analyzed along the resultant factor scores by means of a two-stage clustering procedure, as suggested by Milligan (1980). As a result, two different classifications of entrepreneurs were generated: one according to their HRM approaches, and one according to their strategic orientations. Following, analysis of variance and analysis of covariance were used to investigate the existence of any significant performance differences across strategy types or across HRM types, as well as the existence of any significant interaction effects between HRM types and strategy types on both measures of performance. Finally, several pairwise comparisons of least-squares means between selected HRM treatment effects, within given strategy typologies, were conducted to test hypotheses relative to matching effects.

 

RESULTS FROM THE EXPLORATORY ANALYSIS

Typology of Entrepreneurs' Approaches to Human Resource Management

The underlying dimensions characterizing the entrepreneur's approaches to the management of human resources were examined through the application of principal components factor analysis to a battery of 7 items asking for the entrepreneur's opinion about their motivational value. A three factor solution was chosen based on the Kaiser-Guttman rule, and the solution was rotated using the varimax rotation scheme. Table 1 presents the results obtained from this analysis.

Factor 1 is associated with fixed compensation and close supervision items, which indicates that this dimension is capturing the entrepreneur's relative disposition for behavior-type control. Factor 2 is associated with decentralization of decision making, intrinsic motivation of the job, and pay-for-performance, which indicates that this dimension is capturing the entrepreneur's disposition for outcome-type control. Finally, factor 3 is related to item 1 which captures the extent to which good personal relationships with employees are judged to be an important motivation factor. Although, this item could be considered as an indicator of "clan" or socialization-type of control (Ouchi, 1979), the lack of other items loading also on this factor prevents interpretability beyond that single item. Factor 3 was not used in the subsequent analyses.

Following factor analysis, behavior and outcome control factor scores were computed for all observations in the data set, which were then cluster analyzed on the basis of such scores. The two-stage clustering procedure suggested by Milligan (1980) was employed, where first observations are cluster analyzed using a hierarchical average linkage procedure. As a result of this analysis, the number of clusters underlying these data and their approximate centroids are determined. Next, these centroids are taken as initial seeds for a second non-hierarchical (i.e., K-means) cluster analysis on the same data set. These two consecutive cluster analyses were performed using the CLUSTER and FASTCLUS procedures, respectively, on the SAS statistical package.

TABLE 1
Factor Analysis of Human Resource Practices Items

(N=857)

  'Behavior Control'
Factor 1
'Outcome Control'
Factor 2
'Interpersonal'
Factor 3
1. Develop personal,
friendly relationships
    .896
2. Good wages; raises .792    
3. Extensive fringe benefits .778    
4. Share in decision-making;
more responsibility
  .787  
5. Close supervision .435    
6. Profit-sharing; stock options, etc.   .518  
7. Provide chance to learn new skills   .780  
       
Proportion of Variance Explained: 27.9 17.9 15.2
Cumulative Variance Explained: 27.9 45.8 60.9
* Factor loadings lower than 0.40 are not shown.      

Although frequently reported, the use of analysis-of-variance F tests on the variables used to generate the clusters, is not a valid method to evaluate the adequacy of the number-of-clusters solution finally chosen. Since clustering methods maximize the separation between clusters, it makes little sense to test differences between clusters against the null hypothesis that subjects are assigned randomly to clusters. Given the lack of statistical power of ordinary significance tests, more conservative criteria about the baseline cluster distribution function to be assumed are needed for an appropriate test of the number of clusters underlying the data (Klastorin, 1983). Several criteria based on unimodal or multimodal baseline distributions have been developed by different authors. A simulation study by Milligan and Cooper (1985), comparing thirty different methods for estimating the number of population clusters, showed that the Cubic Clustering Criterion (CCC), the Pseudo-F, and the Pseudo-t2 criteria tend to be more desirable. Accordingly, in the present study, consensus across these three statistics was sought for selecting the number of clusters solution during the non-hierarchical cluster analysis (SAS Institute Inc., 1989). Also, given the important distorting effects of outliers on the clustering procedure (Milligan, 1980), multivariate outliers were selected out of the final average linkage analysis. Multivariate outliers were identified as clusters with less than three members on the initial number-of-clusters solution.

As a result of the first average linkage cluster analysis on the overall sample of 857 entrepreneurs, 5 multivariate HRM outliers were identified and deleted (7 additional observations were judged to be multivariate outliers with respect to the entrepreneurs' strategic approaches, and were also deleted in this step). A consecutive average linkage analysis on the reduced sample produced excellent agreement between the CCC, the pseudo-F, and the pseudo-t2 statistics for a six-cluster solution. Centroids for that average linkage solution were computed and used as initial seeds for the consecutive k-means cluster procedure. Table 2 reports the final results obtained from the non-hierarchical analysis.

TABLE 2

Typology of Entrepreneurs According to their Approaches to Human Resource Management (N=845)

      Mean (standard deviation)
Cluster Number of
entrepreneurs
Sample Proportion 'Behavior
Control'
'Outcome
Control'
1 238 28.2 % 1.022
(0.49)
0.543
(0.60)
2 274 32.4 % -0.490
(0.42)
0.614
(0.53)
3 43 5.1 % - 1.013
(0.57)
-1.812
(0.53)
4 164 19.4 % 0.174
(0.53)
-0.800
(0.40)
5 83 9.8 % - 1.654
(0.48)
0.040
(0.65)
6 43 5.1 % 1.104
(0.54)
-1.890
(0.52)
         
R-Squared:     0.766 0.703
CCC: -17.165      
Pseudo-F: 465.17      

 

Cluster 1, which includes 28 % of the entrepreneurs in the sample, is characterized by a high propensity to use both behavior and outcome forms of control (H/H). Cluster 2, which includes 32 % of the sample, shows a moderately low propensity to use behavior control and a high propensity to use outcome control (the highest in the sample) (L/H). This group can be characterized as the outcome control group, relative to the other entrepreneurs in the sample. Cluster 3 is composed of 5 % of the entrepreneurs, which show a low propensity to use any form of motivational practice (L/L). Cluster 4 includes 19 % of the entrepreneurs in the sample, and it is characterized by an average propensity to use behavior control and a low propensity to use outcome control (--/L). Cluster 5, composed of 10 % of the entrepreneurs, shows a very low propensity to use behavior control, but only an average propensity to use behavior control (L/--). Finally, the 5 % of entrepreneurs included in cluster 6 are characterized by a very high propensity to use behavior control (the highest in the sample) and a very low propensity to use outcome control (the lowest in the sample) (H/L). Accordingly, these entrepreneurs can be characterized as the behavior control group, relative to other entrepreneurs in this sample. Figure 3 shows the plot of cluster memberships in the 2-dimensional space created by the behavior and outcome control factor scores. It is important to note the relative absence of entrepreneurs showing a high propensity to use outcome forms of control in this sample. This result seems to support the observation of the entrepreneur's relative need for tight control of operations and for a "hands-on" leadership style (Smith & Gannon, 1987).

FIGURE 3

Plot of HRM Clusters

Typology of Strategies

Principal components factor analysis on the battery of 9 competitive strategy items revealed 5 factors with eigenvalues higher than 1. Table 3 presents the results of the factor analysis, after varimax rotation of the five-factor solution.

Four items are loading on Factor 1. Emphasis on advertising more effectively, on having superior locations, and on better appearance and facilities load directly onto this factor, while the venture's emphasis on providing better service loads inversely onto it. Accordingly, this emerging strategic dimension captures the extent to which entrepreneurs are attempting to create value through enhancing the image or the perception of the customers relative to the value of the entrepreneur's products and services. Factor 2 is inversely associated with the entrepreneur's emphasis on providing better service and directly associated with his emphasis on delivering products or services that are unique in his marketplace. This dimension seems to reflect upon the creation of value through the performance of the products or services themselves (rather than through the perception of value of the customers). For products/services similar to competitors' offerings, their relative value to the customer may be enhanced through improving the level of service provided. On the other side of the spectrum, innovative entrepreneurs with relatively unmatched products/services will base their value creation strategies on the uniqueness of their products and will have less of an incentive to better their levels of service. Accordingly, this dimension is believed to capture the extent to which the entrepreneur's strategy is based on relative product/service uniqueness. Factor 3 captures a strategic dimension that extends from an emphasis on reputation for quality at the lower end, to an emphasis on offering more product/service choices at the upper end. Thus, this factor illustrates a trade-off for creating value through either higher quality or higher quantity of offerings, which will be labeled the product breadth dimension to value creation. Finally, factors 4 and 5 are associated with single items, so that their interpretation is straightforward. Factor 4 captures the extent to which the entrepreneur is not placing an emphasis on competing on price. Hence, this dimension is the inverse of the extent to which the entrepreneur is pursuing a low-cost strategy, based on the creation of value through the offering of comparable products at lower prices. Factor 5 is directly associated with the emphasis of the entrepreneur on creating value through targeting poorly served customer groups. In the strategy literature, this avenue for value creation is termed the focus dimension of competitive strategy (Porter, 1980). In sum, the 9 original strategy items were reduced to five different fundamental ways in which the sampled entrepreneurs attempted to create a competitive advantage for their offerings: better image, unique products, product breadth trade-offs, lower prices, and focusing on poorly served market niches.

TABLE 3

Factor Analysis of Strategy Items
(N=857)

  'Image'
Factor1
'Unique
Product'
Factor 2
'Product
Breadth'
Factor3
'Low Cost'
(inverted)
Factor 4
'Focus'
Factor 5
1. Keep prices lower       -.978  
2. Provide better service -.447 -.799      
3. Provide more product choices     .746    
4. Build a better reputation
for quality
    -. 832    
5. Advertise more effectively .473        
6. Target customers poorly
served by competitors
        .908
7. Superior location .667        
8. Better appearance/facilities .685        
9. Product/service
otherwise unavailable
  .768      
           
Proportion of Variance Explained: 18.4 14.7 13.5 12.6 11.5
Cumulative Variance Explained: 18.4 33.1 46.6 59.2 70.7
* Factor loadings lower than 0.40 are not shown.        

Following the factor analysis, entrepreneurs were cluster analyzed according to their factor scores for the five strategic dimensions described above. The same procedure followed for the analysis of HRM approaches was used here to distill a typology of strategic orientations of the entrepreneurs in the sample. As a result of the first average linkage cluster analysis, 7 strategic orientation outliers were detected and selected out. Hence, along with the 5 outliers detected during the cluster analysis based on HRM approaches, a total of 12 multivariate outliers were deleted, reducing the sample to 845 entrepreneurs. A second average linkage analysis on this reduced sample produced agreement between the three number-of-cluster criteria for a six-cluster solution. Table 4 reports the final results obtained from the non-hierarchical cluster analysis on the effective sample of 845 entrepreneurs, using the centroids of the average linkage six-cluster solution as initial seeds.

Cluster 1 is composed of a 10 % of the entrepreneurs and is characterized by very low scores on factor 4, and close to average scores on the other strategic dimensions. Hence, this cluster captures a group of entrepreneurs which place a disproportionate emphasis on competing on prices, which is consistent with the pursuit of a low-cost strategy. Clusters 3, 5, and 6, in turn, seem to capture the groups of entrepreneurs in the sample that pursue a differentiation strategy. Cluster 3 entrepreneurs (17 % of the sample) place a high emphasis on enhancing the customers' perceptions of the value of their offerings. Cluster 5 entrepreneurs (9 % of the sample) place a disproportionate emphasis on enhancing customer value by means of innovative offerings. Cluster 6 entrepreneurs (9 % of the sample) place a high emphasis on enhancing the value of their offering through a broader product line. Cluster 4 includes only 5 % of the entrepreneurs; they identify themselves as pursuing a focus strategy. This group of entrepreneurs is pursuing the creation of a competitive advantage by better fitting their offerings to the specific needs of a particular subgroup of customers. However, such competitive advantage may take the form of either a lower cost or a higher perceived value for the target customer group. Finally, Cluster 2 (51 % of the sample) is composed of a large number of entrepreneurs that do not seem to trade-off any strategic dimensions against any others. This most populated strategic profile of entrepreneurs could be characterized as a 'stuck-in-the-middle' strategy (Porter, 1980).

TABLE 4

Typology of Entrepreneurs According to their Strategic Orientations
(N=845)

   

Mean (standard deviation) of factor scores

Cluster

Number of
Entrepreneurs

'Image'

'Unique
Product'

'Product
Breadth'

'Low Cost'
(inverted)

'Focus'

1

81

- 0.267
(0.42)
- 0.005
(0.60)
0.129
(0.67)
- 2.372
(1.01)
- 0.249
(0.60)

2

432

- 0.317
(0.54)
- 0.386
(0.73)
- 0.431
(0.81)
0.245
(0.56)
- 0.126
(0.72)

3

140

1.617
(0.70)
0.056
(0.59)
0.110
(0.71)
0.138
(0.56)
0.272
(0.89)

4

39

- 0.426
(0.60)
0.271
(1.02)
- 0.181
(0.62)
0.152
(0.73)
2.817
(1.07)

5

73

- 0.788
(0.64)
2.091
(0.98)
0.330
(0.87)
0.492
(0.49)
- 0.239
(0.64)

6

80

- 0.188
(0.68)
- 0.081
(0.75)
1.669
(0.71)
0.367
(0.66)
- 0.670
(0.64)
R-Squared:   0.623 0.456 0.387 0.607 0.443
CCC: - 3.362          
Pseudo-F 168.62          

 

RESULTS FROM THE CONFIRMATORY ANALYSIS

Hypothesis 1 will be tested by fitting full factorial ANOVA models (i.e., with interaction term) of strategy typologies and HRM typologies on the two measures of performance. A significant interaction effect between the strategy and the HRM typologies would provide support for this hypothesis. If such an interaction effect does, in fact, exist one can proceed to analyze its nature and form, and hence test the two remaining hypotheses posed in this study. Given the exploratory results described in the previous section, the test of Hypothesis 2 will entail the contrast of the performance differences between entrepreneurs classified as belonging to the HRM cluster 2 (the outcome control group) and those belonging to the HRM cluster 6 (the behavior control group), out of the group of entrepreneurs belonging to the Low-Cost strategy cluster. Similarly, hypothesis 3 will be tested by contrasting the HRM-2 cluster with the HRM-6 cluster, for the groups of entrepreneurs belonging to either of the three differentiation strategy clusters.

Table 5 summarizes the results obtained from the analysis of variance on the subjective measure of performance, and the analysis of covariance on compensation taken out of the business (after controlling for any effects of size). Given, the unbalanced nature of the anova design (i.e., different treatment counts), partial sums of squares (i.e., type III estimable functions in SAS) were employed to test for the significance of effects. These SS are equivalent to the results that would be obtained by using the regression approach to anova (Neter, Wasserman, and Kutner, 1990).

TABLE 5

Results of Analysis of Variance and Analysis of Covariance

 

Source of Variance -- F-values

Dependent variables Strategic
Orientation
HRM
Approach
Interaction Venture Size
TAKEOUT:
Model 1
Model 2

2.18*
1.68

1.85*
1.87*


1.01

30.53***
31.32***
INCPERF:
Model 1
Model 2

0.68
1.03

1.11
1.91*


1.56**
 
         
Significance levels: * p<0.10, ** p<0.05 , *** p<0.01

 

Evidence was found for a significant main effect of strategy orientation on the overall compensation taken out by the entrepreneur. Pairwise comparisons of least-squares means across strategy typologies revealed that the stuck-in-the-middle typology of entrepreneurs exhibited significantly higher short-term performance than entrepreneurs in the other strategic typologies. This result may indicate a lower rate of reinvestment for entrepreneurs who are not trying to build a competitive advantage vis-a-vis existing competition. There is also evidence of a significant effect of the human resource management approach of the entrepreneur on his short-term performance. Pair-wise differences between HRM typologies resulted in two main conclusions. Entrepreneurs belonging to the HRM typology 3, composed of entrepreneurs who are not prone to use either behavior or outcome motivators, show significantly lower short-term performance than entrepreneurs belonging to any other HRM typology. This result indicates that motivation of employees is a factor that entrepreneurs should , in any case, not overlook as it is important to the performance of their businesses. On the other hand, entrepreneurs in typology 4 show significantly higher levels of performance than entrepreneurs in typologies 1 or 2 (besides typology 3). This result indicates that there are only 3 HRM typologies that should be considered by entrepreneurs in order to maximize their short-term performance (i.e., typologies 4, 5, and 6). There are two important characteristics of the human resource management approaches employed by entrepreneurs in these three viable forms of motivation: First, the entrepreneurs that exhibit significantly higher levels of performance are arranged along a downward sloping continuum in the two-dimensional space of behavior and outcome control factor scores. This illustrates that higher performance is obtained when entrepreneurs effect a trade-off between behavior and outcome forms of control. Second, none of the HRM typologies found to deliver superior short-term financial returns to the entrepreneur placed a high emphasis on outcome-based motivation. This result indicates that high levels of decentralization of decision making and of variable pay-for-results produces lower short-term monetary returns for entrepreneurs in the start-up phase. Finally, there is no evidence of any significant matching effects between strategic orientations and HRM approaches that should be taken into account by entrepreneurs striving to maximize their short-term financial returns. Accordingly, no support is found for the hypothesized matching effects bearing upon financial performance.

FIGURE 4

Least Squared Means of HRM approaches, by Strategy Type

LOW-COST STRATEGY

DIFF. BASED ON IMAGE

DIFF. BASED ON UNIQUE PRODUCTS

DIFF. BASED ON PRODUCT BREADTH

With respect to the analysis on the increase in satisfaction measure, there are indications of a significant main effect of control practices. Pair-wise differences between least squares means across HRM typologies resulted again in HRM typology 3 producing significantly lower satisfaction levels than any other motivational approach. The importance of some type of motivational practice is also demonstrated with respect to the entrepreneur's overall satisfaction with the business. Pair-wise least squares differences also showed HRM typology 4 to produce significantly higher satisfaction levels than either typology 1 or typology 5. Hence, once more, there are only three HRM typologies that should be used by entrepreneurs to maximize their overall satisfaction with the business. The viable typologies are also arranged along a downward sloping continuum, illustrating a necessary trade-off between behavior and outcome alternatives to control. Nevertheless, in this case, the three viable typologies include the 'outcome-control' group, the 'behavior-control' group, and HRM typology 4, which will be referred to as the 'balanced-control' strategy.

There are also indications of a significative interaction effect between the form of control and the strategic orientation of the entrepreneur, as they bear upon the entrepreneur's increase in satisfaction. This result provides support for Hypothesis 1. Detailed analysis of the nature of such an interaction effect was undertaken with the aim to test hypotheses 2 and 3. Figure 4 show the plots of the HRM typologies' least squares means for each strategy cluster of interest. Pair-wise comparisons within the group of entrepreneurs following a low-cost strategy, indicated a significantly higher average performance for the behavior-control group than for the outcome-control group (t=1.875, p=0.06). This result provides support for Hypothesis 2. With respect to entrepreneurs following strategies of differentiation, average performance for the outcome-control group was not significantly higher than average performance for the behavior-control group. Hence, no support was found for Hypothesis 3. Further analysis disclosed, however, significantly higher performance for the balanced approach to control versus the behavior control for both the image (t=2.13, p=0.03) and the unique products (t=2.49, p=0.01) differentiation groups. Apart from HRM 3 producing significantly lower performance, no other significant differences were observed across control typologies for the entrepreneurs pursuing a strategy of differentiation based on broader product lines.

Previous Page | Main Menu | Next Page

 

 

1997 Babson College All Rights Reserved
Last Updated 1/15/97 by Geoff Goldman & Dennis Valencia

To sign-up for the Center for Entrepreneurial Studies' publication lists,
please register with the
Entrepreneurship WebTeam.