Data analysis and presentation of results from the pre-recession survey

This part of the chapter presents the data for the research questions/hypotheses in the same order as presented in chapter 2 and includes post-hoc tests where relevant and a brief justification of the statistical measures used where not fully covered in the methodology section.

4.4.1 Adaptability and firm age

A key question concerned the relationship between firm age as an indicator of survival and adaptability. Organisational ecologists maintain that older firms are more inert/less adaptable (Carroll, 1984; Hannan and Freeman, 1989; Harreld, O’Reilly, and Tushman, 2007), while, for organisational strategists, older surviving firms have learned to be more adaptable (Levinthal, 1991; Durand and Coeurderoy, 2001). For the organisational ecologists, inertia is the result of previous successes and a consequence of selection, as well as an attribute that enhances survival, as inertia-disrupting organisational change leads to reduced performance and death. Even if inertia is relative and organisations do change all the time in some way, inertia for the ecologists is still a drag on change such that firms find it hard, if not impossible, to keep up with the ever-changing environment.

Chapter 2 noted that the probability of failure conditional on age (the hazard rate) is shown empirically to decline with age (Phillips and Kirchhoff, 1989; Audretsch and Mahmood, 1995). If the probability of survival of a firm increases with age, does its adaptability increase or decrease with age? For organisational strategists, individual adaptability must increase with age as entrepreneurs and their teams learn to adapt (Levinthal, 1991) and the

more fit firms must somehow be better at reading and interpreting what is going on and then adapting to market and technology changes (Schindehutte and Morris, 2001), generating rising average adaptability. For the organisational ecologists, whatever adaptability is occurring at the individual level, at the population level structural inertia increases with age.

The corollary is that the average adaptability of firms in the population must be decreasing with age where selection processes significantly favour those with high levels of inertia (Hannan and Freeman, 1984).

In order to address this issue, a Welch’s ANOVA was conducted as the purpose of Welch’s ANOVA is to determine whether there are statistically significant relationships between a dependent variable that is continuous (adaptability) and an independent variable that is categorical (firm age), while assuming that the variances of the independent variables are not equal.

There was a significant difference in the adaptability of the firm by the age of the firm, F(4, 58.82) = 4.14, p < .01. In fact, firms ten years old or older had significantly higher adaptability scores than organisations three to five years old. None of the other comparisons were significantly different from one another and the average value for each age group is presented in Table 14. The analysis also shows that firms ten years old or older had significantly higher adaptability scores than firms less than ten years old, F(1, 850.79) = 14.64, p < .01, and the average values for this are shown in Table 15.

Table 14. Mean adaptability scores by age of firm

N M=Mean SD

10 years + 492 2.4920 .59066

6–9 years 160 2.3447 .59915

3–5 years 198 2.3227 .69921

1–2 years 49 2.2878 .52356

First year 10 2.5900 .90025

Source: Analysis of research survey, April 2008

Table 15. Mean adaptability scores over and under ten years trading

N M SD

10 years + 492 2.4920 .59066

Fewer than 10 years 417 2.3335 .64794

Source: Analysis of research survey, April 2008

The result implies that, even if the scope for a firm to change adaptability is small, the more adaptable have some advantage relative to their rivals that confers greater longevity, which is in turn associated with greater survivability, as the hazard rate has been shown to decline with age.

As noted in chapter 2, however, there may be a commonplace explanation for the age-dependency of survival, accounted for by the heterogeneity of the population. As a cohort of firms ages, the risk set becomes increasingly composed of firms with the lowest propensity to exit (Thompson, 2005); those that have not yet exited are those less likely to exit. The mean death rate for the cohort can decline with cohort age, even if the hazard rate does not decline with age for any individual firm. In his shipbuilding study, Thompson (2005, ibid.) shows both that the usual age-dependency of exit is present in the data, and that it disappears with the addition of the quality proxies to the hazard regression, implying the initial age-dependency can be explained by selection bias.

At the macro level, the result presented here may also be accounted for by the heterogeneity of the population, with younger cohorts of firms having a spread of adaptability levels and the less adaptive being weeded out through time. This would conceptually be consistent even with a decline in adaptability for every single firm as it ages. Depending on firm birth and death rates, the average adaptability of the remaining contingents can rise even if adaptability falls for every single firm, as shown illustratively in the diagram at Figure 1.

Figure 1. Individual firm adaptation versus change in population characteristics

The diagram shows the heuristic example of ten companies over 14 time periods, each assigned a life of between six and 14 years and each with constantly declining adaptability after year five. The result demonstrates explicitly the distinction between the adaptability of the individual firm and the average adaptability of the population, between the capacity of an individual firm to survive by adapting to the changing environment and the average level of adaptability in a population of SMEs. Average adaptability can increase even when adaptability is falling for all firms in the population.

A more refined result was demonstrated in a multi-objective optimisation simulation using the Optimisation Toolbox in Matlab, a software package for mathematical computing and visualisation. The assumptions were:

a population of 100 firms;

normally distributed initial adaptation levels, with mean 1.0 and standard deviation r

survival rates for new entrants decline reasonably constantly (ONS) at the rate of 93 per cent, 78 per cent, 63 per cent, 53 per cent, 45 per cent between years one and five;

a death rate fixed at 10 per cent and birth rate at 12 per cent per annum;

in each year the adaptability value of each firm alters by a fractional amount, a, so each firm exhibits an exponential increase or decrease in its adaptability; and