reexamining risk-return relationship in banks using quantile regression

32
Reexamining risk-return rel ationship in banks using qu antile regression By Ming-Yuan Leon Li Service Industries Journal (a SSCI journal), accepted and forthcoming in 2010

Upload: briana

Post on 13-Jan-2016

48 views

Category:

Documents


1 download

DESCRIPTION

Reexamining risk-return relationship in banks using quantile regression. By Ming-Yuan Leon Li Service Industries Journal (a SSCI journal), accepted and forthcoming in 2010. Motivations. banks make money in one of two ways: providing services to customers and taking risks. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Reexamining risk-return relationship in banks using quantile regression

Reexamining risk-return relationship in banks using quantile regressi

on

By Ming-Yuan Leon Li

Service Industries Journal (a SSCI journal), accepted and forthcoming in 2010

Page 2: Reexamining risk-return relationship in banks using quantile regression
Page 3: Reexamining risk-return relationship in banks using quantile regression

Motivations

• banks make money in one of two ways: providing services to customers and taking risks.

• if a bank takes more risk it can expect to make more money

• empirical finding regarding the risk-return relationship is controversial and it has posed a longstanding problem in research field.

Page 4: Reexamining risk-return relationship in banks using quantile regression

Motivations

• a positive risk-return relationship – Pettway, 1976; Kim, 1978; Jahankhani and Lynge, 19

80; Schneller, 1980; Bradley, et al., 1984; Brewer and Lee, 1986; Karels et al., 1989; Hassan and Bashir, 2003) show a positive risk-return relationship

• a negative relationship between risk and bank performance.– Bourke (1989), Molyneux and Thornton (1992), Canto

r and Johnson (1992), Berger (1995), Golin (2001) and Goddard et al. (2004)

Page 5: Reexamining risk-return relationship in banks using quantile regression

Motivations

• This investigation departs from the more conventional research in the way that the parameters of the risk-return regression are modeled and proposes a new approach to questions regarding the relationship dynamics between risk and bank performance.

Page 6: Reexamining risk-return relationship in banks using quantile regression

Motivations

• this study examines whether the risk-return relationship in bank industry is consistent with different levels of bank profitability quantile.

• the quantile is a statistical term describing a division of observations into certain defined intervals based upon the values of the data

• the profitability quartile of a specific bank could show the relative magnitude of its profitability in comparison with the entire set of bank observations .

Page 7: Reexamining risk-return relationship in banks using quantile regression

Motivations

• Our idea – Adizes (1988) proposes the firm life cycle the

ory: business strategies and organizational structures of firms vary according to the problems faced at different life cycle stages of the organization.

• The concept of life-cycle stage has generated considerable applied interest

Page 8: Reexamining risk-return relationship in banks using quantile regression

Motivations

• To analyze the effect of life-cycle stage on profitability, these aforementioned studies invariably apply criteria such as earnings and/or age to segment sample companies into various subsets before performing traditional optimization techniques such as ordinary least squares (OLS) and least absolute deviation (LAD) to fit their subsets.

Page 9: Reexamining risk-return relationship in banks using quantile regression

Motivations

• The analytical framework in these studies is based on unconditional distribution of firm samples.

• The findings of the current study suggest that this form of “truncation of samples” may yield invalid results.

• As demonstrated by Heckman (1979), such methods often exhibit sample selection bias.

Page 10: Reexamining risk-return relationship in banks using quantile regression

Motivations

• A valid alternative is the quantile regression (QR hereafter), which segments the sample into subsets defined by conditioning covariates.

• Moreover, in comparison with the least square method, the QR approach offers a relatively rich description of the conditional mean for extreme cases in the samples

Page 11: Reexamining risk-return relationship in banks using quantile regression

Motivations

• In addition, we posit that the behavior of banks with higher profitability differs from banks with lower profitability.

• First, banks at the growth (decline) stage tend to exhibit higher (lower) profitability.

• Further, according to lifecycle theory, profitable banks should differ from profitless banks in their strategies for enhancing profitability.

Page 12: Reexamining risk-return relationship in banks using quantile regression

Motivations

• this investigation departs from previous related studies in proposing the QR framework to questions regarding the risk-return relationships in banks.

• the empirical results of this study could satisfactorily account for the existing risk-return relationship puzzle among numerous prior studies.

Page 13: Reexamining risk-return relationship in banks using quantile regression

Empirical methods: Non-quantile models: OLS and LAD

iii uxy '

2'2 )(11min ii

ii

i xyu

||1||1min ' ii

ii

i xyu

Page 14: Reexamining risk-return relationship in banks using quantile regression

• One key limitation of OLS and LAD estimates is that they provide only one measure of the central distribution tendency of yi, namely, profitability performance and tail behaviors are not considered.

Page 15: Reexamining risk-return relationship in banks using quantile regression

Empirical methods: Non-quantile models: QR (quantile regression)

model

0)(

)(:inf)( '

'

ii

iiii

iii

xuQuant

xxyFyxyQuant

uxy

Page 16: Reexamining risk-return relationship in banks using quantile regression

0: 0:

''

: :

' '

0 0

||)1(||

)1(||min

ii ii

i bi

xyi xyiiiii

ui uiii

xyxy

uu

Page 17: Reexamining risk-return relationship in banks using quantile regression

• a key feature of the quantile regression technique: the estimator vector of β,θ varies with θ.

• Moreover, by comparing the behaviors with different θ, one could thus characterize the dynamic estimator vector, namely β,θ, in various output-quantile regimes.

• the LAD estimator is a special case of the quantile-varying estimator with a quantile of 0.5.

Page 18: Reexamining risk-return relationship in banks using quantile regression

Data and empirical results

• Samples for publicly-traded U.S. bank holding companies (BHCs hereafter) from March 2001 to September 2007 are analyzed.

• The final sample includes 18,108 quarterly BHS observations.

Page 19: Reexamining risk-return relationship in banks using quantile regression

• Return on equity (RoE) is selected as the proxy variable for bank profitability performance.

• employ loan loss reserves to gross loans ratio (LLRGL) as a proxy variable for the risk taken by a bank – Mansur, et al. (1993), Hassan (1993), Stiroh (2006), H

irtle (2007)

• All data are obtained from the U.S. Federal Reserve Y-9C reports

Page 20: Reexamining risk-return relationship in banks using quantile regression

Table 1 Definition of dependent/independent variables

Variables Definitions

Dependent variable

RoE Return on equity= Net income/ shareholders' Equity

Independent variables

LLRGL The ratio of loan-loss-reserves-to-gross-loans

Tables

Page 21: Reexamining risk-return relationship in banks using quantile regression

Table 2 Descriptive statistics of dependent/independent variables

Variables Mean Standard error Median Minimum Maximum

RoE 0.0808 0.0592 0.0725 -1.3029 0.7027

LLRGL 0.0136 0.0086 0.0127 0.0000 0.2572

Page 22: Reexamining risk-return relationship in banks using quantile regression

Table 3 Impact of LLRGL on bank profitability (RoE) across various quantile levels

Estimation results of the QR and OLS methods Statistic tests of the equality of slope

estimates across various quantiles

Quantile Estimate (p-value) Quantile Estimate (p-value)

Intercept Slope Intercept Slope Quantile F-statistics (p-value)

0.05 0.0290 (0.0000)*** -0.7218 (0.0000)*** 0.95 0.1543 (0.0000)*** 1.5344 (0.0000)*** 0.05 vs. 0.95 64.63 (0.0000)***

0.10 0.0309 (0.0000)*** -0.3406 (0.0000)*** 0.90 0.1315 (0.0000)*** 1.1856 (0.0000)*** 0.10 vs. 0.90 73.46 (0.0000)***

0.15 0.0354 (0.0000)*** -0.3121 (0.0010)*** 0.85 0.1210 (0.0000)*** 0.8426 (0.0000)*** 0.15 vs. 0.85 62.23 (0.0000)***

0.20 0.0396 (0.0000)*** -0.2823 (0.0110)** 0.80 0.1114 (0.0000)*** 0.6648 (0.0000)*** 0.20 vs. 0.80 37.9 (0.0000)***

0.25 0.0444 (0.0000)*** -0.2519 (0.0450)** 0.75 0.1025 (0.0000)*** 0.5721 (0.0010)*** 0.25 vs. 0.75 31.73 (0.0000)***

0.30 0.0494 (0.0000)*** -0.2049 (0.1680) 0.70 0.0967 (0.0000)*** 0.3613 (0.0230)** 0.30 vs. 0.70 16.62 (0.0000)***

0.35 0.0548 (0.0000)*** -0.0848 (0.6080) 0.65 0.0893 (0.0000)*** 0.3202 (0.0830)* 0.35 vs. 0.65 7.86 (0.0050)***

0.40 0.0611 (0.0000)*** -0.0707 (0.6770) 0.60 0.0826 (0.0000)*** 0.2832 (0.0710)* 0.40 vs. 0.60 11.51 (0.0007)***

0.45 0.0657 (0.0000)*** 0.0459 (0.7440) 0.55 0.0762 (0.0000)*** 0.2303 (0.1720) 0.45 vs. 0.55 4.36 (0.0368)**

0.50 0.0716 (0.0000)*** 0.0669 (0.6950) OLS 0.0750 (0.0000)*** 0.4237 (0.0000)***

Page 23: Reexamining risk-return relationship in banks using quantile regression

-1.50

-1.00

-0.50

0.00

0.50

1.00

1.50

2.00

2.50

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

Page 24: Reexamining risk-return relationship in banks using quantile regression
Page 25: Reexamining risk-return relationship in banks using quantile regression

Some explanations

• In theoretical, a bank taking a relatively high risk is supposed to earn high profits, but is also exposed to certain costs; therefore its profitability might be reduced

• In particular, bankruptcy costs may be relatively high for a bank maintaining higher risk exposure.

• A subsequent increase in risk taking should lead to a decrease in profitability by heightening insurance expenses on uninsured debt.

Page 26: Reexamining risk-return relationship in banks using quantile regression

• Our empirical findings show that highly profitable banks can increase their profitability by taking more risks

• by contrast, the superior policy in less profitable banks is to decrease rather than increase their risk exposures

Page 27: Reexamining risk-return relationship in banks using quantile regression

• According to the life-cycle theory, a bank in the growth (declining) stage is denoted by a higher (lower) profitability.

• Moreover, this analysis further reveals that, in banks in a growth stage, namely, banks at higher quantile levels: 0.60 to 0.95, a significantly positive risk-return relationship is defined.

• Therefore, risk exposures positively affect profitability.

Page 28: Reexamining risk-return relationship in banks using quantile regression

• In contrast, in banks at lower quantile levels, from 0.25 to 0.05, or banks in a declining stage

• increased risk exposure has a minimal effect on marginal profit enhancement; thus, the potential bankruptcy cost effect of risk taking would be dominant and would negatively affect profitability.

Page 29: Reexamining risk-return relationship in banks using quantile regression

• Last, in banks at moderate quantiles, from 0.30 to 0.55, or banks in a mature stage,

• the profit enhancement effect derived from risk taking might be fully offset by its cost.

• Consequently, the effect of risk exposures on profitability would be insignificant.

Page 30: Reexamining risk-return relationship in banks using quantile regression

• the existing risk-return puzzle among earlier studies could be satisfactorily accounted for our “V” shape relationship between risk and profitability, as shown in Fig. 2.

• We further indicate that pooling data together without considering the impact of bank life cycle is one of the reasons for the inconsistent bank risk-return relationship findings presented by prior empirical works.

Page 31: Reexamining risk-return relationship in banks using quantile regression

Other applications of QR models

• Li, Ming-Yuan Leon* (2008) Reexamining asymmetric effects of monetary and government spending policies on economic growth using quantile regression, Journal of Developing Areas, accepted and forthcoming 【 SSCI】

Page 32: Reexamining risk-return relationship in banks using quantile regression

• Value Strategy or Volume Strategy? A Dynamic Perspective Using Quantile Regression, submitted to Journal of Empirical Finance 【 SSCI】

• Reexamining the dynamic relationships between capital, size and earnings in banking using the quantile regression model, submitted to Journal of Money, Credit and Banking 【 SSCI】

• Establishing A Hybrid Bankruptcy Prediction Model with Dynamic Loadings Based on Accounting-ratio-based and Market-based Information Using Binary Quantile Regression, submitted to Journal of Financial Services Research 【 SSCI】