CASO PRÁCTICO
C). CONSIDERACIONES PREVENTIVAS EN LA INTERVENCIÓN
Given the central role of firm age with respect to firm survival56, rather than comparing the
entire groups of IAs, OAs and NAs, the groups were split by firm age. Firm age was captured as part of the original BLO Survey (Table 10 – where “Firm age <5yrs” represents the BLO Survey age categories coded as 2<3 years, 3<4 years, 4<5 years, other categories as provided by BLO Survey). Age was coded with reference to 2005, that is, for the 2003 intervention the youngest firms in the BLO Survey must have been two years old.
Table 10 – Age categories as available from BLO Survey
Firm age in BLO Survey 2005 IA OA NA Total
Firm age <5yrs (2003: 0-2 yrs.) 111 64 34 209 Firm age 5<10yrs (2003: 3-7 yrs.) 94 56 49 199 Firm age 10<20yrs (2003: 8-17 yrs.) 103 92 72 267 Firm age 20+yrs (2003: 18+ yrs.) 115 109 144 368
Total 423 321 299 1,043
Source: BLO Survey
In comparison to the BLO survey, where age was coded in above categories, the BSD allows determining age as a continuous variable, using its birth and death variables as discussed in Chapter Three. Table 11 shows age as coded on the basis of the BSD, compared to the BLO survey categories.
56 For a better understanding of the importance of the role of size and age and potential other firm and
owner characteristics, probit regressions were run, with firm exit event as the binary dependent outcome variable. These confirmed the size and age impacts on survival, but provided no coherent results on any other suitable predictors.
- 103 -
Table 11 – Comparison of age categories, BSD and BLO Survey
Firm age 2003, BSD vs. BLO IA OA NA Total Total BSD BLO BSD BLO BSD BLO BSD BLO Firm age 0-2 yrs. 73 111 49 64 26 34 148 209 Firm age 3-7 yrs. 103 94 70 56 51 49 224 199 Firm age 8-17 yrs. 134 103 112 92 103 72 349 267 Firm age 18+ yrs. 115 115 90 109 120 144 325 368
Total 425 423 321 321 300 299 1,046 1,043
Source: BLO Survey and BSD
Table 11 includes the age categories as based on the BLO Survey; they are for their largest part congruent with firm age as recorded in the BSD. Ideally, the figures should be identical. Especially the figures for the youngest firms are considerably lower based on the BSD. One likely reason for this apparent gap is the way the BSD ‘learns’ of a firm’s existence: firm birth will be recorded in the BSD at the time of a firm’s registration for VAT of PAYE, as discussed in Chapter Three. BLO Survey interviewees may therefore have provided birth dates that precede BSD records. Other than that, as with any survey, there is of course the risk of inaccurate responses to the BLO survey, or inaccurate recording of the same.
It was chosen to work with age based on the BSD, therefore with firm age as reflected by official records. This allows for determining age groups that achieve a more even count by age band. It also ensures age was determined in a comparable way to the BSD reference group (introduced and used in Figure 2 and Figure 3). Table 12 shows the age categories as redefined on the basis of the BSD for the survival analysis within this chapter.
Table 12 – Recoded age categories based on BSD
Firm age 2003 (BSD), re-categorised IA OA NA Total
Firm age 0-3 yrs. 102 63 36 201
Firm age 4-9 yrs. 101 75 65 241
Firm age 10-19 yrs. 125 110 85 320
Firm age 20+ yrs. 97 73 114 284
Total 425 321 300 1,046
- 104 -
Figure 2 demonstrates the importance of age on survival for the sample of firms at hand.
Figure 2 – Survival estimates by firm age group (all firms in longitudinal dataset)
Source: BLO Survey and BSD
Figure 3 provides the same breakdown of firms by age group but for a random sample of 197,291 firms from the BSD57. It confirms the trends as observed in the longitudinal dataset to
reflect general survival rates fairly accurately.
57 The random sample was drawn from linked 2004-2011 BSD data. The total sample size of 197,291
divides into 55,531 firm 0-3 years old, 57,721 4-9 years, 47,811 10<19 years. The use of a random sample rather than the full firm population was on the basis of computing power limitations.
20+ yrs. 10-19 yrs. 4-9 yrs. 0-3 yrs.
- 105 -
Figure 3 – Survival estimates by firm age group (random sample from BSD)
Source: BSD
Whilst the overall survival trends are comparable, the firms as captured as part of the BLO survey (Figure 2) enjoyed a considerably higher survival rate than the overall firm population (as represented by the large random BSD sample in Figure 3). The reasons for these observed differences may be manifold – section 4.5.4 looks at this.
Returning to Figure 2, the question arising is to what extent these survival rates differ by age group when comparing assisted and non-assisted groups. Combinations for both IA and OA compared to NA were looked at across age groups.
The results obtained suggest little significant impact of assistance on survival. The following six graphs Figure 4 to Figure9 were chosen to illustrate the discussion of the results, split in pairs of both survival rate and resulting hazard rate graphs58. The results for OA were even
58 This analysis included the survival rates as obtained for the random sample from the BSD (introduced
in Figure 3). They are of purely indicative nature, and no robust conclusions can be drawn on firms outperforming the firm population based on assistance received – again section 4.5.4 discusses this and any observations from the random sample’s survival rates.
20+ yrs. 10-19 yrs.
4-9 yrs. 0-3 yrs.
- 106 -
less remarkable than those for the IAs and are therefore not specifically reviewed in the following.
Figure 4 shows how the youngest firms, which were no more than three years old at the point of intervention, enjoy no distinct advantage for survival as recipients of IA. No IAs exited in 2005, but around six percent of NAs (and ten percent of OAs did so – not shown). Given the upper confidence limit for NAs is at around 98 percent, the early IA survival performance is significantly better for the youngest firms than the NAs. IAs then faced survival rates below those of the NAs, for a period of three years up to 2009, and from 2010 onwards.
Figure 5, showing the according hazard rates, shows a high exit rate for NAs between 2008 and 2009, but far less so for IAs. However, with the exception of the 2005 IA survival performance, no significant differences exist between IA, OA and NA groups for firms of less than three years of age, with confidence limits overlapping nearly entirely59.
59 As the relatively wide confidence limits indicate, sample size for the 0-3 year old firms, which was
then split by category of assistance, is smaller compared to the other age groups defined in the BLO Survey, with particularly few NAs.
- 107 -
Figure 4 – Survival estimates IA vs. NA (for firms aged 0-3 years at intervention)
Source: BLO Survey and BSD
Figure 5 – Hazard rates IA vs. NA (for firms aged 0-3 years at intervention)
Source: BLO Survey and BSD
For the age category of 4-9 year old firms (Figure 6 and Figure 7, age at time of intervention) IA survival nearly always exceeds that of NAs (and OAs, not shown), with the hazard rate for the IAs remaining below that of the NAs until 2008. The approximate 77.5 percent survival rate in 2011 achieved by those 4-9 year old firms that benefitted from IA was near identical to the survival rate of NAs. Again, a significant difference between the survival rates was only found
0.4 0.5 0.6 0.7 0.8 0.9 1 2004 2005 2006 2007 2008 2009 2010 2011 IA NA BSD benchmark IA 95% CL (where outside NA CL) NA 95% CL 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 2004 2005 2006 2007 2008 2009 2010 2011 IA NA BSD benchmark
- 108 -
for IAs immediately post intervention in 2005, with their 100 percent survival lying outside the NA confidence limits. Thereafter, confidence limits overlap, from 2008 onwards to a considerable degree.
Figure 6 – Survival estimates IA vs. NA (for firms aged 4-9 years at intervention)
Source: BLO Survey and BSD
Figure 7 – Hazard rates IA vs. NA (for firms aged 4-9 years at intervention)
Source: BLO Survey and BSD
0.4 0.5 0.6 0.7 0.8 0.9 1 2004 2005 2006 2007 2008 2009 2010 2011 IA NA BSD benchmark IA 95% CL (where outside NA CL) NA 95% CL 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 2004 2005 2006 2007 2008 2009 2010 2011 IA NA BSD benchmark
- 109 -
Figure 8 and Figure 9 show how survival prospects increase by age, compared to the younger age groups discussed above. For firms aged 20 years and above at the time of the intervention, survival rates are better than for any younger cohorts. The hazard rates for IAs and NAs cross at multiple times at overall low hazard levels. Up to 2010, IA and NA survival performance is relatively similar. In 2011 88 percent of IAs remain against 84 percent of NAs, a non-significant difference in survival.
Figure 8 – Survival estimates IA vs. NA (for firms aged 20+ years at intervention)
Source: BLO Survey and BSD
Figure 9 – Hazard rates IA vs. NA (for firms aged 20+ years at intervention)
Source: BLO Survey and BSD
0.4 0.5 0.6 0.7 0.8 0.9 1 2004 2005 2006 2007 2008 2009 2010 2011 IA NA BSD benchmark IA 95% CL (where outside NA CL) NA 95% CL 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 2004 2005 2006 2007 2008 2009 2010 2011 IA NA BSD benchmark
- 110 -
Overall, Section 4.5.2 confirms the role of firm age for survival. For the effectiveness of support it suggests that when considering assistance by firm age group, there may be some limited evidence for a very short-lived significant effect of IA on those firms that are relatively younger. Beyond that, no significant effects of assistance on survival could be identified.