For policymakers to fully comprehend the influence of enterprise deaths on innovation and economic development, in response to a macroeconomic shock, greater understanding of the determinants of enterprise deaths is required. For example, are new small firms more likely to fail than established or mature small firms? Are new large firms more likely to fail than mature large firms? Does industrial sector play a role in firm failure? Are new small firms in cities more likely to fail than new small firms in rural locations? Does it matter when the firm was established? During normal market conditions, small firms, young firms, firms that fail to innovative, modernise and remain competitive, and lifestyle firms with ageing owner-managers are more likely to die (Cefis and Marsili, 2005). While these determinants remain important during crisis periods, other factors, such as restricted access to finance, higher borrowing costs, and increased market risk can result in the death of otherwise healthy firms. While there are a lot of studies of new firm formation, growth and survival, little is known about the determinants of firm deaths and how these respond to changes in the business cycle (Ejermo and Xiao, 2014). In this paper, we examine the exit behaviour of more than 300,000 Irish firms of varying sizes and ages between 2006 and 2010.
A number of studies suggest that, since new, established and mature Typically, new firms are less than 2 years old, established firms are between 2 and 10 years old and mature firms are more than 10 years old.
This paper examines the influence of determinants on the hazard rates of five cohorts of Irish firms, following the major environmental shock posed by the 2007 financial crisis. This shock had a significant effect on all aspects of the economy, with the Irish Government issuing a broad state guarantee of Irish domestic banks in September 2008, followed by the recapitalisation of its two largest banks in February 2009, and an agreement with the International Monetary Fund, the European Central Bank, and the European Commission, for a three-year, €67.5 billion bailout in December 2010 (Bergin
The paper makes two important contributions to the literature on firm deaths. First, by using a very broad representative sample of more than 80% of all enterprises in Ireland (NACE 4-digit Codes A-U between 2006 and 2010), we accurately characterise enterprise deaths over the financial crisis period, using an array of firm, industry, economic and regional determinants. Few empirical studies that focus on firm deaths include a broad-based sectoral analysis, except for Germany (Fackler
Briefly, this paper is structured as follows: we begin by locating our hypotheses in the extant literature. Next, we introduce our data and key variables, before reporting the results of the CPH models. Finally, we present our conclusions.
Theoretical approaches taken in studies of firm deaths have generally, but not exclusively, adopted an organisational ecology (Geroski
The competing theoretical explanations focus largely on the struggle of new firms to survive post-entry, given their ‘newness’. This is referred to as the ‘liability of newness’ (Stinchcombe, 1965). This struggle to acquire resources is likely to be more severe during an economic shock. In addition, according to Barron et al. (1994), ageing itself brings two liabilities, which increase the risk of firm death: (1) the ‘liability of obsolescence’, where the structural inertia in the face of environmental change limits the firm’s ability to make the necessary adjustments to survive; and (2) the ‘liability of senescence’, where increasing bureaucracy reduces the firm’s efficiency, thus threatening its ability to survive. Mature firms face both these liabilities in a period of economic distress.
There is some evidence in support of the liability of ageing. Fackler
dependence solely among younger firms, thus supporting the underlying premises of learning models of industry dynamics (Jovanovic, 1982, Ericson and Pakes, 1995) and capabilities, and positional advantage arguments based on the resource-based view of the firm. However, they found no evidence to support the liability of ageing. Our first hypothesis captures the uncertain and complex non-linear relationship between age and the hazard rate.
H1: The probability of firm exit decreases with age but rises again in the later stages of the firm lifecycle. H2: Smaller firms face a higher hazard rate than larger firms, but the hazard rate of mature smaller firms is unaffected or increases with reductions in size. H3: The hazard rate of younger firms is heightened by negative structural market conditions, whereas the hazard rate of mature firms is unaffected or positively affected by structural market conditions. H4: The hazard rate of young firms in the high-technology manufacturing sectors and knowledge-intensive sectors is lower than that in other sectors (e.g. low-technology manufacturing, low-knowledge services). H5: Firms founded in adverse economic conditions face a higher hazard rate. H6: Hazard rates of mature firms are more sensitive to economic downturns, whereas hazard rates of newly founded firms are less sensitive to economic downturns. H7: Young firms located in urban areas face a higher hazard rate than firms located in rural regions.
Irish business demography data covering all sectors (NACE Rev. 2 Sectors A–U), collated annually from administrative sources by the Irish Central Statistics Office (CSO) between 2006 and 2010, is adopted in this paper. Principle variables recorded in the data are the number of active enterprises, enterprise births and deaths, and their associated NACE sector and contribution to employment. Enterprises are active if they have paid VAT, registered employees, or filed a corporation/income (with over Є50,000 turnover) tax return. Enterprise deaths do not include exits from the population due to mergers, acquisitions, divestments or restructuring, or enterprises, which reactivated within two years (Eurostat, 2007). Since most firms in our sample are geocoded, we use firm location to incorporate attributes of the region. We use the enterprise birth year to identify the enterprise’s birth cohort. Our final sample consists of 218,880 active enterprises (84% of Irish active enterprises) in 2010 and 104,031 enterprise deaths (82% of Irish enterprise deaths) 2006–2009, where the geographical location of the enterprise is known to county level. Further details on the sample are available from the authors.
We adopt a cohort approach similar to that of Box (2008) to examine how the determinants impact the probability of exit by birth cohort. We examine five cohorts of firms: (1) firms founded prior to 1980 (5,332 firms, 14% exited); (2) firms founded between 1980 and 1989 (39,578 firms, 11.7% exited); (3) firms founded between 1990 and 1999 (74,410 firms, 29.6% exited); (4) firms founded between 2000 and 2005 (75,469 firms, 35% exited); and (f) firms founded between 2006 and 2010 (140,477 firms, 41.2% exited). Focusing on these cohorts enables a detailed examination of the impact of firm and wider environmental factors on the exit rates of young, established and mature cohorts over the crisis period. Figure 1 shows that more firms exited from young and adolescent cohorts. The percentage exiting declines with age across the cohorts but rises again for the oldest cohorts who were founded before 1980 supporting hypothesis 1 and 2.
Figure 2, a smoothed hazard rate, provides further support for these hypotheses. This figure demonstrates that firm failure risk was higher for younger firms. It continued to rise for firms that were almost 20 years in business before falling again for firms over 55 years, before rising for firms trading up to 70 years. The fall and rise in the failure risk are repeated for older or third-generation businesses. The higher failure risk for older firms in the crisis coincides with periods when businesses require successors, either within or outside the family. Next, we briefly examine some of our key variables. Detailed information on the calculation of the determinants included in our estimations along with their associated summary statistics are available from the authors.
Size is approximated by the total number of people currently engaged in the business. Initial size is not available for firms founded in mature cohorts. Most firms in Ireland are micro firms (<10 employees), with very few having more than 50 employees. We break the total number of people currently engaged in the business into five dummies (0, 1–4, 5–9, 10–49, 50+ employees) to capture the effect of micro, small- and medium-sized firms on the probability of exit. Figure 3 illustrates that the vast majority of exits from each cohort throughout the crisis had 1–4 employees.
Minimum efficient scale is measured, in accordance with Baldwin and Gellatly (2003), as a ratio of average new birth to firm size for each NACE Rev. 2 four-digit code to capture the influence of economies of scale. We also include the logarithmic of average capital intensity in each NACE Rev. 2 four-digit code. Average sectoral growth is measured by logarithmic difference in industry employment in each NACE Rev. 2 4-digit sector code between 2006 and 2010. The average entry rate is approximated in accordance with Lin and Huang (2008). If industries are cumulatively characterised by growth and high entry rates, the likely duration of firms is shorter. We include an interaction term to capture this effect. We also include a measure of the sunk costs associated with the industry, equating to 1 minus the minimum average entry and exit rate (Bernard and Jensen, 2007). The Herfindahl Index was measured at the NACE Rev. 2 four-digit level similar to that of Pe’er and Keil (2013), to capture the effect of concentration. The proxy for R&D intensity, as with Strotmann (2007), was captured by three dummy variables; high-and medium-high-technology (1/0) manufacturers, knowledge-intensive services (1/0) and the reference category (1/0), which includes low-technology manufacturing, low-knowledge services and other sectors, as defined by the Eurostat (2014) indicators of high-tech industry and knowledge-intensive services. Broad sectoral dummies for the significant sectors within Ireland were also included in the estimation. Figure 4 shows that the construction (1/0) and other services sectors (1/0) suffered particularly among younger and adolescent cohorts, along with the professional and scientific sectors (1/0), whereas firms in agriculture/fishing/forestry (1/0), manufacturing (1/0) and wholesale/retail/repairs (1/0) had the greater shares of exits in mature cohorts.
A measure of the growth in local unemployment over the period 2007 to 2010 is included, to capture local economic distress caused by the financial crisis in Ireland. Local unemployment is computed using county-level unemployment rates (2005 to 2012) for Ireland, compiled by Fitzpatrick and McQuinn (2014). We also include a dummy variable for firms that were founded before the financial crisis. This variable has a value of ‘1’ for firms founded before 2007 and ‘0’ otherwise. Such firms are expected to have a higher likelihood of survival (see Geroski et al. (2010).
An urban dummy was included, which took on a value of ‘1’ if the electoral district (DED) population was more than 1,500 and zero otherwise. Most deaths were in urban districts, with little variation across cohorts (see Figure 5). We also control for attributes of the region in which the firm is located, using data on the local labour market in each DED obtained from the census 2006. We include the percentage of the local population at work, the percentage of the local population with third-level education, and the percentage of the local population employed across broad sectors. The location of the firm is controlled for using 25 county dummies. The sample size falls when the regional variables are controlled, as not all firms in the sample give their exact address. The model is therefore estimated twice – with and without these factors.
We examine influences on the hazard rates of mature and young cohorts of firms during the financial crisis, using a CPH models with an Efron correction for ties. We compare these estimates to accelerator failure time (AFT) models, to ensure the stability of the coefficients across estimation techniques. Result available from the authors.
The CPH model specifies the hazard function
where
We present the CPH estimates in Table 1. A basic and extended model is estimated for each cohort and for the full sample, where the extended model includes characteristics of the region. Our dependent variable is the hazard rate. A positive (negative) coefficient means that the equivalent variable increases (reduces) the instantaneous probability of exit, thereby decreasing (increasing) the firms’ chance of survival.
Estimates of Cox Proportional Hazard Model across cohorts (with Efron correction for ties)
Zero employees (dummy) | 2.8070*** (0.2336) | 2.8000*** (0.2944) | 2.5559*** (0.1243) | 2.7103*** (0.1802) | 2.2888*** (0.0633) | 2.3643*** (0.0950) | 1.8708*** (0.0604) | 1.7485*** (0.0837) | 1.5619*** (0.0522) | 1.4044*** (0.0757) | 2.3399*** (0.0319) | 2.3515*** (0.0456) |
1–4 employees (dummy) | 2.3839*** (0.2267) | 2.2396*** (0.2806) | 2.0433*** (0.1182) | 2.0209*** (0.1676) | 1.8728*** (0.0613) | 1.9320*** (0.0907) | 1.5236*** (0.0591) | 1.4739*** (0.0813) | 1.4025*** (0.0515) | 1.2854*** (0.0741) | 2.0167*** (0.0312) | 1.9798*** (0.0443) |
5–9 employees(dummy) | 0.6038** (0.2901) | 0.4123 (0.3647) | 0.4429*** (0.1475) | 0.5626*** (0.2052) | 0.5546*** (0.0731) | 0.6117*** (0.1067) | 0.4683*** (0.0699) | 0.4649*** (0.0941) | 0.4088*** (0.0592) | 0.4190*** (0.0840) | 0.6365*** (0.0367) | 0.6768*** (0.0514) |
10–49 employees (reference) | ||||||||||||
50+ employees (dummy) | −0.7813 (0.5476) | −0.9160 (0.7595) | −0.5952* (0.3534) | −0.7864 (0.5282) | −1.2315*** (0.2569) | −1.6639*** (0.5082) | −0.7745*** (0.2211) | −0.5609* (0.2884) | −1.1967*** (0.2479) | −1.4514*** (0.4148) | −1.1194*** (0.1248) | −1.0761*** (0.1905) |
Minimum efficient scale | 0.5259* (0.2786) | 0.2766 (0.4136) | 0.4961*** (0.1286) | −0.0704 (0.2357) | 0.1682*** (0.0567) | 0.1692* (0.0976) | 0.3113*** (0.0472) | 0.3809*** (0.0693) | 0.3784*** (0.0333) | 0.3543*** (0.0559) | 0.3379*** (0.0240) | 0.5294*** (0.0378) |
Entry rate | −0.0631 (0.0399) | −0.0904 (0.0582) | −0.0824*** (0.0184) | −0.0441 (0.0332) | −0.0624*** (0.0082) | −0.0943*** (0.0147) | −0.0389*** (0.0068) | −0.0818*** (0.0116) | −0.0582*** (0.0041) | −0.0639*** (0.0077) | −0.0146*** (0.0030) | −0.0267*** (0.0054) |
Sectorial growth | 0.0059 (0.4335) | −0.1751 (0.5975) | −0.0008 (0.2698) | −0.2557 (0.4520) | −0.3820*** (0.1243) | −0.7363*** (0.2161) | −1.0207*** (0.0989) | −1.2379*** (0.1529) | −0.6106*** (0.0613) | −0.6844*** (0.0961) | −0.7563*** (0.0413) | −0.9628*** (0.0628) |
Entry size | 0.1802* (0.1071) | 0.2062 (0.1287) | 0.1499*** (0.0424) | 0.2648*** (0.0701) | 0.1282*** (0.0202) | 0.1738*** (0.0317) | 0.1213*** (0.0235) | 0.1853*** (0.0341) | 0.1489*** (0.0149) | 0.1329*** (0.0276) | 0.1283*** (0.0102) | 0.1194*** (0.0178) |
Herfindahl Index | −0.5713** (0.2641) | −0.5779* (0.3366) | 0.3506 0.2318) | −0.0652 (0.4111) | 0.3360*** (0.1119) | 0.7006*** (0.1899) | −0.1750** (0.0864) | −0.2123* (0.1155) | 0.1574*** (0.0611) | 0.2390** (0.0961) | 0.1419*** (0.0435) | 0.2214*** (0.0643) |
Entry and exit cost | −0.2417*** (0.0701) | −0.3151*** (0.1015) | −0.2386*** (0.0304) | −0.1392** (0.0584) | −0.1917*** (0.0136) | −0.2164*** (0.0260) | −0.1575*** (0.0115) | −0.2106*** (0.0199) | −0.1906*** (0.0071) | −0.2230*** (0.0132_ | −0.2238*** (0.0052) | −0.2657*** (0.0093) |
Entry rate x sectorial growth | −0.0191 (0.0541) | 0.0169 (0.0888) | 0.0276 (0.0357) | 0.0551 (0.0660) | 0.0679*** (0.0162) | 0.1270*** (0.0314) | 0.1175*** (0.0124) | 0.3809*** (0.0693) | 0.0521*** (0.0069) | 0.0613*** (0.0118) | 0.0798*** (0.0046) | 0.1123** (0.0078) |
Capital intensity | −0.0305 (0.0349) | −0.0108 (0.0475) | −0.0284* (0.0158) | 0.0278 (0.0260) | −0.0019 (0.0069) | 0.0011 (0.0122) | −0.0063 (0.0059) | −0.0818*** (0.0116) | −0.0101*** (0.0040) | −0.0257*** (0.0066) | 0.0006 (0.0029) | 0.0014 (0.0048) |
Low technology/low knowledge services/other (reference) | ||||||||||||
High and medium/high technology (dummy) | 0.2550 (0.3090) | 0.0696 (0.4206) | 0.4068** (0.1713) | 0.5186** (0.2430) | −0.0166 (0.1238) | −0.1268 (0.2054) | 0.1022 (0.1027) | 0.1464 (0.1381) | 0.0638 (0.0819) | 0.1184 (0.1192) | 0.0291 (0.0529) | 0.0763 (0.0771) |
Knowledge-intensive services (dummy) | 0.0176 (0.1739) | 0.0573 (0.2493) | 0.0711 (0.0761) | −0.1434 (0.1468) | −0.0030 (0.0328) | 0.0071 (0.0621) | 0.0260 (0.0263) | 0.0228 (0.0432) | 0.0424** (0.0177) | 0.0591* (0.0319) | 0.0666*** (0.0130) | 0.0986*** (0.0228) |
Born pre-financial crisis (dummy) | 0.1714*** (0.0098) | 0.1348*** (0.0164) | −1.4809*** (0.0099) | −1.5047*** (0.0164) | ||||||||
Regional unemployment growth | 0.0383 (0.1068) | 0.1014 (0.2556) | 0.0782 (0.0497) | 0.2485** (0.1229) | 0.0467** (0.0225) | 0.0420 (0.0555) | 0.0179 (0.0187) | 0.0259 (0.0389) | 0.0243* (0.0134) | 0.0167 (0.0299) | 0.0146 (0.0096) | 0.0554*** (0.0213) |
Urban (dummy) | 0.0169 (0.0888) | 0.0757 (0.0716) | 0.0174 (0.0311) | 0.0600*** (0.0225) | 0.0620*** (0.0183) | 0.0163 (0.0125) | ||||||
Percentage at work | 0.0136 (0.1250) | −0.0061 (0.0065) | 0.0032 (0.0029) | −0.0002 (0.0021) | −0.0017 (0.0016) | 0.0042*** (0.0012) | ||||||
Percentage with third-level education | 0.0056 (0.0120) | 0.0137 (0.0099) | −0.0001 (0.0045) | −0.0008 (0.0031) | 0.0018 (0.0025) | 0.0013 (0.0017) | ||||||
Percentage working in agri forestry, fishing | 0.0112 (0.0187) | 0.0072 (0.0173) | −0.0224*** (0.0080) | −0.0143*** (0.0053) | 0.0004 (0.0042) | −0.0070** (0.0031) | ||||||
Percentage working in construction | −0.0040 (0.0308) | 0.0042 (0.0099) | 0.0067 (0.0043) | 0.0027 (0.0031) | 0.0032 (0.0025) | 0.0100*** (0.0017) | ||||||
Percentage working in manufacturing | −0.0101 (0.0179) | −0.0151** (0.0073) | 0.0029 (0.0030) | −0.0018 (0.0023) | −0.0007 (0.0018) | −0.0001 (0.0013) | ||||||
Percentage working in commerce trade | −0.0028 (0.0127) | −0.0120 (0.0081) | −0.0030 (0.0036) | −0.0020 (0.0027) | 0.0032 (0.0022) | −0.0006 (0.0015) | ||||||
Percentage working in transport commerce | 0.0037 (0.0144) | 0.0067 (0.0168) | 0.0142 (0.0074) | −0.0006 (0.0054) | −0.0044 (0.0043) | 0.0083*** (0.0030) | ||||||
Percentage working in public administration | 0.0271 (0.0310) | 0.0128 (0.0121) | −0.0026** (0.0060) | −0.0129*** (0.0044) | 0.0024 (0.0033) | 0.0003 (0.0023) | ||||||
Percentage working in professional services | 0.0294 (0.0186) | −0.0023 (0.0076) | 0.0001 (0.0034) | −0.0089*** (0.0025) | −0.0039** (0.0020) | −0.0038*** (0.0014) | ||||||
Agriculture, fishing and extractive (dummy) | −0.3651 (0.2263) | −0.6301** (0.2871) | 0.2061** (0.1006) | −0.3567* (0.1875) | 0.0806* (0.0490) | −0.0224*** 0.0080 | −0.1330** (0.0468) | −0.3979*** (0.0794) | −0.3221*** (0.0374) | −0.4112*** (0.0678) | −0.5427*** (0.0226) | −0.5816*** (0.0420) |
Manufacturing (reference) | ||||||||||||
Construction (dummy) | −0.1958 (0.1876) | −0.2187 (0.2517) | 0.0676 (0.0982) | 0.3564** (0.1561) | 0.2832*** (0.0478) | 0.4141*** 0.0739 | 0.2371*** (0.0375) | 0.2828*** (0.0494) | 0.1819*** (0.0288) | 0.1182** (0.0440) | 0.4709*** (0.0198) | 0.4370*** (0.0284) |
Wholesalers, retail and repair of motor vehicles (dummy) | −0.5627*** (0.1575) | −0.7999*** (0.2092) | −0.0840 (0.0896) | 0.0579 (0.1368) | 0.0127 (0.0455) | 0.0449 0.0689 | 0.0595 (0.0395) | 0.0369 (0.0516) | 0.0893*** (0.0291) | 0.1573*** (0.0439) | −0.0522*** (0.0200) | −0.0840*** (0.0287) |
Accommodation and food (dummy) | −0.7694*** (0.1974) | −0.8601*** (0.2443) | −0.0701 (0.1041) | 0.0356 (0.1543) | 0.2204*** (0.0497) | 0.2486*** 0.0733 | 0.1835*** (0.0491) | 0.1258** (0.0633) | 0.1629*** (0.0344) | 0.1761** (0.0536) | −0.0182 (0.0232) | −0.1295*** (0.0332) |
Professional, scientific, technical (dummy) | −0.4876* (0.2607) | −0.5067 (0.3590) | −0.1712 (0.1194) | 0.1905 (0.2119) | −0.0812 (0.0565) | −0.0512 0.0985 | −0.1169** (0.0458) | −0.0607 (0.0683) | −0.1302*** (0.0336) | −0.0933* (0.0551) | −0.0789*** (0.0235) | −0.0959** (0.0375) |
Other services (dummy) | −0.2750* (0.1654) | −0.3281 (0.2142) | 0.0048 (0.0884) | 0.1095 (0.1367) | −0.0205 (0.0439) | − −0.0301 0.0664 | −0.0451 (0.0366) | −0.0996 (0.0472) | −0.0515* (0.0278) | −0.0371 (0.0420) | −0.0119 (0.0191) | −0.0582** (0.0273) |
N | 5258 | 3432 | 39310 | 14625 | 73212 | 32544 | 74299 | 42065 | 130832 | 37687 | 322911 | 130353 |
Event | 753 | 479 | 4547 | 1440 | 21220 | 7649 | 25595 | 14136 | 51916 | 22328 | 104031 | 46032 |
2 LOG L (with covariates) | 10847 | 6503 | 88314 | 24270 | 425889 | 142146 | 530876 | 277168 | 1142381 | 443823 | 2428677 | 1002415 |
Wald | 405*** | 282*** | 1294*** | 519*** | 3637*** | 1829 | 4045 *** | 2477 *** | 6737*** | 2401*** | 54276*** | 24654 *** |
Notes:
1%;
5%;
10%
Size class at the time of the crisis had the largest relative impact on the probability of exit. A significant positive coefficient on the dummy variables for micro firms with zero employees, 1–4 employees and 5–9 employees in the full sample estimates (Columns XI–XII), and also in the estimates for each cohort (Columns I–X), indicated that micro firms face a lower likelihood of survival during the crisis period, in comparison with small firms with 10 to 49 employees, thus providing some support for Hypothesis 1. Firms with zero employees face the highest hazard rates, which can be obtained by calculating
Examining the full sample estimates (Columns XI–XII), firms operating in an industry with a higher minimum efficient scale and average size of new entrants over the period of the crisis are significantly more likely to exit, consistent with Strotmann (2007), whereas firms operating in industries with greater sunk costs over the crisis period face a significantly a lower likelihood of exit. Firms engaged in more concentrated markets over the crisis period captured by greater values of the Herfindahl index are also more likely to exit, while firms operating in growing sectors face a lower likelihood of exit. The latter effect is attenuated by higher entry rates into the sector over the crisis period, captured by the positive coefficient on the interaction variable between entry rate and sectorial growth, as found by Lopez-Garcia and Puente (2007). Contrary to expectations, firms operating in sectors with greater firm entry during the crisis period face a lower likelihood of exit, but similarly, this effect is diminished by higher sectorial growth, captured by the positive coefficient on the interaction variable between entry and sectorial growth.
Similar results are found for the individual cohorts. However, the hazard rates of firms founded prior to 1980, and to a slightly lesser degree those founded between 1980 and 1989, are not as sensitive to changes in industry conditions over the crisis period. Cohorts founded during the crisis, or just prior to the crisis period, are much more sensitive to changes in these conditions, providing strong indicative support for the Geroski
Focusing on industry technological conditions, we find that knowledge intensive services firms are significantly more likely to exit, relative to firms operating low-technology and low-knowledge sectors over the crisis period (
When examining sectorial effects in the full sample estimates, we find that construction firms face a significantly higher probability of exit than manufacturing firms, while firms in agriculture, fishing and extractive industries, wholesalers, retail and repair of motor vehicles and professional, scientific and technical sectors face a significantly lower probability of exit than manufacturing firms. Firms operating in other services and in accommodation and food also face a significantly lower probability of exit, when agglomeration variables at DED level are controlled for. The pattern is similar for the cohort of firms founded amid the crisis (2006–2010), except for wholesalers, retailers and repair of motor vehicles firms and firms operating in the accommodation and food sector, which were significantly more likely to exit, in comparison with manufacturing firms also founded at that time. The significant and positive coefficient on construction and accommodation and food sector dummies was clearly evident in firms founded from 1990 onwards. Agriculture, fishing and extractive industries in general show a lower probability of exit in younger establishments founded from 2000 onwards. Mature firms (founded before 1980) operating in the agriculture, fishing and extractive (extended model only), wholesalers, retailers and repair of motor vehicles, accommodation and food, professional, scientific and technical and other services sectors were significantly less likely to exit, in comparison with manufacturing firms. The survival of mature manufacturing and construction firms was threatened more than services industries.
Firms founded before the financial crisis are significantly less likely to exit during the crisis and face a much lower hazard rate when examining the full sample estimates, as is consistent with Esteve-Pérez and Mañez-Castillejo (2008). The magnitude of the impact on the hazard rate is sizeable at 77%. This indicates that macroeconomic conditions at founding have a significant influence on a firm’s chances of survival, supporting Hypothesis 5. The result for the 2006 to 2010 cohort differs. A significant and positive coefficient was obtained, when those firms founded immediately prior to the crisis faced a higher probability of exit. It is thus not just the conditions in the year of founding, but in the immediate period following this event, that can influence the firm’s survival chances. When examining the effect of regional unemployment growth per county on the hazard facing firms, we find that it raises the hazard rates of firms in the full sample (Column XII), when attributes of the local region are controlled for (impact on hazard rate of 6%). The coefficient on regional unemployment growth is also positive and significant for young firms (2006–2010 cohort, Column IX) and more mature cohorts, when examining estimates of the reduced model for firms founded 1990 and 1999 (Column V), and the extended model for firms founded between 1980 and 1989 (Column IV). The magnitude of the impact on the hazard rates is considerably larger for firms founded in 1980 and 1989 at 28%, in comparison with the younger cohorts at 5% and 2% for firms founded in 1990–1999 and 2006 to 2010. Thus, older cohorts seem to be more sensitive to the economic cycle, which corresponds with the findings of Boeri and Bellmann (1995) and supporting hypothesis 6.
Examining the influence of agglomeration approximated by the urban dummy, we find that being located in an urban area raises the likelihood of exit for younger cohorts, as is consistent with Strotmann (2007) and thus supporting Hypothesis 7. The coefficient of the urban dummy is not significant for the full sample estimates but is positive and significant for firms founded after 2000. Thus, agglomeration seems to affect the survival chances of younger cohorts, as opposed to more mature cohorts. The magnitude of the percentage change in the hazard rate is low for firms founded between 2006 and 2010 (6.2%) and for firms founded between 2000 and 2005 (6.4%).
When examining attributes of the local region at DED level, we find that firms operating in districts that have a higher percentage of individuals in employment face a higher hazard rate over the crisis period (Column XII). This effect is not significant for any distinct cohort. Its impact on the hazard rate is low, at 0.4%. Districts with a higher percentage of third-level graduates have no significant influence on the survival chances of firms located in these areas during the period of the crisis. Urban districts have a significantly higher mean percentage of people in employment (F Statistic=666, p-value<0.0001) and with third-level education (F Statistic=13627, p-value<0.0001).
In the full sample, firms located in districts with a high percentage of construction workers (1%) and workers in transport (8%) face a higher probability of exit over the crisis period, whereas firms located in districts with a high percentage of agriculture, forestry and fishing workers (−0.7%) or/and professional services workers (−4%) face a lower hazard rate over the crisis period (with associated impacts on the hazard rate of exiting in parentheses). Isolating cohort effects, we find that firms in younger cohorts (founded from 2000 onwards), located in districts with a high percentage of professional services workers, face a lower hazard rate over the crisis period. Firms in less-mature cohorts (founded from 1990 to 2005), located in districts with a high percentage of agriculture, forestry and fishing, and public administration workers, also face a lower hazard rate over the crisis period. Mature firms (1980–1989 cohort) operating in districts with a high percentage of manufacturing workers also face a significantly lower probability of exit. Urban districts have a significantly higher mean percentage of people in trade commerce (F Statistic=666, p-value<0.0001) and in transport commerce (F Statistic=13627, p-value<0.0001), otherwise the mean percentage workers is higher in rural regions, with respect to the industries above. The reduction in the probability of exit is tentatively in support of evidence of a higher likelihood of survival in rural regions, except unsurprisingly, the construction industry. In general, agglomeration seems to affect younger cohorts more than mature cohorts, but the percentage change in the hazard rate is low in magnitude, in comparison with those of firm size and industry conditions.
In this paper, we examine the influence of an array of firm, industry, economic and regional determinants on the hazard rates of a large representative sample of young, established and mature firms during the crisis period 2006 to 2010 in Ireland. From our empirical investigation, we find that the exit behaviour of young and mature firms is explained by different determinants. Specifically, we find that the liability of smallness is more pronounced for more mature micro firms than for younger counterparts founded amid the crisis. This is contrary to expectations, but supportive of the liability of ageing. By contrast, the hazard rates of mature firms are not as sensitive to changes in industry conditions over the crisis period. Younger cohorts are much more sensitive to changes in these conditions, as is consistent with Bellone
Exit behaviour from knowledge-intensive industries confirm the liability of newness. Firms founded during the crisis period in industries which are knowledge intensive faced a significantly higher probability of exit than those which entered medium-low and low- technology manufacturing and low-knowledge or other services industries. The estimates suggest a routinised regime in favour of the innovative activity of incumbents, rather than innovative entry in knowledge intensive services. The significantly higher probability of exit facing established high and medium-high-technology manufacturing firms confirms the liability of ageing. The higher probability of exit of construction and accommodation and food firms related to manufacturing firms during the crisis can largely be explained by a narrowing of the market within these sectors. Firms that are more than 30 years old at the time of the study in the services industries are significantly less likely to exit, in comparison with manufacturing firms, thus indicating that manufacturing firms are much more subject to liability ageing due to obsolescence and/or senescence during the crisis.
From a regional perspective, it is also the younger firms that seem to suffer. Agglomeration and industry-related skills at a regional level affect younger cohorts more than mature cohorts. However, their impact on the hazard rate is low, in comparison with the impact of firm size and industry conditions. In response to economic conditions, firms founded after the crisis have a much higher probability of exit than those founded prior to the crisis, but not in comparison with those founded immediately prior to the crisis (at the end of the good times). Entrepreneurs who founded firms during the crisis perhaps did so knowingly. The hazard rates of young firms are influenced by regional unemployment, but to a lesser degree than those influencing mature cohorts. This, again, demonstrates the importance of considering the influence of regional determinants on mature, as well as young, firms.
The novel comparison of the estimate’s pre-crisis, with those during the onset of the crisis, enables an analysis of true crisis effects and effects that are invariant with macroeconomic shocks. Larger firms, firms located in rural regions, and knowledge-intensive service business are less likely to exit pre-crisis and during the crisis. Smaller firms and younger firms are more likely to exit pre-crisis and during the crisis. Furthermore, firms that are jointly small and new are also more likely to exit under different economic environments. Older firms, in line with the findings of Boeri and Bellmann (1995), are more acutely affected during a crisis. They are culled earlier on in the crisis and hazard rates faced by firms with joint liabilities of smallness and ageing are more acute during a crisis.
Our results provide significant evidence that small and young firms face increased hazard rates, which means they might benefit from targeted policy support, in order to improve their probability of survival. While the OECD (2019) has highlighted that there are already significant supports for SMEs and entrepreneurial endeavours in Ireland, there is no unified SME and entrepreneurship policy document. There is an overarching enterprise policy document in place for Ireland,
Though we can identify firms that are more likely to exit during a crisis, it is challenging for policymakers to develop programmes to assist struggling businesses during severe downturns. It is hard to put together a comprehensive or even a targeted set of criteria, in terms of who should be supported and who should not. On aggregate, there is suggestive evidence that younger firms seem to suffer more than mature and established firms, but this is also true, though lower in intensity, in times of economic prosperity. Any policy – whether supportive of all or targeted groups (e.g. specific industries, new starts) – that is implemented is likely to have an associated deadweight and/or substitution effect (see Santarelli and Vivarelli (2007). Subsidies to new starts that would have survived despite such support represents deadweight, whereas subsidies for established firms that would have exited the market without the support represents a substitution effect. Relying on market selection, improvements in business and institutional conditions for all firms (e.g. reducing barriers to entry, cutting red tape and bureaucracy, and ensuring an efficient banking system and an attractive tax system), and reducing capital market imperfections, is arguably superior to directly subsidising mature, established or young businesses to prevent firm exits. In general, building an industrial base that is resistant to economic shocks is also a preferable approach to industrial policy (e.g. encouraging firms to increase scale, supporting knowledge-based services firms, encouraging rural entrepreneurship).
While we have presented, in this paper, the influence of determinants on the hazard rates of a large representative sample of firms during the crisis period, we cannot distinguish between voluntary closures and involuntary closures, as in the recent work by Mueller and Stegmaier (2015) and Balcaen