A. Introduction

1. The UAE banking system has weathered well so far the recent period of lower oil prices. The UAE banking system remains well capitalized, liquid, profitable, with high asset quality. However, sustained lower oil prices and higher short-term interest rates pose challenges. Asset quality could deteriorate in tandem with lower economic activity if lower oil prices are offset by cuts in expenditures with a high fiscal multiplier. In addition, liquidity could come under further pressure if the government continues to withdraw its banking deposits to finance the deficit. Banks have reacted to lower government deposits by attracting nonresident deposits and increasing wholesale funding to sustain robust private sector credit growth without raising lending rates on average. However, these sources of wholesale funding are more costly² and this, together with lower asset quality,³ can affect the profitability and capital buffers of banks. If there are further reductions in government deposits and increases in US policy rates, banks will need to manage a combination of liquidity pressures, higher funding rates, and lower profitability

2. The paper aims at analyzing the financial stability implications of lower oil prices and higher short-term interest rates for the UAE. The paper focuses on the effects of lower oil prices and higher short-term interest rates on liquidity and solvency. To capture its different dimensions,⁴ liquidity is described by three indicators: the inverse of the loan-to-deposit ratio, the interbank loans-to-interbank deposits ratio, and the liquid assets-to-customer deposits and short-term debt ratio. Their determinants are analyzed with a view to assess the impact of lower oil prices, higher short-term interest rates, and other bank-specific characteristics. Solvency is proxied by probabilities of default under adverse macroeconomic scenarios. They are obtained with a forward intensity model, which is flexible enough to incorporate not only defaults but also exits of firms from mergers and acquisitions. Finally, the paper provides policy recommendations to mitigate the adverse effects of lower oil prices and higher short-term interest rates on the banking system.

B. Banking Sector Developments

3. The UAE banking system has attracted new deposits and tapped wholesale markets to mitigate the effect of lower deposit growth on funding and maintain robust loan growth. While credit to the private sector remained robust at 8.7 percent at end-2015 (Figure 1), total deposit growth—including nonresident deposits—was substantially weaker at one percent. In particular, the contribution of government deposits to total deposit growth has been negative since 2015:Q2. As a result, the gap between private sector credit and total deposit growth has been felt in the interbank market where the spread between the interbank offer rates EIBOR (UAE interbank rate) and Libor temporarily widened in October 2015 after a two-year period of low spreads. Given the implementation of the new liquidity standards,⁵ banks have not reduced their liquid assets to sustain private sector credit growth. Instead, banks have tapped wholesale funding and expanded their corporate and non-resident deposit base.

View Full Size

Recent Banking Sector Developments

Citation: IMF Staff Country Reports 2016, 266; 10.5089/9781475522075.002.A002

Sources: Country authorities; Haver; NBS; and IMF staff estimates.

• Download Figure
• Download figure as PowerPoint slide

4. Despite slowing non-oil growth amid lower oil prices, UAE domestic banks remain well capitalized and profitable, with high asset quality. Asset quality has substantially improved since the completion of major GRE debt restructurings during 2012-14. The average nonperforming loan ratio in the UAE domestic banks declined from 8 percent at end-2012 to 4.6 percent at-end 2015 (Figure 2), even though there has been a slight increase in nonperforming loans in the SME portfolio associated with the recent slowdown in non-oil growth. In addition, the higher-than-average nonperforming loan ratios were concentrated in a fewer than 40 percent of domestic banks at end-2015. Average profitability has improved since 2009 with a slight reversal in 2015 due to a higher cost of funding, including higher interest rate on deposits and more competitive rates on credit, while the percentage of banks with above-average return on assets was high at more than 60 percent at-end 2015. The average capital adequacy ratios in the UAE banks remain high at more than 18 percent at end-2015 well above the regulatory requirement of 12, even though they have been on a declining trend since 2009. In terms of distribution, more than 60 percent of banks hold capital at levels lower than the average but still higher than the 12 minimum capital requirement.

View Full Size

Financial Soundness Indicators

(In Percent)

Citation: IMF Staff Country Reports 2016, 266; 10.5089/9781475522075.002.A002

Source: Bankscope; and author’s calculations.

• Download Figure
• Download figure as PowerPoint slide

5. Liquidity has been high but on a declining trend since 2007. The average liquid assets to customer deposits and short-term debt ratio in the UAE domestic banks has declined from 35 percent at end-2007 to about 20 percent at end-2015. More than 60 percent of domestic banks had lower liquid assets as a percentage of customer deposits and short-term debt than the average ratio at end-2015. The average loan-to-deposit ratio has been stable at around 100 percent since 2012 after declining for more than four years. About 60 percent of banks had granted loans that represented more than 100 percent of their deposits at end-2015. Consistent with decreasing liquid assets as a percentage of customer deposits and short-term debt, the average interbank placements-to-interbank deposits ratio has declined since 2012. More than 60 percent of the banks had interbank placements that represented less than 200 percent of interbank deposits at end-2015.

6. The high level of liquidity is also a result of central bank regulations. The Loan-to-Stable Resources Ratio (LSRR) was introduced in 1986 to promote a stable funding ratio in which net loans, guarantees, and short-term (less than 3 months) interbank loans were required to be fully funded with customer deposits, medium- and long-term (more than 6 months) interbank loans, and capital and reserves. In addition, reserve requirements on demand, savings, and call accounts were raised to 14 percent in 1999 while reserve requirements on time deposits were reduced to 1 percent.⁶ Finally, the central bank has aligned its regulatory framework with Basel III by introducing the Eligible Liquid Assets Ratio (ELAR) and the Liquidity Coverage Ratio in 2015. The new ELAR was implemented in July 2015 and requires eligible banks to hold liquid assets (cash, central bank CDs, reserve requirements, UAE and foreign government bonds and Sukuk with a zero-risk weight) equivalent to 10 percent of total liabilities while the new LCR became effective on January 1, 2016.

7. Given the comfortable liquidity position of the banking system, the central bank’s collateralized loans to the banking system have been marginal since 2012. The central bank has established collateralized dirham and dollar liquidity facilities for conventional and Islamic banks. The framework is flexible enough to include not only domestic but also investment grade foreign securities as collateral. A rough estimate of the stock of repoable securities—including foreign securities and claims on the government—that could be pledged to the central bank with no rehypothecation is about 25 percent of total assets in the banking system as of end-2015.

C. Assessing the Impact of Lower Oil Prices and Higher Short-term Interest Rates on Liquidity

8. Even though the empirical literature on the determinants of bank liquidity is new, it can help shed light on the effect of lower oil prices and higher short-term interest rates on the liquidity buffers. The 2007-8 financial crisis is a reminder that fragilities resulting from liquidity and funding mismatches can lead to liquidity pressures and trigger bank failures. As a result, it is important to know how banks adjust their liquidity buffers in response to different shocks. Aspachs, Nier, and Tiesset (2005), Bonfim and Kim (2012), and Delechat, Henao, Muthoora, and Vtyurina (2014) provide a summary and guidance on the determinants of liquidity buffers.

9. The empirical literature has highlighted the role of bank-specific characteristics as important determinants of bank liquidity. Bank-specific characteristics such as profitability and capital can have a negative impact on liquidity buffers as banks with high profitability and capital levels can fund their operations with internal resources or can have easy market access, requiring them to hold less liquidity buffers. On the other hand, high levels of capital can also have a positive impact on liquidity as a result of capital and liquidity requirements in which larger holdings of government securities with zero-risk weight are also considered high-quality liquid assets. Other bank-specific characteristics such as lending specialization and net interest margins can also have a negative impact on liquidity buffers as they encourage lending and lead to high loan-to-deposit and low interbank ratios. In the same vein, high cost to income can have a positive impact on liquidity buffers as banks with high cost to income are less prone to fund themselves with internal resources, requiring banks to hold more liquidity buffers. Finally, the effect of bank size on liquidity buffers is ambiguous. On one hand, large bank size implies easy market access, requiring banks to hold less liquidity buffers. On the other hand, large bank size also implies high scrutiny by supervisors and markets, requiring banks to hold more liquidity buffers.

10. In addition to bank-specific characteristics, the empirical literature has also included macroeconomic variables as important determinants of bank liquidity. Macroeconomic variables such as economic growth can have a negative impact on liquidity buffers as banks expand their loan portfolio during economic expansions, resulting in less liquidity buffers, while higher short-term interest rates on deposits discourage banks from holding large liquidity buffers. In an oil exporting country, higher oil prices can also play a role in lubricating the financial system. When oil prices are higher, oil-related bank customers can make more deposits in the banking system, helping banks increase their liquid assets.

11. To assess the determinants of liquidity buffers in the UAE domestic banks, a parsimonious specification should include not only bank-specific characteristics but also oil prices, short-term interest rates, and non-oil GDP as explanatory variables. As suggested in Bonfim and Kim (2012), liquidity buffers can be proxied by (the inverse of) the loan-to-deposit ratio, the interbank loans-to-interbank deposits ratio, and the liquid assets-to-customer deposits and short-term debt ratio. Bank-specific characteristics include total assets (in log), return on asset, capital adequacy ratio, cost-to-income ratio, the net loans-to-total assets ratio (specialization), and the net interest margin. Important macroeconomic variables in the context of an oil producer with a peg to the US dollar are: changes in the three-month Libor rate (in percentage points), the percentage change in Brent oil prices, and the real non-oil GDP. The proxies for liquidity buffers and their explanatory variables are collected for a sample of 17 domestic banks for the period 2005:Q1-2015:Q4. However, not all proxies and explanatory variables are available for the full period range, which limits the number of observations to a minimum of 290 observations. The estimation method is unbalanced panel data regression with fixed effects and cluster-robust standard errors to account for autocorrelation within time and across banks. Seasonal dummy variables are also included to account for quarterly seasonality.

12. The results of the panel data regression estimation indicate that bank-specific characteristics play an important role in determining bank liquidity buffers. Table 1 reports the coefficient estimates and their respective p-values obtained with cluster-robust standard error terms to account for autocorrelation in the residuals within time and across banks. The results indicate that selective bank-specific characteristics are statistically significantly different from zero. Capital in specification (6), return on asset in specifications (3), (4), and (6), asset size in specifications (1), (4), and (5), specialization in specifications (1), (3), (5), and (6) have a negative impact on liquidity buffers as expected. However, the estimate for the net interest margin in specifications (1), (2), and (4) has a positive but opposite sign than expected. In addition, while the estimate for the cost to income in specification (1) implies that higher cost to income has a positive impact on liquidity buffers, the respective estimate in specification (5) does not have the sign as suggested by the literature. Finally, the estimates for lagged dependent variables as explanatory variables in specifications (2), (4), and (6) are larger than 0.5, implying a high degree of persistence driving the dynamics of liquidity buffers.

Table 1.

Determinants of Liquidity Buffers

Source: Bankscope, Alvarez, Barbero and Zofio (2013), and author’s calculations. Note: P-values reported between parenthesis are associated with t stastics adjusted for serial correlation with clustered standard errors in specifications (1), (3), and (5). Green and yellow cells indicate statistically significant different from zero at 5 percent and 10 percent, respectively. Both the F test of individual effects and the Baltagi and Li (1990) version of the Breusche and Pagan (1980) test rejected the null hypothesis of no fixed effects for all specifications. Wooldridge’s test for serial correlation and Pesaram test for cross correlation rejected the null hypothesis of no serial correlation in specifications (1), (3), and (5). Cluster-robust standard erros are calculated according to Gow, Ormazabal, and Taylor (2010) and Cameron and Miller (2015).

Table 1.

Determinants of Liquidity Buffers

	Inverse of the Loan-to-deposit ratio		Interbank loans-to-interbank deposits ratio		Liquid assets-to-customer deposits and short-term debt ratio
	(1)	(2)	(3)	(4)	(5)	(6)
Capital adequacy ratio	−0.0001	0.0014	5.3000	4.8000	−0.1161	−0.1960
	(0.98)	(0.36)	(0.43)	(0.18)	(0.35)	(0.07)
Log of total assets	−0.0556	−0.0128	−89.6000	−61.4000	−5.4805	−1.2406
	(0.00)	(0.41)	(0.11)	(0.08)	(0.00)	(0.28)
Return on asset	0.0036	−0.0106	−50.0000	−57.7000	−0.9967	−1.5099
	(0.67)	(0.17)	(0.06)	(0.00)	(0.19)	(0.01)
Cost to income ratio	0.0018	0.0022	−4.6000	−1.7000	−0.2246	−0.0468
	(0.19)	(0.03)	(0.13)	(0.42)	(0.03)	(0.54)
Specialization	−0.0046	−0.0008	−3.6000	0.4000	−0.3337	−0.1126
	(0.00)	(0.24)	(0.05)	(0.76)	(0.00)	(0.02)
Net interest margin	0.0418	0.0374	36.1000	52.1000	−1.6130	0.8144
	(0.00)	(0.00)	(0.44)	(0.01)	(0.17)	(0.26)
3 month Libor rate change	0.0065	0.0021	−14.1000	−22.3000	−1.8748	−0.9585
	(0.60)	(0.88)	(0.78)	(0.44)	(0.05)	(0.35)
Oil price change	0.0001	0.0001	0.7000	−0.1000	0.0441	0.0185
	(0.61)	(0.77)	(0.39)	(0.84)	(0.02)	(0.33)
Nonoil growth rate	−0.0029	−0.0043	−1.3000	14.0000	0.8037	0.1536
	(0.76)	(0.59)	(0.97)	(0.41)	(0.19)	(0.79)
Lagged dependent variable		0.6050		0.5000		0.5948
		(0.00)		(0.00)		(0.00)
Number of observations	353	353	303	292	353	353
Adjusted R-squared	0.20	0.41	0.04	0.27	0.19	0.42
Cluster-robust standard errors	Yes	No	Yes	No	Yes	No
Seazonal dummy variables	Yes	Yes	Yes	Yes	Yes	Yes

Source: Bankscope, Alvarez, Barbero and Zofio (2013), and author’s calculations. Note: P-values reported between parenthesis are associated with t stastics adjusted for serial correlation with clustered standard errors in specifications (1), (3), and (5). Green and yellow cells indicate statistically significant different from zero at 5 percent and 10 percent, respectively. Both the F test of individual effects and the Baltagi and Li (1990) version of the Breusche and Pagan (1980) test rejected the null hypothesis of no fixed effects for all specifications. Wooldridge’s test for serial correlation and Pesaram test for cross correlation rejected the null hypothesis of no serial correlation in specifications (1), (3), and (5). Cluster-robust standard erros are calculated according to Gow, Ormazabal, and Taylor (2010) and Cameron and Miller (2015).

View Table

Table 1.

Determinants of Liquidity Buffers

	Inverse of the Loan-to-deposit ratio		Interbank loans-to-interbank deposits ratio		Liquid assets-to-customer deposits and short-term debt ratio
	(1)	(2)	(3)	(4)	(5)	(6)
Capital adequacy ratio	−0.0001	0.0014	5.3000	4.8000	−0.1161	−0.1960
	(0.98)	(0.36)	(0.43)	(0.18)	(0.35)	(0.07)
Log of total assets	−0.0556	−0.0128	−89.6000	−61.4000	−5.4805	−1.2406
	(0.00)	(0.41)	(0.11)	(0.08)	(0.00)	(0.28)
Return on asset	0.0036	−0.0106	−50.0000	−57.7000	−0.9967	−1.5099
	(0.67)	(0.17)	(0.06)	(0.00)	(0.19)	(0.01)
Cost to income ratio	0.0018	0.0022	−4.6000	−1.7000	−0.2246	−0.0468
	(0.19)	(0.03)	(0.13)	(0.42)	(0.03)	(0.54)
Specialization	−0.0046	−0.0008	−3.6000	0.4000	−0.3337	−0.1126
	(0.00)	(0.24)	(0.05)	(0.76)	(0.00)	(0.02)
Net interest margin	0.0418	0.0374	36.1000	52.1000	−1.6130	0.8144
	(0.00)	(0.00)	(0.44)	(0.01)	(0.17)	(0.26)
3 month Libor rate change	0.0065	0.0021	−14.1000	−22.3000	−1.8748	−0.9585
	(0.60)	(0.88)	(0.78)	(0.44)	(0.05)	(0.35)
Oil price change	0.0001	0.0001	0.7000	−0.1000	0.0441	0.0185
	(0.61)	(0.77)	(0.39)	(0.84)	(0.02)	(0.33)
Nonoil growth rate	−0.0029	−0.0043	−1.3000	14.0000	0.8037	0.1536
	(0.76)	(0.59)	(0.97)	(0.41)	(0.19)	(0.79)
Lagged dependent variable		0.6050		0.5000		0.5948
		(0.00)		(0.00)		(0.00)
Number of observations	353	353	303	292	353	353
Adjusted R-squared	0.20	0.41	0.04	0.27	0.19	0.42
Cluster-robust standard errors	Yes	No	Yes	No	Yes	No
Seazonal dummy variables	Yes	Yes	Yes	Yes	Yes	Yes

Source: Bankscope, Alvarez, Barbero and Zofio (2013), and author’s calculations. Note: P-values reported between parenthesis are associated with t stastics adjusted for serial correlation with clustered standard errors in specifications (1), (3), and (5). Green and yellow cells indicate statistically significant different from zero at 5 percent and 10 percent, respectively. Both the F test of individual effects and the Baltagi and Li (1990) version of the Breusche and Pagan (1980) test rejected the null hypothesis of no fixed effects for all specifications. Wooldridge’s test for serial correlation and Pesaram test for cross correlation rejected the null hypothesis of no serial correlation in specifications (1), (3), and (5). Cluster-robust standard erros are calculated according to Gow, Ormazabal, and Taylor (2010) and Cameron and Miller (2015).

13. The results also indicate that macroeconomic variables are important determinants of liquidity buffers. As expected, the estimate for changes in the oil prices in specification (5) imply that higher oil prices should encourage banks to hold liquidity buffers. On the other hand, the estimate for changes in the three-month Libor rate in specification (6) suggests that higher short-term interest rates should encourage banks to lend and hold lower liquidity buffers. A possible explanation for the low statistical significance of non-oil growth rates in all specifications is that higher or lower non-oil growth rates affect simultaneously deposits, loans, and, as a result, liquid assets as business confidence rises or deteriorates. In addition, the interpolation of the annual non-oil GDP series might not be adequate to reflect its dynamics in a quarterly context. Finally, fiscal financing might also affect deposits and liquid assets simultaneously, which could make the liquidity ratios insensitive to changes in non-oil growth.

D. Assessing the Impact of Lower Oil Prices and Higher Short-term Interest Rates on Solvency

14. The effect of lower oil prices and higher short-term interest rates on credit quality and solvency in the UAE banking system has so far been modest. Despite strong fiscal consolidation during 2015, credit growth has remained robust, asset quality stable and capital buffers comfortable. However, sustained depressed oil prices and further tightening of monetary conditions could have an adverse impact on confidence, investment, and consumption, leading to an economic downturn and rising probabilities of default in the corporate and banking sectors.

15. Solvency of the UAE corporate sector can be assessed by analyzing probabilities of default (PDs) under adverse macroeconomic scenarios. In particular, probabilities of default (PDs) based on a forward intensity model, developed by the National University of Singapore (NUS), are a reduced form in which defaults have a Poisson distribution and the intensity of default events is a function of variables with predictive power.⁷ The NUS intensity model is flexible enough to incorporate not only defaults but also exits of firms arising from mergers and acquisitions. In this model customized to the UAE, the PDs are a function of common independent and firm-specific risk factors such as stock market index, short-term interest rate, distance-to-default (the expected difference between the asset value and the default barrier, adjusted and normalized by asset volatility), profitability, size, market-to-book value, and idiosyncratic volatility. Both common and firm-specific risk factors are then driven by macroeconomic risk factors such as oil price changes, real non-oil GDP growth, consumer price inflation, and changes in the three-month EIBOR interbank rate. The NUS model is calibrated with monthly data for the period January 1990-December 2015 for a total of 74 UAE listed firms, of which 17 firms are banks while the remaining 57 ones are either private firms (47) or GREs (10). The annual macroeconomic data are interpolated accordingly.

16. The projections for the PDs start by defining scenarios for macroeconomic risk factors. The first scenario is a baseline consistent with the 2016-21 macroeconomic framework in which oil Brent prices continue to decline to $45.3 per barrel in 2016 but to gradually recover to $60.1 per barrel in 2021, real non-oil GDP growth slows down to 2.4 percent in 2016 and gradually improve to 4 percent in 2021, CPI inflation declines to 2.8 percent in 2017 and gradually increases to 3.6 percent in 2021, and the EIBOR interbank rate changes according to the Libor projections throughout the period. The adverse oil shock scenario consists of a sudden but permanent $10 reduction in oil prices through 2021, a 1.7 percent decline in non-oil GDP over the period 2016-21 associated with a gradual fiscal consolidation that would eliminate the additional fiscal deficit resulting from lower oil revenues,⁸ a stable consumer price inflation and EIBOR interbank rate increases similar to the baseline. Finally, the interest rate shock scenario consists of an increase of 100 bps in the EIBOR interbank rate both during 2016-17.

17. Overall, the PDs for the 74 UAE listed firms were low in the period preceding the oil shock and would increase under adverse scenarios. At the height of the financial crisis, the average PD for the 17 banks spiked to 100 bps while the average PD for other firms peaked at 38 bps. While the average PD for banks gradually declined to 7.2 bps in May 2013, the average PD for other firms remained volatile and above the average PD for banks in the period September 2011-March 2015. Since then, the average PD for banks has been higher than the PD for other firms and is also projected to remain higher under both adverse scenarios, with a minimum of 21 bps and a maximum of 50 bps on average under the adverse oil price shock scenario. Looking forward, under both the lower oil price and higher interest rate scenarios, PDs for UAE banks and other UAE firms are projected to significantly increase, though from a lower base. Under the adverse oil shock scenario with economic slowdown, PDs are forecast to spike and reach their highest levels since the 2009 crisis.

View Full Size

Probabilities of Default for Banks and Corporates under Adverse Macroeconomic Scenarios

Citation: IMF Staff Country Reports 2016, 266; 10.5089/9781475522075.002.A002

• Download Figure
• Download figure as PowerPoint slide

E. Conclusion and Policy Recommendations

18. Stressed macroeconomic conditions are expected to put pressures on liquidity and solvency of the banking sector. The empirical evidence shows that a severe scenario with oil prices lower by $10 than the baseline (24 percent lower at the peak) will bring the liquid assets-to-customer deposits and short-term debt ratio down by 0.9 percentage points with respect to the baseline and will increase banks’ probability of default by 7.5 bps in addition to the baseline on average. If the real non-oil GDP declines by 1.7 percent over 2016-21 as a result of fiscal policies to restore fiscal balance as described above, the probability of default would peak at 16.2 bps higher than the baseline. Similarly, an interest rate hike of 200 bps will reduce the liquid assets-to-customer deposits and short-term debt ratio by 3.7 percentage points in addition to the baseline and will lead to a marginal increase in the probability of default by 0.23 bps with respect to the baseline on average.

19. The results above shed light on the importance of timely implementing the central bank’s plans aimed at further strengthening banking liquidity and solvency, and enhancing corporate governance. Increased PDs in the corporate sector call for adequate provisioning of banks’ portfolio as recently required by the central bank, strengthening the enforcement of loan-concentration limits and of tight control of related-party lending, and phasing in Basel III capital requirements. Regarding liquidity, supervisors should encourage banks to increase transparency on maturity mismatches in their market disclosures, diversify their funding sources, strengthen their treasury function, and maintain a cushion of high-quality liquid assets. For example, the limited information available on projected cash flows on a contractual basis published as part of annual reports and Basel II pillar III disclosures indicate that, overall, there is a short-term liquidity gap within the three-month maturity bucket in UAE banks and liquidity surpluses in the longer time horizon maturity buckets. It is also important to further develop safety nets and resolution frameworks.