The Limits to Credit Growth: Mitigation Policies and Macroprudential Regulations to Foster Macrofinancial Stability and Sustainable Debt

Abstract

In this paper we study an economy with a high degree of financialization in which (non-financial) firms need loans from commercial banks to finance production, service debt, and make long-term investments. Along the business cycle, the economy follows a Minsky base cycle with firms traversing through the various stages of financial fragility, i.e. hedge, speculative and Ponzi finance (cf., Minsky in The financial instability hypothesis: a restatement. Hyman P Minsky archive paper, vol 180, pp 541–552, 1978; Stabilizing an unstable economy. Yale University Press, 2nd edn 2008, McGraw-Hill, New York, 1986; The financial instability hypothesis. Economics working paper archive wp74. The Jerome Levy Economics Institute of Bard College, 1992). In the speculative financial stage cash flows are insufficient to finance the repayment of principle but sufficient for paying interest, so banks are willing to roll-over credits in order to prevent loan defaults. In the Ponzi financial position even interest payments cannot be served, but banks my still be willing to keep firms alive through “extend and pretend” loans, also known as zombie-lending (Caballero et al. in Am Econ Rev 98(5):1943–1977, 2008). This lending behavior may cause credit bubbles with increasing leverage ratios. Empirical evidence suggests that recessions following such leveraging booms are more severe and can be associated to higher economic costs (Jordà et al. in J Money Credit Bank 45(s2):3–28, 2013; Schularick and Taylor in Am Econ Rev 102(2):1029–1061, 2012). We study macroprudential regulations aimed at: (i) the prevention and mitigation of credit bubbles, (ii) ensuring macro-financial stability, and (iii) limiting the ability of banks to create unsustainable debt bubbles. Our results show that limiting the credit growth by using a non-risk-weighted capital ratio has slightly positive effects, while using loan eligibility criteria such as cutting off funding to all financially unsound firms (speculative and Ponzi) has strong positive effects.

Appendices

Appendix A: Business Cycle Dating Algorithm and Recession Analysis

The algorithm that was used to obtain the results for this paper is based on well-established methods from the empirical literature to study macroeconomic time series data. A classic reference to business cycle dating algorithms is the original BB algorithm developed by Bry and Boschan (1971). A quarterly Bry-Boschan algorithm, known as the BBQ-algorithm, was proposed by Harding and Pagan (2002). We adopt a similar methodology to time series data analysis as in Claessens et al. (2012). The only difference is that we use the synthetic data being generated by our simulation model, while normally such algorithms are used on empirical data.²⁴

1.1 Terminology and Definitions

The following definitions are taken from Claessens et al. (2012, pp. 10–12). The definitions can either be based on the time series of the units of output produced, or on actual sales levels. In our time series analysis, we have used the output-based definitions.²⁵

Peaks and troughs A peak in a timeseries $y_t$ occurs at time t if there are 2 periods of increase before, and 2 periods of decrease after t:

$$\begin{aligned} (y_t - y_{t-2}>0 , y_t - y_{t-1}>0) \text { and } (y_{t+2} - y_t<0 , y_{t+1}-y_t<0) \end{aligned}$$

(20)

A trough in a timeseries $y_t$ occurs at time t if there are 2 periods of decrease before, and 2 periods of increase after t:

$$\begin{aligned} (y_t-y_{t-2}<0 , y_t-y_{t-1}<0) \text { and } (y_{t+2}-y_t>0 , y_{t+1}-y_t>0)) \end{aligned}$$

(21)

Recession A recession/downturn is the period between a peak a trough.

Expansion An expansion/upturn is the period between a trough and a peak.

Recovery A recovery is the early part of the expansion phase, defined as the time it takes for output to rebound from the trough to the peak level before the recession.

Duration of recessions The duration of a recession/downturn is the number of quarters, k, between a peak ($y_0$) and the next trough ($y_k$) of a variable.

Duration of recoveries The duration of a recovery/upturn is the number of quarters (r) it takes for a variable to reach its previous peak level after the trough: $\{r>k: y_r \ge y_0\}$.

Amplitude for recessions The amplitude of a recession/downturn $A_c$, measures the change in $y_t$ from a peak ($y_0$) to the next trough ($y_k$): $A_c = y_k - y_0$

Amplitude for recoveries The amplitude of a recovery/upturn, $A_u$, measures the change in $y_t$ from a trough to the level reached in the first four quarters of an expansion ($y_{k+4}$): $A_u = y_{k+4} - y_k$.

Slope for recessions The slope of a recession/downturn is the ratio of the amplitude to the duration of the recession/downturn: $S_c = A_c/D_c$.

Slope for recoveries The slope of a recovery/upturn is the ratio of the change of a variable from the trough to the quarter at which it attains its last peak divided by the duration: $S_r = (y_r - y_0)/D_u$. The slope is an approximation of the average economic growth rate over the duration of the recovery.

Cumulative loss for recessions The cumulative loss for a recession with duration k combines the duration and amplitude as a measure for the overall costs of recession: $F^c = \sum _{j=1}^k (y_j - y_0) - A^c/2$, where $y_0$ is the level of output at the start of the recession, and $y_j$ are the successive terms during the recession.

1.2 Detecting Peaks and Troughs

Figure 7 shows the detection of peaks and troughs in the time series of output for the business cycle (panel a) and for the time series of total debt for the financial cycle (panel c). Figure 7b shows expansions and recessions from peak to trough for the business cycle. This plot does not coincide exactly with the peaks and troughs detected in Fig. 7a due to the fact that sometimes two peaks can follow each other without having a trough in the middle. This is because the trough does not necessarily signal a recession, since it might be too short. In such cases the event is censored, i.e. removed from the plot. Figure 7c, d provides the same type of analysis for the credit cycle. Here the solid lines coincide with peaks in the credit cycle, i.e. with the start of a downturn. Dotted lines indicate troughs in the credit cycle, i.e. the start of an upturn or recovery.

Fig. 7

Peaks and troughs for the business cycle and the financial cycle, for 500 months (167 quarters). Solid lines peaks, or start of a recession; dotted lines: troughs, or start of an expansion. a, c Detection of peaks and troughs. b, d Recessions and expansions (for the business cycle), and upturns and downturns (for the financial cycle)

Fig. 8

Boxplots of a recession amplitudes and b cumulative economic losses. Parameter settings: $(\alpha ,\beta )=\{1,8\}\times \{0,0.1,1\}$. The scenarios A till G for the default setting $(\alpha =8$, $\beta =0.1)$ correspond to sets 29–35. Scenario I for $(\alpha =1$, $\beta =0)$ corresponds to set 1 and scenario H for $(\alpha =1$, $\beta =1)$ corresponds to set 15. Downturns less than 500 units are censored

Appendix B: Robustness Analysis

In this section we provide a robustness analysis of the results presented in the main text. As parameter settings we consider all combinations for $\alpha =(0,8.0)$ and $\beta =(0,0.1,1)$, yielding 6 cases. Combining this with the scenarios A through G (7 scenarios), this yields 42 scenarios in total.

Table 3 Summary statistics for recession data across 100 simulation runs

Table 4 Summary statistics for recession amplitudes in 42 scenarios

Table 5 Summary statistics for recession cumulative losses in 42 scenarios

As statistics for the analysis we consider the amplitude and cumulative loss as metrics for recession severity. Figure 8 shows boxplots for the recession amplitudes and cumulative losses for each of the 42 scenarios. In Table 3 we report the extreme value of the lower whisker of the distribution of each metric. For the recession duration we report the median values. All distributions are based on 100 simulation runs and the full ensemble of recessions observed during the runs is considered. The minimal required recession duration is set to two quarters, and the minimal required amplitude is set to 500 units. Recessions with downturns smaller than 500 units are censored from the distribution.

The scenarios A to G from the main text correspond to group 5 (rows $29-35$). Scenario I $(\alpha =1$, $\beta =0)$ corresponds to row 1 in group 1, and scenario H $(\alpha =1$, $\beta =1)$ corresponds to row 1 in group 3. Table 4 shows descriptive statistics for the recession amplitudes, while Table 5 shows the descriptive statistics for cumulative losses. In each table the overall best policy within each group is indicated with a star ($\star$), while a dagger ($\dagger$) indicates the best single policy (from A, B, C, D, E). The ranking criterion for Table 3 is the smallest absolute value of the lower whisker, for Table 4 it is the smallest absolute value of the minimum amplitude, and for Table 5 it is the smallest absolute value of the cumulative loss.

Three main conclusions can be drawn from this robustness analysis. First, Table 3 shows that the ranking is rather robust between groups. The ranking we obtain for the default parameter setting (group 5) is maintained for three other parameter settings, including scenario I (groups 1, 2 and 4). For these cases, the overall best scenario is F (a mixed scenario), followed by the best single policy D (cutting off funding to all financially unhealthy firms). Second, for the remaining two parameter settings (group 3 and 6) the overall best scenario is the single policy C (using non-risk weighted capital ratios). The latter two groups have in common that they feature the full reserve requirement ($\beta =1$). Third, if we consider the cumulative loss as the ranking criterion, the overall best scenario for most parameter settings is again policy D (cutting off funding to all financially unhealthy firms). The only groups for which this is not the case are groups 3 and 6, for which A is the overall best policy (no intervention).

Therefore, the main conclusion to draw from this extensive robustness analysis is that cutting off the funding to financially unhealthy firms and using a non-risk-weighted capital ratio are the two most promising candidates for reducing the amplitude and cumulative losses due to recessions.