Changes in New Zealand's business insolvency rates after the GFC

Abstract

We examine the question of whether the rate of business insolvencies in New Zealand is related to overall macroeconomic conditions. In particular, our interest is in whether the rate of business insolvencies changed in the wake of the Global Financial Crisis (GFC). We find that there was a large increase in insolvencies in New Zealand following the onset of the GFC in 2008. We also find that the timing of the change did not occur uniformly over the country but occurred at different times in four key regional centres. Sharply rising relative costs were the most important macroeconomic factor influencing corporate insolvencies in New Zealand, Auckland, Waikato and Wellington, but have been immaterial in determining New Zealand's total personal insolvencies. It is employment growth and house price inflation that have been significant in explaining total personal insolvencies.

KEYWORDS:

• Bankruptcy
• business cycles
• structural breaks
• New Zealand
• Global Financial Crisis

JEL CLASSIFICATIONS:

• G33
• E32
• R11

Acknowledgements

We acknowledge valuable comments from Gael Price, Arthur Grimes and Dennis Wesselbaum, from presentations at Motu and the April 2019 NZ Macroeconomic Dynamics Workshop, and from an anonymous referee.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Much previous business cycle work for New Zealand's post-Second World War period has been focussed on identifying and explaining movements in classical business cycles, e.g. Hall and McDermott (2007, 2009, 2016), Reddell and Sleeman (2008), and Williams (2017a, 2017b). The latter two papers focus particularly on the GFC period and its aftermath. Fabling and Grimes’ (2005) study, examining the determinants of forced insolvency in New Zealand at national and regional levels, is for a sample period prior to the onset of the GFC.

2 https://www.insolvency.govt.nz/support/about/statistics.

3 A theoretical model justifying the assessment of macroeconomic variables can be found in Vlieghe (2001), who acknowledges his model as a stylised version of Wadhwani (1986) and in the style of Scott (1981).

4 Hall and McDermott (2016, Table 1).

5 We explored the possibility of joint modelling the breaks. Unfortunately, using a Markov-switching model for this task presents a couple of problems. First, the distributional assumptions required for count data are not readily applicable to the standard Markov models, but the more material problem is that the span of the data covers too few transitions. However, provided the database is maintained through another business cycle it would be very interesting to re-visit this question with a longer span of data.

6 It is not possible to start our sample period prior to July 2003, due to an observable discontinuity in company liquidations/insolvency data series around 2003. The lack of consistency between the former involuntary company liquidations series as utilised by Fabling and Grimes (2005) for the sample period 1988q1–2003q2, and the currently published corporate insolvency statistics would seem associated with: (1) New Zealand government Minister Lianne Dalziel's announcement of 18 February 2003 of projected Insolvency Law changes; and (2) the New Zealand Companies Amendment Act 2006, s 6, being subsequently inserted on 1 November 2007 into the Companies Act 1993 in Part 15A (Voluntary Administration).

7 https://www.insolvency.govt.nz/support/about/statistics/corporate-insolvency-statistics/monthly-its-administered-liquidations/. Insolvency data on this website are listed by the regional centre they are lodged in but we label them by region so as to align their terminology with other data we use in the paper.

8 For detail on these two aspects, see Fabling and Grimes (2005).

9 https://www.insolvency.govt.nz/support/about/statistics/insolvency-procedure-statistics/monthly-bankruptcy-figures/. Unfortunately, associated with the passing of the Insolvency Act 2006 implemented on 3 December 2007, and the administering and reporting on the sizeable numbers of accepted ‘no asset procedure’ insolvencies being passed to the Official Assignee, total personal insolvency numbers for our four regions are not published consistently for the full sample period.

10 The GFC is believed by many to have begun from July 2007 with the U.S. credit crunch, and by others to have related more directly to Lehman Brothers filing for Chapter 11 bankruptcy protection on 15 September 2008.

11 For example, insolvent companies which previously went into liquidation have since had the opportunity to trade out of difficulties through some form of rehabilitation regime, and the No Asset Procedure and Summary Instalment Order regime under the Insolvency Act then became an alternative to bankruptcy for personal low-income debtors with few or no realisable assets.

12 Hall and McDermott (2016, Table 1).

13 One drawback of the Poisson model is that it imposes the assumption of mean-variance equality which is clearly violated in the sample we have. It is possible to relax this assumption by using a negative binomial distribution at the cost of introducing another parameter to be estimated. The negative binomial would also be appropriate if insolvencies are contagious since contagious insolvencies have positive correlated occurrences causing larger variances than if the occurrences were independent. That said, the salient feature of the data is the sharp change in the mean rate of insolvency rather than inflated variances. Therefore, the best practical solution is to use the Poisson model in a quasi-likelihood setting and calculate the standard errors using robust methods.

14 This is not the case for corporate insolvencies in the Waikato, Wellington and Canterbury regions.

15 No breaks were detected in any region using the pre-GFC period.

16 This break period is consistent with Williams (2017a) having categorised the post-GFC period mid-2010 to late-2012 as ‘domestic caution and global uncertainty’. The following key events can then be noted as potentially contributing to the accompanying slowdown in economic activity and the subsequent somewhat increased insolvencies: the OCR increase of 50 basis points during June and July 2010; the deterioration in global sentiment over 2011 and 2012, and drought conditions during the summer of 2012/13.

17 As a robustness check we performed a Bai-Perron (2003) test for multiple breaks. The results are reported in Table A1 and provide supporting evidence of a break in the rate of insolvencies around 2008 (where the rate of insolvencies increased) and a further break between 2010 and 2014 (where the rate of insolvencies decreased). In contrast to our Bayesian Poisson model, the Bai-Perron tests find evidence of multiple breaks in Waikato, Wellington and Canterbury insolvency rates. These additional breaks are at the start of the GFC for Waikato and Canterbury and in 2011 for Wellington. The timing of these breaks is consistent with the break in other regions and so seems plausible. However, given the Bai-Perron test was not specifically set up for the type of application we are using and to avoid the temptation to over fit the model to every outlier we prefer to rely on the results from the Poisson model.

18 It can be noted that in Fabling and Grimes work, neither a terms of trade variable nor an exchange rate variable had a significant influence.

19 The previous U.K. and U.S. studies referred to above, and Fabling and Grimes (2005) were able to use the rate of total insolvencies as their dependent variable. In Fabling and Grimes (2005), this was because a series for the total number of companies registered by the Companies Office was available for the denominator. We have not found a similar readily available series for our sample period, so our results are restricted to those using the number of corporate insolvencies. Results reported by Fabling and Grimes are similar, whether the dependent variable is the number or the rate of insolvencies. It is further the case that disaggregating their number of total forced insolvencies variable so as to provide separate equations for the number of personal bankruptcies and the number of involuntary company liquidations provided very similar results.

20 Fabling and Grimes (2005) and Hall and McDermott (2007) were able to report results using the National Bank of New Zealand's quarterly measures of National and Regional Economic Activity.

21 Hess, Grimes, and Holmes (2009) have found a lagged bank credit expansion variable significant in explaining credit losses in Australasian banking; and Grimes and Hyland (2015) have used the ratio of non-performing loans to total assets of New Zealand registered banks as an exogenous indicator of supply-side credit restrictions to assist in explaining credit losses in Australasian banking.

22 The number of personal insolvencies is sufficiently large that they could be well approximated as a continuous variable in a standard linear model and thus estimated by OLS. However, such an approximation is not appropriate for the corporate insolvency data where the number of insolvencies per quarter can be small, especially in the regions. Since we need to use maximum likelihood for the corporate insolvencies we chose to use it for both data sets and avoid any approximations altogether.

23 To check for any misspecification, we examined the correlogram of the residuals and found no evidence of serial correlation. The lagged insolvencies variable is statistically significant for all regions except Wellington and it is this variable that is soaking up any possible serial correlation. Excluding the lagged dependent variable leads to serious serial correlation problems in the specification. However, the size of the coefficient is extremely small. For example, consider the case of New Zealand: for every 100 extra insolvencies in the previous quarter there is approximately one extra expected insolvency in the current quarter.

24 Fabling and Grimes found that economic activity, real private sector credit, CPI inflation, and at regional levels real property price inflation were all significant in influencing insolvency rates.