Boom and Bust. A Brief Analysis of the Financial Cycle and its Lessons

Table of Contents

List of Tables/Figures

1. Introduction

2. Theory of the Financial Cycle
2.1 Identification of the Financial Cycle: Measurement and Empirical Regularities
2.1.1 Indicators and Approaches
2.1.2 Basic Features and Characteristics
2.2 Financial Cycle and Business Cycle: Interactions

3. The Financial Cycle: Challenges and Implications
3.1 Theoretical Derivation and Reflections
3.2 Challenges, Warnings and Coping Strategies

4. Conclusions

References

Annex Table

Abstract

This paper mainly analyzes the implications of the financial cycle and its interactions with the traditional business cycle. Using frequency-based filter and turning-point analysis to measure duration, amplitude and evolution of the financial cycle we show that the results of both approaches for the financial cycle are similar and fit the actual dates well. Further, we also find that although financial and economic cycles are completely different, they are closely related.

The financial cycle significantly amplifies fluctuations in the real economy. The peaks of the financial cycle often indicate the build-up of the financial crisis, and the sharp adjustment on the peaks of the cycle will influence the length and depth of business recessions.

Keywords: financial cycle, business cycle, medium term, financial crisis, credit, asset price, real interest rates, monetary policy

List of Tables/Figures

Abbildung in dieser Leseprobe nicht enthalten

1. Introduction

Before the 20th century, Keynesian economic cycle theory based on IS-LM framework has been the main paradigm in the field of macroeconomic cycle research. Keynesians divided macroeconomics into two categories, namely, short-term fluctuation components and long-term growth trend, to discuss them separately. They argue that short-term (1-8 years) fluctuations in the real economy represented by real GDP are deviations from the long-term potential GDP trend. However, the two subsequent oil crises and the lack of explanatory power of stagflation made Keynesianism questioned by the neoclassical macroeconomic school represented by Lucas critique (Lucas, 1976). After the 1980s, the Real business cycle theory (RBC), which used exogenous impact of actual factors to explain the source of business cycle fluctuations, became the core of new classical macroeconomic theory and dominated the research on cycles for more than two decades (Kydland and Prescott (1982); Nelson and Plosser (1982)). However, both Keynesian economic cycle theory and real business cycle theory ignore the role of financial frictions in macroeconomics unwittingly. The outbreak of the financial crisis in 2007 just exposed the disconnection between the traditional economic cycle theory and the real macroeconomic development due to its emphasis on economic fluctuations and over-simplification of the financial market setting (Woodford (2003)).

Economic activity is indeed characterized by periodic fluctuations, with no permanent booms and no permanent recessions. There is a joke that 10 economists have 11 different views, so there are naturally different perspectives on the law of economic fluctuations. A review of the economic and financial development of the United States in the past 100 years shows that with the tightening of monetary policy, adjustment of regulatory system, expansion and contraction of credits and innovation of financial instruments, the financial market also presents cyclical fluctuations. In the United States, for example, there have been three relatively complete and large financial cycles. All three rounds of financial cycles were accompanied by severe economic fluctuations and profound economic recessions, corresponding to the Great Depression (1929-1933), the savings and loan crisis (1988-1989) and the subprime mortgage crisis (2007-2008). From the historical experience of the United States, the three financial cycles show us some near coincidental similarities. The first half of the financial cycle is often accompanied by financial deregulation, credit expansion and surge of asset prices. The peak of the financial cycle is often accompanied by asset price bubbles, banking crises and significant financial repression. The second half of the financial cycle is typically marked by rising interest rates, tighter regulation and a prolonged economic recovery.

The earliest research on financial cycles and economic cycles can be traced back to 1873. Bagehot introduced financial factors into the analytical framework of economic fluctuations. He believes that the growth of bank credit and rising asset prices will lead to asset bubbles. After the Great Depression, Fisher (1932, 1933) put forward the famous "debt-deflation theory", which systematically elaborated the continuous impact of corporate liability (credit) behavior on asset prices and business cycles. In the 1980s, economists such as Bernanke & Gertler (1989), Bernanke et al (1996) and Kiyotaki & Moore (1997,2002) proposed the theory of “financial accelerator” and Kiyotaki–Moore model of credit cycles on the basis of credit theory respectively, further emphasizing the crucial role of financial factors in the economic cycle. Since the early 1980s, especially after the financial liberalization and outbreak of the international financial crisis in 2007, the financial cycle has played an increasingly important role globally and attracted the attention of an increasing number of economists and policymakers. Based on the reflection of mainstream macroeconomics (e.g., new Keynesian and new classical models) and business cycle-oriented theories, the research on the mechanism of the financial cycle and its interactions with the economic cycle returned to the central position of economic research (e.g., Claessens et al (2011, 2012), Aikman et al (2013), Borio et al (2013), etc).¹

History does not repeat itself, but it always follows the same rhyme and blesses those who are willing to risk lifting its veil. Thus, the main purpose of this paper is to clarify the development and detect the inherent characteristics of financial cycles in the past decades from both theoretical and empirical aspects. On this basis, I try to give readers a more comprehensive and profound understanding of the financial cycle theory, so as to distinguish it from the traditional economic cycle theory readily. And, where possible, nest it into appropriate economic paradigms in a more suitable way. At the same time, based on the existing empirical data and analysis of previous financial shocks in various countries, I try to make a review of the rule-of-thumb policies and an outlook on future economic policy guidelines, especially the (accommodative) monetary policy and (expansionary) fiscal policy, which focuses on stimulation and stabilization of the real economy before the onset of and during financial crises, will be highlighted here.

Our theorical and empirical parts are principally inspired by Borio (2014) and Claessens et al (2011, 2012) respectively. The structure of the paper is organized as follows: The second part first introduces and explains the widely used methods and approaches to estimate the financial cycle, and then discusses the basic features and characteristics of the financial cycle grounded on two widely accepted methods of estimating cycles mentioned above and its interaction with the business cycle. Section three explains and discusses the main problems that economic researchers and economic policymakers could face under the background of financial cycle theory, especially the identification and the confirmation of financial crises and the effectiveness of policy instruments in the period of financial distress. The last section concludes.

2. Theory of the Financial Cycle

2.1 Identification of the Financial Cycle: Measurement and Empirical Regularities

2.1.1 Indicators and Approaches

According to the existing theories and literature, the definition of the financial cycle is still rough and relatively imprecise, and a consensus definition has not been reached. As a result, economists choose a variety of financial indicators when measuring the financial cycle. Our definition of the financial cycle here and below is mainly cited from Borio (2014). According to his explanation, he defines the financial cycle as self-reinforcing and procyclical interactive movements, whose peaks (booms) and troughs (busts) are sensitive to risk perceptions and external financial environment. In a nutshell, the financial cycle refers to the periodic fluctuations caused by the expansion and contraction of financial variables and is a boom-bust cycle caused by the interaction between risk perception and financing constraints. In the selection of variables, unlike the business cycle, which is typically approximated and estimated by real GDP, the main indicators for the financial cycle that widely used are (quarterly or seasonally adjusted) credit, credit-to-GDP ratio, real estate prices, equity prices, aggregate asset prices as well as their variations². Among them, the credit cycle and house price cycle are regarded as the most important proxy variables of the multivariate financial cycle (Claessens et al (2011, 2012), Borio (2014), Drehmann et al (2012)). First, credit is a determinant of the analysis of the financial cycle, because it bridges the connection between savings and investment. As the most traditional way of credit creation, credit closely links savings and investment, which makes it easier to measure the cyclical expansion and contraction of the financial sector. At the same time, considering the fact that most countries have financial intermediaries as the main body of financing structure, credit is undoubtedly the most intuitive indicator to measure the overall operation situation of the financial system. Therefore, it is not surprising that credit has proved to be one of the most effective variables in many literatures on measuring the financial cycle.³ Second, real estate assets are a kind of physical assets, and are often used as collateral for credit and the basic assets of financial innovation (asset securitization), not to mention that it was the trigger for the 2008 financial crisis. From the reality of various countries, the cycle in the housing market is not only closely related to the credit cycle and the economic cycle, but also the direct source of economic fluctuations as well as an important mechanism for the transmission of financial risks between the real economy and financial systems. Other financial variables such as the cycle in equity markets is excluded from the dominant factors of the financial cycle analysis because of its relatively high volatility and dissonant frequency with credit and house prices.

In terms of the empirical identification of the financial cycle, two approaches are widely used. One of the fascinating insights into this issue are materialized with the help of frequency-based filer analysis (e.g., Drehmann et al (2012), Alikan et al (2015)), which is based on the framework of Baxter and King (1999) and subsequent improvements to this method by Christiano and Fitzgerald (2003).⁴ To isolate the frequency properties of each proxy variable in the financial cycle, Drehmann et al (2012) calculate the ratio of the standard deviation of above-mentioned individual variable in the medium term to its own standard deviation in the short term. Specifically, the medium frequency of each variable has greater amplitude than its high frequency if:

Abbildung in dieser Leseprobe nicht enthalten

where i={credit,credit/GDP,house prices,aggregate asset prices,equity prices }; m and s are abbreviations of medium term and short term respectively. Further, to see whether each cycle changed differently over time, the observation period is then divided into two intervals bounded by 1985 with focus on the medium frequency.⁵ Analogously, standard deviation ratio could also be used here to trace the evolution of different cycles over time. All relevant results are recorded in Table 1 and Table 2 below, which will still be used in the following discussion.

Table 1: Relative Volatility of Short- and Medium-term Cycles

(Individual Series with Frequency-based Method)

Abbildung in dieser Leseprobe nicht enthalten

Source: Drehmann et al (2012) p.6

Notes: AAP=aggregate asset price; AU=Australia; DE=Germany; GB=United Kingdom; JP=Japan; NO=Norway; SE=Sweden; US=United States

The second approach is equipped with the turning point analysis (Burns and Mitchell (1946), Bry and Boschan (1971), Harding and Pagan (2002)).⁶ Like frequency-based analysis, several screening criteria need to be set manually here to obtain a series of maxima (peaks) and minima (troughs). First, (local) maxima and (local) minima over a given period of time should be identified. After that, the data need to be filtered according to the preset minimum length of the cycle. These obtained extreme points are then peaks and troughs when such conditions are met. For the financial cycle, the local maxima (minima) are usually defined as the peak (trough) within 9 quarters and the length of the financial cycle is set at a minimum threshold of 20 to 32 quarters from a practical and empirical point of view (Drehmann et al (2012)).⁷ According to Claessens et al (2011), the advantage of this approach is that it is robust, even when more new data are included. It also makes it possible and easy to compare between countries horizontally. Based on the aforementioned frequency-based analysis, Drehmann et al (2012) estimate each cycle mainly by median and peak-start-peak-end method.⁸ Similar analysis has also been conducted by Claessens et al (2011, 2012).⁹ Their results are plotted in Table 3 and Table 4 using the median of the distribution as the baseline.¹⁰

Table 2: Relative Volatility of Medium-term Cycles before and after 1985

(Individual Series with Frequency-based Method)

Abbildung in dieser Leseprobe nicht enthalten

Source: Drehmann et al (2012) p.7

2.1.2 Basic Features and Characteristics

By using the two commonly used descriptive statistical approaches described above, we find several interesting and crucial facts about the financial cycle.

First, as shown in the Table 1 above, the intermediate frequency components of all variables exceed their respective low frequency components (i.e., the ratio of standard deviation in the medium term to the short term one is greater than unity). In another words, their higher volatility in the medium frequencies suggests that cycles with lower frequency are better at shaping financial cycles. Among them, the standard deviation ratio of credit and credit-to-GDP ratio is especially large, which is almost one or two times as large as other variables. In stark contrast, the standard deviation ratio for equity prices is much smaller and is almost as much as that for GDP, which is a clear indicator for the business cycle. This finding shows that the medium-term components of each variables are way better to describe the financial cycle than that in the short-term approach. Especially, the credit cycle co-varies with the house price cycle. According to the results measured by the concordance index (Harding and Pagan (2002)),¹¹ The average value of this index between the credit cycle and the house price cycle amounts to the highest 73 percent (Drehmann et al (2012)), turning out that these two medium-term cycles are well synchronized and are suitable to identify a well-behaved financial cycle. Other variables, such as equity prices and aggregate asset prices, are excluded from the construction of the financial cycle because of their relatively high volatility and low concordance level with credit and house prices to some extent.

Second, the financial cycle tends to have a long duration and has gained power from its medium-term components over time. On average, the financial cycle has a length of about 64 quarters (16 years). By dividing the observation period into two parts, there is a clear signal that both the amplitude and the duration of the medium-term cycle has significantly increased, especially those for credit and house prices (Table 2 and Table 4). For example, the duration of the financial cycle has significantly increased from 11 years (pre-1985) to almost 20 years since 1985 (Annex Table A).

Third, the periodic length of each financial cycle is not uniform (Table 3), varying from the longest 72 quarters (credit) to the shortest 38 quarters (equity prices). In terms of the internal composition of each cycle, the financial recessions (downturns) last about 5 to 8 quarters, on the one hand; the financial recoveries (upturns) last usually 2 to 10 quarters longer than financial contractions, on the other hand. Moreover, the length and the intensity of the financial cycle change over time as well. Taking 1985 as the dividing line, the length of the financial recovery for all cycles after 1985 has become markedly shorter, although the duration of financial downturns has remained roughly the same. Besides, the amplitude of almost every financial variable during downturns and upturns is smaller after 1985,¹² ranging from -2.87 to -22.74 percent and 2.88 to 21.91 percent respectively (Annex Table B).

Table 3: Financial Cycles: Medium-term and Short-term (Turning-point Approach)

Abbildung in dieser Leseprobe nicht enthalten

Source: Drehmann et al (2012) p.9

Notes: Results are based on the median of the distribution of every individual series.

Fourth, there also exists a multiplier (synchronization) effect between different variables. Before refining our observation, two concepts are first introduced here: disruptions and booms (Claessens et al (2011)).¹³ Generally, the descriptive statistical data becomes more extreme both for disruptions and booms. For instance, it is harder and takes around 10 quarters longer for economic to recover from severe house market disruptions. Further, the joint analysis of synchronized downturns and upturns of different cycles confirms our previous findings that the adverse shocks from the side of credit cycles (financial market) or house price cycles (real economy) may amplify each other and lead to a downward spiral. This discovery, to some extent, has proved the correctness of the financial accelerator theory that is put forward by Bernanke and Gertler (1989). Additionally, this amplifying effect can also be seen across countries and becomes severer over time. While the duration of joint downturns remains almost unchanged, the price drop is nearly twice as large as it would have been without the effect.

Table 4: Characteristics of individual cycles in the medium term (pre- and post-1985) (Turning-point approach)

Abbildung in dieser Leseprobe nicht enthalten

Source: Drehmann et al (2012) p.10

The last point worth mentioning here is that the duration and the amplitude of the financial downturns are statistically influenced by several other factors (Annex Table C). By applying regressions and the Weibull distribution model, Claessens et al (2011) confirm that disruptions in the house price cycles are positive correlated with the duration and the amplitude of downturns in credit cycles and negative correlated with that in equity price cycles.¹⁴ To be specific, the credit crunch lasts longer and is more severe in the presence of house price bust, and the downturns in the equity market seems to be shortened and eased when house price bust is included. Beyond that, cross-country factors such like high level of trade openness and financial openness could also help to pull economies out of financial traps quickly (Annex Table C and Table D).

Similar analysis has been carried out by Strohsal et al (2015) using frequency domain approach instead of more descriptive methods conducted by Claessens et al (2011) and Drehmann et al (2012). By focusing on Germany, United Kingdom and the United States, they suggest that the financial cycle dose evolve and becomes increasingly important in the light of its variance contribution.¹⁵

Table 5: Financial Cycles (Turning-point Approach)

Abbildung in dieser Leseprobe nicht enthalten

Source: Claessens et al (2012) p.182

Notes: ***= significance at 1% level; **= significance at 5% level; *= significance at 10% level

[...]

¹ For more recent studies, see e.g. Juselius et al (2017) and Kunovac et al (2018).

² These variables are credit quality, credit composition, default rates, financial leverage, non-performing loan ratio and so on.

³ In addition, Drehmann et al (2012) suggest choosing credit to the non-financial private sector as the proxy variable of credit, because this indicator can better reflect the connection between the financial sector and the real economy.

⁴ As the naming convention suggests, a specified frequency range should be preset. This may lead to potential bias due to subjective settings.

⁵ The dating here is not arbitrary, because the main thing we want to examine here is whether financial liberalization has a direct impact on the financial cycle, and the mid-1980s happened to be the first phase of the global financial liberalization.

⁶ This approach was first designed by Burns and Mitchell (1946) to detect the business cycle. Later, Bry and Boschan (1971) and Harding and Pagan (2002) proposed algorithms to process monthly and quarterly data separately (BB algorithm and BBQ algorithm respectively).

⁷ By contrast, the lower bound of the length of the business cycle is limited to five quarters and the local maxima (minima) is also defined as the highest (lowest) point within a five-quarter interval.

⁸ Their analysis includes seven major developed countries from 1960 to 2011.

⁹ Claessens et al (2011, 2012) use quarterly adjusted time series data to examine the evolution of the financial cycle in 21 OECD countries from 1960:1 to 2007:4.

¹⁰ Since the median is at the center of the distribution of all data and reflects the central tendency, even if the distribution is not symmetrical, this method avoids several errors caused by the mean method, such as the possibility of bias if extreme values or skewed data occur in the whole sample.

¹¹ The concordance index is defined as follows: where x and y stand for candidate variables, t represents the time period from 1 to T. and take the value 0 or 1 relying on their location on the cycle. Two cycles could be treated as fully dependent and synchronized if the concordance index is equal to 100 percent.

¹² The only exception is equity price cycle. The amplitude of its upturns has even increased from 19.09 to 21.91 percent. This seems to be consistent with the fact that stock market is more volatile after financial liberalization.

¹³ The financial disruptions (booms) are defined as the bottom (top) 25 percent of the downturn (upturn) range.

¹⁴ Credit crunches, house price busts and equity price busts appear in regression models in the form of dummy variables.

¹⁵ However, the detection of the financial cycle in Germany is to some extent fuzzy compared to that in the US and United Kingdom.