Journal Issue
Share
Article

The Impact of Oil Prices on the Banking System in the GCC

Author(s):
Padamja Khandelwal, Ken Miyajima, and Andre Santos
Published Date:
August 2016
Share
  • ShareShare
Show Summary Details

I. Introduction

Oil prices have rebounded vigorously from their trough in early-2016 and are expected to stage a more moderate rally in the next two years (Figure 1). The WTI spot price fell close to $30 a barrel in February 2016 (monthly average data), losing nearly ¾ of its value since June 2014. However, by mid-May, they have rebounded by more than 50 percent. Pricing of futures contracts suggests that oil prices are expected to recover at a more measured pace, by 5-10 percent over the coming two years.

Figure 1.WTI Oil Prices (U$ a barrel, monthly average)

Sources: Haver and IMF staff calculations.

Note: Actual through end-April 2016. Red broken line based on 3, 6, 12, 24 month futures prices, close of May 12.

Generally, the GCC economies can be adversely affected by low oil prices for at least two reasons:

  • First, the GCC economies are highly dependent on oil and gas exports. During 2011–14, hydrocarbon exports represented about 70 percent of exports of goods and services on average (Table 1). Fiscal dependence on hydrocarbon revenues was even greater, accounting for over 80 percent of total fiscal revenues on average. Over the past decade and a half, the dependence on hydrocarbon fiscal revenues did not decline despite efforts at economic diversification.
  • Second, macro-financial linkages in the GCC can amplify the effects of oil price movements over the financial cycle. Oil price movements and government spending policies create feedback loops between asset prices and credit that can lead to the buildup of systemic vulnerabilities in the financial sector. Oil price upturns lead to higher oil revenues, and stronger fiscal and external positions. Equity market returns are larger as investors anticipate the impact of higher oil prices on the corporate sector, and generally stronger government spending growth. In turn, stronger government spending leads to higher non-oil output growth, greater banking sector liquidity and credit growth, higher real estate prices, and stronger bank balance sheets. Asset price appreciation also has positive wealth effects. In the event of an oil price downturn, these developments can reverse. With financial sectors being fairly large in the GCC (Table 2), and oil prices being highly volatile, the unraveling of systemic financial sector vulnerabilities could have significant adverse effects on the real economy.
Table 1.GCC: Dependence on Oil Revenues
2000–052006–102011–14
Oil export revenues as percent of total exports of goods and services
Bahrain58.760.565.1
Kuwait82.780.587.6
Oman76.969.464.3
Qatar88.585.988.9
Saudi Arabia83.483.183.0
United Arab Emirates45.038.732.6
Fiscal oil revenues as percent of total fiscal revenues
Bahrain71.782.287.2
Kuwait72.779.283.6
Oman83.483.488.7
Qatar90.588.390.7
Saudi Arabia82.888.390.3
United Arab Emirates60.265.169.9
Source: IMF staff calculations.Note: Fiscal data is of the general government for UAE and central government in other GCC countries
Source: IMF staff calculations.Note: Fiscal data is of the general government for UAE and central government in other GCC countries
Table 2.GCC: Size of the Financial System(Percent of GDP)
Bank AssetsMarket Capitalization
Bahrain259.764.1
Kuwait165.572.8
Oman121.428.1
Qatar148.773.7
Saudi Arabia92.671.0
United Arab Emirates193.459.7
Source: GCC authorities, BloombergNote: United Arab Emirates market capitalization is sum of Abu Dhabi and Dubai stock markets; Market capitalization data is as of Sept. 16, 2015. Banking sector data for Bahrain excludes wholesale banks.
Source: GCC authorities, BloombergNote: United Arab Emirates market capitalization is sum of Abu Dhabi and Dubai stock markets; Market capitalization data is as of Sept. 16, 2015. Banking sector data for Bahrain excludes wholesale banks.

As an important reminder, systemic financial sector risks rose in the GCC countries with the oil price upswing in the years before the global financial crisis. An expanding deposit base and high liquidity (owing to high oil prices and short-term capital inflows) resulted in credit and asset-price booms before the global financial crisis.2 Bologna and Prasad (2010) document a sharp increase in household leverage between 2004 and 2008 in Oman. Al-Hassan and others (2010) suggest that the bursting of a domestic real estate bubble and tightening global liquidity conditions played a role in the United Arab Emirates’ 2009 financial crisis, while defaults in 2008 by two of the largest investment companies in Kuwait imposed strains on the banking system, with the third-largest bank having to be recapitalized. IMF (2010) discusses Qatar’s preemptive recapitalization of banks and other measures to support the banking sector in 2009. As the global financial crisis hit, asset prices and credit declined in several GCC countries, although fiscal stimulus and liquidity support helped cushion the impact.

More generally, evidence suggests that oil price performance has been an important driver of business and financial variables in the GCC economies.3 First, stronger performance of real and financial variables tends to be associated with oil price upturns. For instance, during 1991–2014, the growth rates of real government spending and non-oil GDP were much stronger during oil price upturns than during downturns. Second, the timing of downturns in business and financial variables in some cases coincides with that of oil prices. Contractions in credit and equity markets reflected oil price movements, along with global financial market developments and the underlying domestic vulnerabilities. Importantly, contractions in real government spending occurred as often as real oil price downturns in the 1990s, but less so since 2000, likely aided by greater fiscal buffers.

Strong banking sector soundness provides an important buffer in the GCC to the oil price decline since mid-2014. GCC banks have strong capital and liquidity buffers as of end-2014. Capital buffers and provisioning levels were above those in many other commodity exporting countries. NPL ratios are low and both loan loss provisions and profits are strong. In fact, provisions fully cover NPLs, on average. Strong macroeconomic performance helped, so did strengthened regulatory frameworks and improved risk management. However, liquidity conditions have started to tighten more recently. While credit growth has remained robust, deposit growth has slowed, largely as governments and government-related entities have withdrawn deposits from the banking system (Figure 2). Interbank rates have edged higher since the beginning of the summer of 2015.

Figure 2.Growth Rates of Bank Credit and Deposits in the GCC Countries

(Year on year, percent)

Sources: Haver and IMF staff calculations.

Against this backdrop, this paper addresses three interlinked issues: (i) how oil prices affect NPLs in the GCC; (ii) the links between oil prices and real and financial developments in these countries; and (iii) observed behavior of bank capital and provisioning with respect to indicators of the business and financial cycles. Section II empirically models bank asset quality relying on a range of multivariate approaches. It also employs a panel vector autoregression (VAR) approach to identify oil-macro-financial linkages. Section III empirically assesses the observed behavior of bank capital and provisions with respect to indicators of business and financial cycles. Section IV concludes and provides policy recommendations. The analyses in this paper generally consider macroeconomic, banking, and financial market data up to end-December 2014.

II. Econometric Analysis

Two related empirical models of oil-macro-financial linkages are estimated for the GCC economies. A brief review of the literature and discussion of data are followed by estimation of (i) a multivariate model to assess the impact of macroeconomic outcomes on bank asset quality (NPLs), and (ii) a panel VAR model to investigate oil-macro-financial linkages.

A. Literature

Empirical studies on bank asset quality and macro-financial linkages have expanded to include those of the Middle East and North Africa region. In the area of asset quality assessment, Nkusu (2011) analyzes country-level data spanning 1998–2009 for 26 advanced economies and confirms that adverse macroeconomic developments are associated with rising NPLs. De Bock and Demyantes (2012) use country-level annual observations for 25 emerging markets during 1996–2010 and find that the NPL ratio increases when economic growth declines, the exchange rate weakens, the terms of trade deteriorate and debt-creating capital inflows decline. Klein (2013) uses data spanning 1998–2011 for the ten largest banks in 16 countries (thus a total of 160 banks) in Central, Eastern and South-Eastern Europe and finds that macroeconomic conditions are relatively more important than banks specific factors in explaining NPLs.

Espinoza and Prasad (2010) represent the first attempt to model NPLs in the GCC countries using both macroeconomic and bank-level data. Using data spanning 1995–2008 for about 80 banks in the GCC region, the authors find that the NPL ratio rises as economic growth declines and both interest rates and risk aversion increase. Love and Ariss (2014) analyze a panel of Egyptian banks over 1993–2010 and find that larger capital inflows and stronger GDP growth improve bank loan portfolio quality. Many of the studies, but not all, also investigate macro-financial linkages. Earlier work relies on macro-level data (Nkusu (2011), De Bock and Demyantes (2012), Klein (2013)). Espinosa and Prasad (2010) study the GCC economies and find a strong, albeit short-lived feedback effect from weaker bank balance sheet conditions to economic activity. Studies applying a panel VAR approach to bank-level data are emerging. Recent studies focusing on the Middle East and North Africa include Love and Ariss (2014) for Egypt and Miyajima (2016) for Saudi Arabia.

B. Data

Our empirical analyses rely on macroeconomic and bank-level data which cover the six GCC economies and span 1999–2014 (Figure 3).

Figure 3.Macroeconomic and Bank Level Variables

Sources: Bankscope, Haver, and authors’ calculations.

Note: The green bands represent a 50 percent range around the median value

The macroeconomic and financial sector variables included are real oil price growth, nonoil private sector real GDP growth, real equity price growth, and US interest rates. Real oil price growth averaged 17 percent year-on-year during 1999-2011 despite having declined sharply in 2001 and 2009. More recently, it fell increasingly into negative territory, reaching close to -10 percent in 2014. Nonoil private sector real GDP growth jumped in 2004, from around 4 percent year on year on average during the preceding several years to near 20 percent and, despite decelerating, remained above 10 percent through 2008. However, as lower oil prices took their toll on economic activity, the variable’s performance fell to below 6 percent in 2014. Stock returns were impressive during the earlier part of the sample period on strong oil price performance.4 Real equity price growth was close to an average of 40 percent every year during 2001–05.5 However, following a sharp 40 percent drop in 2009, real stocks returns have been more subdued, registering an average of 3 percent decline per year during 2009–15. The U.S. Fed funds rate went through several cycles during the period. After increasing through 2000, it declined sharply in 2001. It rose during 2004–06, and declined sharply in 2008 and remained unchanged through 2014.

Table 3.Sampled GCC Banks(Number and share of domestic banking system)
Number of banksPercent share
Bahrain444
Kuwait560
Oman683
Qatar781
Saudi Arabia977
United Arab Emirates1162
Average768
Total42
Sources: Bankscope and authors’ calculations.Note: Percent share measured in terms of the stock of domestic credit.
Sources: Bankscope and authors’ calculations.Note: Percent share measured in terms of the stock of domestic credit.

The bank-level variables include NPLs as a share of total loans (the NPL ratio), real credit growth, and real deposit growth. The analysis incorporates 42 GCC banks for which sufficient time series data are available from Bankscope (Tables 3 and A1). These banks represent on average some 70 percent of the individual banking system in terms of the stock of credit. In terms of broad trends, the NPL ratio broadly declined during the sample period, at an accelerating pace during the first half of the 2000s when the oil price strengthened noticeably. The NPL ratio increased in 2009 coinciding with a sharp oil price decline, but due likely also to the realization of domestic vulnerabilities that had built up during the run up to the global financial crisis. The growth rates of credit and deposits appear to broadly mirror movements of oil prices, economic activity and equity prices.

C. A multivariate model of NPLs

A multivariate model is used to empirically assess the determinants of NPLs in the GCC. As is common in the literature, the dependent variable, the bank-level NPL ratio, was used after a logit transformation – this makes the NPL ratio a more normally-distributed variable and captures the empirical regularity that the variable tends to vary most for banks that start out with higher starting levels (Figure 4). A range of explanatory variables was considered, guided by the discussion in the earlier part of the paper.6

Figure 4.Distribution of NPL Ratios in GCC

Sources: Bankscope and authors’ calculation.

Dummy variables are introduced to help control for potential country effects as well as time effects not directly related to oil price shocks. As discussed earlier, many of the GCC countries experienced such stress events.

The determinants of the NPL ratio are estimated using the following multivariate panel data specification for bank i in year t.

where logit_NPLri,t is the logit transformation of the ratio of nonperforming loaks to total loans and the lagged regressor logit_NPLri,t-1 captures its persistence commonly found in the literature. Globalj,t-1 (j =1) is the US Fed funds rate. Macrok,t-1 represents macroeconomic variables (k = 1,2,3) - real growth rates of oil prices, real nonoil private sector GDP, and equity prices, lagged by one period. NPL ratios are expected to rise as US interest rates rise and the value of the macro variables declines. Bankii,i,t-1 is real credit growth (l = 1) lagged by one period. Positive and healthy credit growth would help support economic activity and lead to lower NPL ratios. That being said, in the medium term, higher leverage in the economy could build vulnerabilities. Year and country dummy variables are introduced in the regressions to control for events other than oil price developments that potentially led to an increase in NPL ratios and potential country effects. θi and εi,t are bank fixed effects and random errors.

Table 4.Determinants of Bank NPL Ratios in the GCC
System GMM results
Model number123456789101112
Logit of NPL ratio (L1)0.932***0.923***0.941***0.714***0.716***0.806***0.872***0.842***0.874***0.620*0.649*0.775***
Real oil price growth (L1)−0.005***−0.005***−0.005***−0.005***−0.005***−0.005***−0.012*−0.008*−0.008***−0.020*−0.015−0.013
nonoil GDP growth (L1)0.010.0010.006−0.055*−0.059**−0.037−0.058−0.076*−0.069*−0.276−0.234−0.155
Real equity price growth (L1)−0.004−0.004−0.004−0.005−0.005*−0.003−0.012−0.001−0.003−0.012−0.003−0.005
Real credit growth (L1)−0.005−0.003−0.00300.0010−0.004−0.004−0.003−0.004−0.003−0.002
Real US fed funds FD (L1)0.0840.0650.0860.0310.0150.08
Year dummyNNNNNNYYYYYY
Country dummyNNNYYYNNNYYY
Lag depth123123123123
p-values
AR(1)0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0230.0140.002
AR(2)0.2060.1960.1410.1030.1030.0960.3640.1920.2510.1440.1800.204
Hansen0.0020.0120.0300.0050.0260.1000.5460.6340.7900.9460.8500.802
Passes tests?NNNNNNYYYYYY
Note: The dependent variable is bank-by-bank (logit transformed) NPL ratio for selected GCC banks spanning 2000-2014 (annual frequency). Relying on a system GMM approach, with the collapsing method. The coefficients represent non-liner effect that depends on starting levels. ***, **, and * signify statistical significance at the 1%, 5% and 10% levels. L1 signifies one period lag. AR(1) and AR(2) signify p-values associated with the null hypothesis of lack of first and second order serial correlation. Hansen signifies p-value associated with the null hypothesis that the instruments are exogenous. A model is cosidered to pass tests if pvalues of both AR(2) and Hansen tests are 10% or greater.Sources: Bankscope, Haver, Bloomberg, and staff estimates.
Note: The dependent variable is bank-by-bank (logit transformed) NPL ratio for selected GCC banks spanning 2000-2014 (annual frequency). Relying on a system GMM approach, with the collapsing method. The coefficients represent non-liner effect that depends on starting levels. ***, **, and * signify statistical significance at the 1%, 5% and 10% levels. L1 signifies one period lag. AR(1) and AR(2) signify p-values associated with the null hypothesis of lack of first and second order serial correlation. Hansen signifies p-value associated with the null hypothesis that the instruments are exogenous. A model is cosidered to pass tests if pvalues of both AR(2) and Hansen tests are 10% or greater.Sources: Bankscope, Haver, Bloomberg, and staff estimates.

The results suggest that some of the macroeconomic and bank level variables are key determinants of NPL ratios in GCC (Table 4). A number of models were estimated using a system GMM approach. Models 7-12 pass certain statistical tests (Hansen, AR(1) and AR(2)). The NPL ratio exhibits strong persistence. Growth rates of real oil prices and nonoil private sector GDP are significant, suggesting that an increase in oil prices or private sector output leads to a decline in the NPL ratio. The coefficients on real equity prices growth and real credit growth come out with the correct sign but are not statistically significant. Real government spending and real U.S. interest rates do not directly affect NPL ratios in a systematic way (real government spending not shown).

The results are subject to some caveats. First, the information content of publicly available bank-level balance sheet data is relatively limited compared with the more granular regulatory data typically used for some other IMF assessments (such as FSAP). Second, any analysis based on historical data might not always account for the effects of recent improvements to risk-management and supervisory frameworks. Third, the data spanning 2000–14 may not capture a sufficient number of oil price and financial cycles. Fourth, as inherent in any econometric analysis, there is parameter uncertainty surrounding the estimated relationship between macroeconomic shocks and NPL ratios.

The empirical results imply that the actual and projected declines in oil prices and slowing of GDP growth could lead to an increase in the NPL ratio (Figure 5). For illustrative purposes, the NPL ratios in 10, 50 and 90 percentile in 2014 are taken as starting points to which the coefficients obtained from model 8 are applied. Data on actual and projected oil price performance for 2015–20 (see Figure 1) suggest that oil prices will, on average, remain 50-60 percent below the 2014 peak in the medium term. Recognizing the degree of uncertainty surrounding the oil price projections and that the risks are tilted to the upside, for this simulation, oil prices are assumed to decline by 50 percent in t = 0, and remain unchanged for three years. Similarly, non-oil GDP growth is assumed to weaken by 3 percentage points. Figure 5 shows that, starting from 0.8 percent, the NPL ratio would rise by 1 percentage point to 1.9 percent in three years. Banks with lower asset quality would witness larger increases. Starting from 2.5 (8.5) percent, the NPL ratio would rise by about 3 (10) percentage points to about 6 (19) percent. Nevertheless, their strong capital buffers, profitability, and provisions provide an important source of resilience for GCC banks.

Figure 5Dynamics of NPL Ratios After a 50 Percent Decline in Oil Prices and 3 Percentage Point Reduction in Nonoil GDP Growth in t=0

Source: Bankscope, Haver, and IMF staff calculations.

Note: The three cases illustrate banks in the 10th, 50th, and 90th percentile of NPLS in 2014 and the impact of an adverse oil price and growth shock scenario over a three year horizon.

Two alternative dynamic panel models were used to check the robustness of the baseline estimates – Fixed Effects (FE) and Least Squares Dummy Variable Correction (LSDVC) (Kiviet, 1995; Bruno, 2005). The fixed effects model incorporates the data’s panel structure but ignores the correlation between the lagged dependent variable and the regression error, thus yielding a downward-biased coefficient estimates for the lagged dependent variable. LSDVC corrects the biased in fixed effects-estimated coefficients, assuming that the independent variables are exogenous.7 The lagged dependent variable should lie between OLS and FE.8

Table 5.Determinants of Bank NPL Ratios in the GCC-Alternative Approach 1
Fixed-effects model results
Model number1314151617
Coeff.seCoeff.seCoeff.seCoeff.seCoeff.se
Logit of NPL ratio (L1)0.686***0.0340.681***0.0380.694***0.0390.685***0.0430.688***0.036
Real oil price growth (L1)−0.004***0.001−0.004***0.001−0.004***0.001−0.003***0.001−0.019***0.001
nonoil GDP growth (L1)−0.033***0.007−0.030***0.008−0.023**0.008−0.023**0.008−0.0090.007
Real equity price growth (L1)−0.002***0−0.002***0−0.002***00.0010.001
Real credit growth (L1)−0.0010.002−0.0010.002−0.0040.002
Real US fed funds FD (L1)−0.0230.018
Year dummyNNNNY
N575551499499499
R^20.5950.580.5810.5810.642
Note: The dependent variable is bank-by-bank (logit transformed) NPL ratio for selected GCC banks spanning 2000-2014 (annual frequency). Relying on a fixed effects approach. The coefficients represent non-liner effect that depends on starting levels. ***, **, and * signify significance at the 1%, 5% and 10% levels. L1 signifies one period lag.Sources: Bankscope, Haver, Bloomberg, and staff estimates.
Note: The dependent variable is bank-by-bank (logit transformed) NPL ratio for selected GCC banks spanning 2000-2014 (annual frequency). Relying on a fixed effects approach. The coefficients represent non-liner effect that depends on starting levels. ***, **, and * signify significance at the 1%, 5% and 10% levels. L1 signifies one period lag.Sources: Bankscope, Haver, Bloomberg, and staff estimates.
Table 6.Determinants of Bank NPL Ratios in the GCC-Alternative Approach 2
Bias corrected least square dummy variable dynamic panel estimator results
Model number1819202122
Coeff.seCoeff.seCoeff.seCoeff.seCoeff.se
Logit of NPL ratio (L1)0.685***0.0450.679***0.0590.790***0.0620.764***0.0620.864***0.051
Real oil price growth (L1)−0.0040.009−0.0040.006−0.004***0.001−0.003***0.001−0.004***0.001
nonoil GDP growth (L1)−0.0290.058−0.0260.053−0.017***0.006−0.017***0.007−0.0010.008
Real equity price growth (L1)−0.0020.007−0.002**0.001−0.002**0.00100.001
Real credit growth (L1)−0.0010.002−0.0010.002−0.0030.002
Real US fed funds FD (L1)−0.0230.023
Year dummyNNNNY
Note: The dependent variable is bank-by-bank (logit transformed) NPL ratio for selected GCC banks spanning 2000-2014 (annual frequency). Bias corrected least squares dummy variable estaimtor. Anderson-Hsiao. The coefficients represent non-liner effect that depends on starting levels. ***, **, and * signify significance at the 1%, 5% and 10% levels. L1 signifies one period lag.Sources: Bankscope, Haver, Bloomberg, and staff estimates.
Note: The dependent variable is bank-by-bank (logit transformed) NPL ratio for selected GCC banks spanning 2000-2014 (annual frequency). Bias corrected least squares dummy variable estaimtor. Anderson-Hsiao. The coefficients represent non-liner effect that depends on starting levels. ***, **, and * signify significance at the 1%, 5% and 10% levels. L1 signifies one period lag.Sources: Bankscope, Haver, Bloomberg, and staff estimates.

The results support the key findings of our baseline analysis (Tables 5 & 6). The NPL ratio exhibits similar levels of persistence. The coefficients of the lagged dependent variable are in a range of 0.68–0.69 based on the fixed effects model (Table 5, models 13–17), below 0.68–0.86 estimated using LSDVC (Table 6, models 18–22). A lower autoregressive coefficient generates a steeper path of a projected NPL ratio in a downside scenario analysis.

Growth rates of real oil prices, nonoil private sector GDP and real equity prices remain key determinants of NPL ratios in the GCC economies. The coefficients on real credit growth come out with the correct sign but are not statistically significant. Real government spending (not shown) and real U.S. interest rates do not directly affect NPL ratios in a systematic way.

D. A panel VAR model of oil-macro-financial linkages

To identify a positive feedback loop between the macroeconomic and bank-level balance sheet variables, a panel vector autoregression (VAR) model, which accounts for bank-level heterogeneity, was estimated. The multivariate model in the previous section considered the uni-directional effects of macroeconomic shocks on the bank NPL ratio. A panel VAR model employed in this section goes one step further and captures the spillback from the bank NPL ratio and other balance sheet variables to the macroeconomy.:

where yi,t is a vector of macroeconomic and bank-level variables, B0 is the deterministic component, (L) is a lag operator and uit is the residual. The model was estimated using a panel VAR routine pvar developed by Love and Zicchino (2006), which exploits a System-General Method of Moments (GMM) estimator as in Arellano and Bover (1995).9 The number of lags is set at two in view of the short time series dimension (2000–14).

Five macroeconomic and bank level variables were included: macroeconomic variables include real oil price growth and real equity price growth; bank level variables are NPL ratios, real credit growth and real deposit growth for 42 GCC banks for which sufficient time series data are available. The data spans the period 2000–14.

The identification of shocks is based on a Cholesky decomposition where the variables are stacked to explore how macroeconomic shocks affect bank-level variables first, and how the latter affect the former in the second round. In particular, one macroeconomic variables {real oil price growth} was stacked at the top. The bank-level variables {NPL ratio, real credit growth, real deposit growth} are stacked below the macro-level variables. Finally, real equity price growth is stacked at the bottom as commonly done in the literature.

The estimated results are summarized as follows. Table 7 presents the estimated coefficients from the system GMM approach in the panel VAR model. The estimated coefficients are often statistically significant. Table 8 includes impulse response values after normalizing by the size of each variable’s one standard deviation shock. Figure 6 displays the values visually.

Table 7.GCC: Estimated Panel VAR Coefficients and T-Statistics
roilpgnplrrcrgrdpgreqpg
coef.t-stat.coef.t-stat.coef.t-stat.coef.t-stat.coef.t-stat.
L.roilpg−0.006−0.0780.0070.8600.0701.8830.0831.894−0.580−8.010
L.nplr1.9383.1490.8806.169−0.281−1.034−0.583−1.8230.4530.668
L.rcrg0.0490.4960.0010.0820.0450.559−0.012−0.1480.3432.276
L.rdpg0.2412.395−0.030−3.7280.3283.7820.2412.7540.0840.573
L.reqpg−0.020−0.632−0.003−0.8940.0040.132−0.023−0.8460.3775.238
L2.roilpg0.0631.1500.0030.5830.0611.7690.0310.851−0.040−0.603
L2.nplr0.3740.956−0.223−2.4230.8592.9250.8843.1401.7382.834
L2.rcrg0.1992.3740.0060.7200.0991.4610.0410.4570.2071.271
L2.rdpg0.0110.1390.0192.185−0.007−0.1080.0360.385−0.259−2.180
L2.reqpg0.0110.363−0.011−3.3550.0662.5170.0843.406−0.014−0.284
Sources: Bankscope, Haver, and IMF staff calculations.Note: Estimated using a panel VAR routine with two lags. Annual data spanning 2000-14. Bank level data for NPL ratio, real credit and deposit growth for 42 GCC banks. “L.” and “L.2” are lag operator indicating the first and second lag. roilpg is real oil price growth, nplr is NPL to total loans ratio, rcrg real credit growth, rdpg is real deposit growth, reqpg is real equity price growth.
Sources: Bankscope, Haver, and IMF staff calculations.Note: Estimated using a panel VAR routine with two lags. Annual data spanning 2000-14. Bank level data for NPL ratio, real credit and deposit growth for 42 GCC banks. “L.” and “L.2” are lag operator indicating the first and second lag. roilpg is real oil price growth, nplr is NPL to total loans ratio, rcrg real credit growth, rdpg is real deposit growth, reqpg is real equity price growth.

There is strong empirical evidence of feedback loops between oil price movements, bank balance sheets, and asset prices in the GCC (Figure 6). Results from a panel VAR model suggest that, first, oil price movements affect bank balance sheets in a significant way. A drop in the growth rate of oil prices results in a rise in the ratio of nonperforming loans (NPL) to gross loans, and a reduction in the real growth rates of bank credit and deposits (Figure 1). A 1 percent decline in oil prices leads to a 0.2–0.3 percentage point decline in real credit growth and a 0.1–0.2 percentage point decline in real deposit growth—with timing varying from immediate to 2–3 year lags. The NPL ratio would increase by about 0.1 percentage point in the long run. There is also a feedback effect within bank balance sheets, as a higher NPL ratio leads to lower real bank credit and deposit growth—as solvency risk rises, banks reduce lending to boost capital adequacy ratios, while the customers lose confidence in the banks, and vice versa. These results are consistent with other studies on the GCC economies (see, for instance, Espinoza and Prasad, 2010).

Table 8.GCC: Response of Macro and Bank Level Variables
(Shock variable)
timeroilpgnplrrcrgrdpgreqpg
(Responses)
0roilpg22.2410.0000.0000.0000.000
1roilpg−0.2403.7842.2192.907−0.538
2roilpg2.6114.0012.7280.277−0.189
3roilpg0.6973.3931.0131.059−0.197
4roilpg0.6862.6620.8560.367−0.393
5roilpg0.3301.9380.4610.376−0.475
6roilpg0.3271.3380.3060.261−0.454
0nplr−0.5172.2420.0000.0000.000
1nplr−0.4562.037−0.221−0.378−0.075
2nplr−0.3651.331−0.102−0.204−0.366
3nplr−0.1710.692−0.116−0.074−0.492
4nplr0.0320.288−0.0930.040−0.393
5nplr0.1090.078−0.0430.094−0.249
6nplr0.120−0.0030.0020.096−0.128
0rcrg5.824−1.16113.5620.0000.000
1rcrg3.327−1.3722.9184.0410.100
2rcrg3.7370.8582.5481.5541.569
3rcrg1.0672.0171.6661.0821.351
4rcrg0.3992.2541.1580.2120.668
5rcrg−0.0041.9700.5760.0620.138
6rcrg−0.0281.4840.278−0.005−0.184
0rdpg3.967−2.1056.97812.2850.000
1rdpg2.632−1.7971.3682.898−0.610
2rdpg3.0380.5461.7891.7481.842
3rdpg0.6351.4501.2960.8121.274
4rdpg0.1781.6960.8980.0750.625
5rdpg−0.0941.4940.408−0.0170.155
6rdpg−0.0821.1200.180−0.052−0.127
0reqpg17.497−0.1596.4472.85226.333
1reqpg−4.2110.3837.6582.1019.920
2reqpg−2.1582.4593.519−2.6653.635
3reqpg−1.6773.1150.390−1.2981.938
4reqpg−1.4212.486−0.086−1.1570.359
5reqpg−1.0131.501−0.340−0.692−0.460
6reqpg−0.4490.662−0.372−0.322−0.647
Source: IMF staff calculations.
Source: IMF staff calculations.

Second, equity price developments are a channel for amplification of the bank liquidity feedback loop stemming from an adverse oil price shock. A one percent reduction in oil price growth leads to a 0.8 percent decline in the rate of equity price inflation, which in turn leads to a reduction of bank credit and deposit growth by 0.1 percentage point, further depressing equity price performance.

Figure 6.GCC: Response of Macro and Bank Level Variables

Sources: Bankscope, Haver, and IMF staff calculations.

Note: Estimated using a panel VAR routine with two lags. Annual data spanning 2000–14. Bank level data for NPL ratio, real credit and deposit growth for 42 GCC banks, roilpg is real oil price growth, nplr is NPL to total loans ratio, rcrg real credit growth, rdpg is real deposit growth, reqpg is real equity price growth.

III. Movements in Bank capital and provisions in the GCC

The existence of oil-macro-financial feedback loops suggests greater needs to build buffers in good times in the GCC. The importance of building buffers to cushion against negative shocks has been well documented. Increasing capital and provisions in good times helps enhance the resilience of the financial system and reduce procyclical feedback effects between asset prices and credit. In Saudi Arabia, empirical evidence confirms the view that bank capital and provisioning buffers have been moved counter-cyclically (Abusaaq et al, 2015). Both the capital and provisioning ratios increase as indicators of business and financial cycles strengthen.

Table 9.Correlation of NPL Provisions and Capital Ratios with Indicators of Business and Financial Cycles: Country Level Analysis(Correlation and p-values)
Credit to non oil GDP gapReal credit growthReal non oil GDP growthReal oil price growth
BahrainProvisions/NPLCoeff.0.070.62 ***0.210.29
p-val.0.800.020.460.31
CARCoeff.−0.82 ***0.45 ***0.73 ***0.57 *
p-val.0.000.100.000.03
KuwaitProvisions/NPLCoeff−0.74 ***0.57 ***0.450.28
p-val.0.000.030.100.32
CARCoeff.−0.340.63 ***0.43−0.16
p-val.0.220.010.120.59
OmanProvisions/NPLCoeff.0.51 ***0.28−0.11−0.22
p-val.0.060.330.700.45
CARCoeff.−0.90 ***0.300.63 ***0.49 *
p-val0.000.290.010.07
QatarProvisions/NPLCoeff.−0.46 ***0.35−0.090.34
p-val0.100.210.760.23
CARCoeff.−0.77 ***0.250.290.31
p-val0.000.460.390.34
Saudi ArabiaProvisions/NPLCoeff.0.69 ***0.090.35−0.09
p-val0.010.760.220.75
CARCoeff.−0.040.300.030.14
p-val0.880.300.910.63
UAEProvisions/NPLCoeff.0.51 ***0.28−0.11−0.22
p-val0.060.330.700.45
CARCoeff.−0.62 ***−0.440.210.12
p-val0.020.110.460.68
Sources: Bankscope and IMF staff estimates.Note: ***, **, and * signify statistical significance at the 1 percent, 5 percent, and 10 percent level.
Sources: Bankscope and IMF staff estimates.Note: ***, **, and * signify statistical significance at the 1 percent, 5 percent, and 10 percent level.

A simple empirical approach is used to shed light on the countercyclical behavior of loan loss provisions and capital ratios in the GCC. Developments in loan loss provisioning and capital adequacy ratios (CAR) are compared to movements in key business and financial cycle indicators, including the credit-to-nonoil GDP gap (estimated as percent deviations from HP trends), real credit growth, and real nonoil GDP growth.10 To account for GCC-specific factors, real oil price growth is also considered. The degree of countercyclical movement is assessed using correlation coefficients between provisions to NPLs or CAR with each of the four business and financial cycle indicators.11 Positive correlation coefficients, when statistically significant, signal the potential that provisions to NPLs and CAR are countercyclical – these ratios increase during good times. This exercise was conducted using country level data and bank level data.

Table 10.Correlation of NPL Provisions and Capital Ratios with Indicators of Business and Financial Cycles: Bank level analysis(Number of banks with statistically significant positive correlation)
Credit to non oil GDP gapReal credit growthReal non oil GDP growthReal oil price growth
Saudi ArabiaProvisions/NPL7112
CAR4220
United Arab EmiratesProvisions/NPL3330
CAR0220
Sources: Bankscope and IMF staff estimates.Note: The total number of banks analyzed is seven for both Saudi Arabia and the United Arab Emirates, respectively.
Sources: Bankscope and IMF staff estimates.Note: The total number of banks analyzed is seven for both Saudi Arabia and the United Arab Emirates, respectively.

There is tentative evidence to suggest these tools have moved countercyclically for some GCC banks (Table 10). Country-level loan loss provisions to NPL ratios are calculated using bank level data from Bankscope.12 This ratio is countercyclical relative to the credit to nonoil GDP gap in Oman, Saudi Arabia, and the United Arab Emirates; and relative to real credit growth in Bahrain and Kuwait. Country-level CARs are assessed based on published country-level financial soundness indicators data. They are countercyclical in Bahrain, Kuwait, and Oman with respect to real growth of credit, nonoil GDP, or oil prices. No systematic linkage was found for Qatar, Saudi Arabia, and the United Arab Emirates. In four instances either provisions to NPL ratios or CARs are found to be procyclical. The results need to be interpreted with caution as the annual data provide a relatively limited number of observations. Additionally, the estimated credit-to-nonoil GDP gap may not represent sufficiently the financial cycle.

Similar assessments using bank-level data reveal heterogeneity across individual banks. Given the relatively large sample size, a bank level analysis focuses on banks in Saudi Arabia and the United Arab Emirates. Table 11 reports the number of banks with statistically significant positive correlation is reported. In Saudi Arabia, provisions to NPLs ratios are countercyclical, in a statistically significant way with respect to the credit-to-nonoil GDP gap for 7 of the 11 banks, consistent with the evidence based on the aggregate data. The CAR is countercyclical with respect to the credit to nonoil GDP gap for 4 banks. In the United Arab Emirates, provisions to NPL ratios are countercyclical with respect to most macroeconomic variables for 3 out of 7 banks. The CAR is countercyclical with respect to real growth rates of credit and nonoil GDP for 2 of the 7 banks analyzed.

IV. Concluding Remarks

The paper has demonstrated the existence of oil-macro-financial linkages in the GCC countries. First, the performance of key indicators of business and financial cycles has generally strengthened during oil price upturns. The timing of downturns in those variables tends to coincide with oil price downturns, even though greater fiscal buffers have attenuated the linkage. Second, this paper’s econometric analysis confirmed that oil prices and economic activity significantly affect bank asset quality. It also identified feedback loops between oil price movements, bank balance sheets, and asset prices in the GCC.

Tentative evidence suggests that banks in the GCC countries have been setting the capital ratio and provisioning for NPLs countercyclically. A simple empirical analysis, using both country- and bank-level data, shows that loan loss provisions and capital adequacy ratios are positively correlated with indicators of business and financial cycles. This has helped strengthen the resilience of the financial system to the oil price decline since mid-2014.

Annex 1. Tables
Table A1.GCC Banks Included Econometric Analyses
CountryBank
Bahrain1Ahli United Bank BSC
2Gulf International Bank BSC
3BBK B.S.C.
4National Bank of Bahrain
Kuwait5National Bank of Kuwait S.A.K.
6Kuwait Finance House
7Gulf Bank KSC (The)
8Commercial Bank of Kuwait SAK (The)
9Industrial Bank of Kuwait K.S.C.
Oman10Bank Muscat SAOG
11Bank Dhofar SAOG
12National Bank of Oman (SAOG)
13HSBC Bank Oman
14Oman Arab Bank SAOC
15Ahli Bank SAOG
Qatar16Qatar National Bank
17Commercial Bank of Qatar (The) QSC
18Qatar Islamic Bank SAQ
19Doha Bank
20Ahli Bank QSC
21International Bank of Qatar Q.S.C.
22Qatar Development Bank Q.S.C.C.
Saudi Arabia23National Commercial Bank (The)
24Samba Financial Group
25Riyad Bank
26Banque Saudi Fransi JSC
27Saudi British Bank JSC (The)
28Arab National Bank Public Joint Stock Company
29Saudi Hollandi Bank
30Saudi Investment Bank (The)
31Bank AlJazira JSC
United Arab Emirates32National Bank of Abu Dhabi
33First Gulf Bank
34Abu Dhabi Commercial Bank
35Abu Dhabi Islamic Bank - Public Joint Stock Co.
36Mashreqbank PSC
37Union National Bank
38Commercial Bank of Dubai P.S.C.
39Bank of Sharjah
40National Bank of Fujairah PJSC
41Commercial Bank International P.S.C.
42National Bank of Umm Al-Qaiwain PSC
References

    Abusaaq, H., A.Alfi, P.Khandelwal, K.Miyajima, and B.Piven,2015, “Countercyclical Macroprudential Policies in Saudi Arabia,” International Monetary Fund, Washington DC.

    • Search Google Scholar
    • Export Citation

    Al-Hassan, A., M.Khamis, and N.Oulidi,2010, “The GCC Banking Sector: Topography and Analysis,” IMF Working Paper 10/87.

    Arellano, M., and O.Bover(1995). “Another look at the instrumental variable estimation of error-components models,” Journal of Econometrics, 68 (1), pp 29–51.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Arvai, Z., A.Prasad, and K.Ketayama,2014, “Macroprudential Policy in the GCC Countries,” IMF Staff Discussion Note 14/01.

    Bologna, P. and A.Prasad,2010, “Oman: Banking Sector Resilience” IMF Working Paper 10/61.

    Bruno, G. S. F.2005. Approximating the bias of the LSDV estimator for dynamic unbalanced panel data models. Economics Letters 87: 361–366

    • Search Google Scholar
    • Export Citation

    Callen, T., P.Khandelwal, K.Miyajima, and A.Santos,2015, “Oil Prices, Financial Stability, and the Use of Countercyclical Macroprudential Policies in the GCC,” Prepared for the Annual Meeting of Ministers of Finance and Central Bank Governors.

    • Search Google Scholar
    • Export Citation

    De Bock, R., and A.Demyanets (2012). “Bank Asset Quality in Emerging Markets: Determinants and Spillovers,” IMF Working Paper 12/71.

    Espinoza, R., and A.Prasad,2010, “Nonperforming Loans in the GCC Banking Systems and their Macroeconomic Effects,” IMF Working Paper 10/224.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Flannery, M. and K. W.Hankins,2013, “Estimating dynamic panel models in corporate finance,” Journal of Corporate Finance No 19, 1–19.

    • Search Google Scholar
    • Export Citation

    IMF, 2010, “Qatar: 2009 Article IV Consultation-Staff Report,” Country Report No. 10/41.

    Kiviet, J. F.1995. On bias, inconsistency, and efficiency of various estimators in dynamic panel data models. Journal of Econometrics68: 53–78.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Klein, N. (2013). “Non-Performing Loans in CESEE: Determinants and Impact on Macroeconomic Performance,” IMF Working Paper 13/72.

    Love, I., and R. T.Ariss (2014). “Macro-financial linkages in Egypt: A panel analysis of economic shocks and loan portfolio quality,” Journal of International Financial Markets, Institutions and Money, 28 (C), pp 158–81.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Love, I., and L.Zicchino,2006, “Financial development and dynamic investment behavior: evidence from panel vector autoregression,” The Quarterly Review of Economics and Finance, 46, pp 190−210.

    • Crossref
    • Search Google Scholar
    • Export Citation

    Miyajima, K.,2016, “An Empirical Investigation of Oil-Macro-Financial Linkages in Saudi Arabia,” IMF Working Paper 15/xxx

    Nickell, S.,1981, “Biases in Dynamic Models with Fixed Effects,” Econometrica, Vol 49, No 6, 1417–26.

    Nkusu, M. (2011). “Nonperforming Loans and Macrofinancial Vulnerabilities in Advanced Economies,” IMF Working Paper 11/161.

1

The authors thank Chikako Baba, Alberto Behar, Tim Callen, Sally Chen, Sean Craig, Phil De Imus, Mary Goodman, Aasim Husain, Deniz Igan, Etibar Jafarov, Taline Koranchelian, Inutu Lukonga, Camelia Minoiu, Erlend Nier, Marco Piñón, Ananthakrishnan Prasad, Damien Puy, Martin Sommer, James Walsh, and various MCD seminar participants for helpful comments. All remaining errors and opinions expressed in this paper are sole responsibility of the authors.

2

Al-Hassan and others (2010) explain the role of short-term capital inflows and high oil prices in fueling financial sector vulnerabilities in the run up to the global financial crisis. Arvai and others (2014) also discuss the near doubling of private sector credit as a share of non-oil GDP in the GCC during 2003-08 which contributed to systemic vulnerabilities.

4

The strong stock price performance may partly reflect the pent up demand from domestic investors to help diversify their asset allocation after some of the stock markets were established in the early 2000s.

5

Data on real estate prices are unavailable for GCC countries.

6

Real US fed funds rate is de-trended as unitroot tests suggest the variable is not stationary.

7

The Anderson and Hsiao approach is used.

8

The combination of fixed effects and lagged dependent variables can introduce econometric bias. In particular, OLS estimates of the lagged dependent variable’s coefficient in a dynamic panel model are biased due to the correlation between the fixed effects and the lagged dependent variable (Nickell (1981)). Such bias declines as panel length increases. Moreover, data property can affect different estimators’ performance (Flannery et al, 2013).

9

As the fixed effects are correlated with the regressors due to lags of the dependent variables, the mean-differencing procedure commonly used to eliminate fixed effects would create biased coefficients. The orthogonality between transformed variables and lagged regressors is preserved by forward mean-differencing (the Helmert procedure in Arellano and Bover, 1995), which removes the mean of the future observations. Then, lagged regressors are used as instruments to estimate the coefficients by system GMM.

10

The “credit to nonoil GDP gap”, defined here as the deviation in the ratio of credit to non-oil GDP from its long run trend, is a key indicator of the financial cycle, but should be complemented by additional indicators and judgment. It replaces the credit-to-GDP gap which is influenced by oil prices and does not provide a robust indicator of financial sector vulnerabilities in the GCC. The empirical literature finds that, currently, it is the single best early-warning indicator of crises, signaling crises five to three years in advance. To help address the well-known end-point issues with HP filters, the gaps for 2000–14 are calculated with two extra years of data, including forecasts through 2015–16. A two-sided approach is used.

11

The latter are lagged by one period (i.e. one year) to help reduce the chance of capturing reverse causality stemming from macro variables.

12

Bankscope-based country aggregates are consistent with IMF desk data and, importantly, available with longer time series. The sampled banks represent 50–96 percent (an average of 67 percent) of domestic banking systems measured in terms of percent of the stock of credit.

Other Resources Citing This Publication