## _wp1541

## Source details

**Canonical URL:** [_wp1541](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1541.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1541.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1541.pdf.json)

---

### Introduction and research question
- Maturity transformation—the conversion of short term liabilities into long term assets—creates liquidity mismatch risk that can lead to bank runs and insolvency (Diamond and Dybvig (1983)).
- Research objective:
  - Examine whether a rumor-driven financial panic in Pakistan (September–October 2008) affected deposit and lending behavior of Islamic and conventional banks differently in a country with no explicit deposit insurance.
- Two-step empirical approach:
  - Detect heterogeneous liquidity shocks using weekly bank-level deposit movements during the panic.
  - Assess whether changes in deposits affected banks’ lending behavior, and whether that effect differed for Islamic versus conventional banks, controlling for bank fundamentals and using fixed effects (bank fixed effects for deposit changes; borrower fixed effects for lending).

### Event: September–October 2008 financial panic (Pakistan)
- Timing and headline magnitude:
  - Demand deposits fell continuously over a period of seven weeks (from September 27, 2008 to November 14, 2008).
  - In three weeks (from September 27, 2008 through October 18, 2008), demand deposits declined by 4 percent or 131 billion PKR.
  - Aggregate demand deposits fell from a pre-panic maximum of 3,397 bn PKR to a low of 3,233 bn PKR, or by 4.8 percent.
- Cross-bank heterogeneity:
  - Some banks experienced deposit outflows up to 46 percent; others experienced inflows up to 58 percent.
  - The shock was positive for 18 banks/branches and negative for 34 banks/branches.
  - CBs mainly experienced negative shocks; IBIs (in particular ISs) mainly experienced positive shocks.
- Market stress indicators and policy response:
  - Several banks failed to meet weekly cash reserve requirements.
  - Annualized overnight interbank lending rate reached 48 percent in October 2008.
  - SBP actions: decreased reserve requirement from 9 percent to 5 percent in October 2008; exempted short-term liabilities from CRR; reductions in CRR and SLR infused almost PKR 270 billion of liquidity; conducted thirteen OMOs injecting temporary liquidity in excess of PKR 300 billion; several banks borrowed through the discount window facility.
  - It took more than two months for deposits to reach the pre-crisis level.

### Data and empirical strategy
- Data:
  - Weekly deposit data for all banks: July 11, 2008 to January 2, 2009.
  - Loan-level data (newly granted loans) from the Credit Information Bureau at the SBP (matched to bank balance-sheet data).
  - Other bank-level covariates measured as of June 30, 2008 to avoid endogeneity.
  - Exchange rate contextual note: 1 USD ~ PKR 79 in October 2008.
- Key regression specifications:
  - Deposit-change regression (Specification (1)): dependent variable Δ log(deposits) over the seven-week crisis (difference between November 14, 2008 and September 27, 2008); key regressors include IB_i (Islamic Banks), IS_ij (Islamic subsidiaries/branches), and bank characteristics X_ij (CAMEL fundamentals, size, reach, age, credit rating, non-deposit funding share).
  - Loan-level lending regression (Specification (2)): dependent variable ln(new loan) by bank i, branch type j, to firm k during the panic; includes Δ log(deposits), IBI_ij, and controls X_ij, F_k, L_l.
  - Identification enhancement (Specification (3)): adds borrower (firm) fixed effects to isolate credit-supply effects from credit-demand variation (approach following Khwaja and Mian (2008)).

### Main empirical findings — deposits
- Aggregate and baseline:
  - CBs experienced a large withdrawal of demand deposits during September–October 2008 (to the tune of 7 percent in the parsimonious specification).
- Islamic vs conventional:
  - Full-fledged IBs: coefficient statistically insignificant in parsimonious regression — demand deposit outflows from IBs were similar to CBs in that specification.
  - Islamic subsidiaries/branches (ISs): estimated coefficient positive and statistically significant — ISs recorded an increase in demand deposits during the period to the tune of 19 percent for the average IS (with mean demand deposits at PKR 64 million, 19.8 percent is economically significant).
  - Within mixed banks (bank fixed effects): positive and statistically significant IS effect persists — Islamic operations performed better than conventional operations of the same bank during the panic.
- Credit rating effect:
  - Bank credit rating (numerical 1 to 10 scale) is positively related to deposit retention: a one-notch improvement in the bank’s credit rating results in a deposit withdrawal smaller by 2.1 percent.
  - IS effect remains significant after controlling for credit rating, suggesting an additional “religious branding” effect.

### Main empirical findings — lending (transmission to real economy)
- Parsimonious lending results:
  - IBIs on average extended 77 percent more credit than conventional counterparts during the liquidity crisis (parsimonious model).
  - Change in deposits is positively and significantly associated with new lending (statistically significant at the 1 percent level) in early specifications.
- Sensitivity to deposit shocks:
  - Interaction results: for IBIs, lending decisions are less sensitive to change in deposits — among banks subject to the same deposit withdrawals, IBIs reduce supply of new loans less than CBs.
- Conditional results with controls:
  - After controlling for bank, borrower, and loan characteristics, IBIs still grant more new loans during the crisis.
  - Bank characteristics positively related to new loan volume: capital adequacy, size of branch network, age, liquidity.
  - Bank characteristics negatively related to loan volume: NPLs, profitability, reliance on non-deposit funding.
- Within-bank contrasts:
  - Bank fixed effects: within the same bank, Islamic operations extended larger loan volumes during the financial panic compared to conventional operations.
- Borrower-fixed-effect specification:
  - After controlling for borrower fixed effects and bank unobservables:
    - Banks that experienced an increase in deposits during the financial panic raised the supply of loans (coefficient statistically significant at the 10 percent level).
    - Loan supply by ISs does not appear more sensitive to change in deposits: linear combination of estimated coefficients on deposit change and interaction with IS dummy yields 0.022 with a standard error of 0.021.
- Interpretation:
  - IBIs can help shield the real sector from financial sector shocks by maintaining credit supply; this may occur despite a “religiosity premium” (IBIs possibly charging higher financing rates and offering lower deposit returns).
  - Regulatory context: SBP sets a floor rate of interest for conventional deposits, which is not applicable to profit and loss sharing Islamic deposits.

### Mechanisms and interpretation
- Potential channels making IBIs more resilient:
  - Religious branding: depositor affiliation and religiosity reduce propensity to withdraw; survey evidence supports depositor loyalty and willingness to absorb losses.
  - Contractual design: PSIAs (profit & loss sharing accounts) operate like puttable equity and can provide an extra cushion when losses occur; asset-backed financing limits speculative exposures.
  - Bank fundamentals: IBIs exhibit stronger average balance-sheet metrics in the Pakistan sample (reported ratios below).
- Countervailing risks:
  - Scarcity of shariah-compliant liquidity instruments and underdeveloped Islamic money markets can raise vulnerability; absence of lender-of-last-resort facilities in some jurisdictions can compound risk.
  - IBIs hold higher cash buffers in practice (cost-inefficient), offset by regulatory relief (lower SLR for Islamic banks in Pakistan as of May 22, 2008).

### Comparative bank strength (reported ratios)
- Capital to asset ratios: IB 23%, IS 12%, CB 7%.
- NPL ratios: IB 1.27 percent, IS 1.26 percent, CBs 13.37.
- SLR and CRR (as of May 22, 2008):
  - CRR for all banks: 9 percent of demand and time liabilities (with less than 1 year tenor).
  - SLR: 9 percent for Islamic banks, 19 percent for conventional banks.

### Robustness checks and alternative specifications
- Alternative loan aggregation: defining loans as Firm-Bank pairs (aggregating multiple loans of a firm from the same bank) — sample of 17,606 loans to 15,224 borrowers — results maintain sign and significance patterns supporting primary findings.
- Results robust to controls for bank fundamentals and to specifications that include bank fixed effects and borrower fixed effects.
- Limitations in statistical power noted for some bank-characteristic coefficients (small sample size concerns).

### Policy implications and recommendations
- Financial inclusion:
  - Greater financial inclusion through faith-based financial institutions (Islamic banking) may enhance banking system stability and dampen transmission of financial shocks to the real economy in predominantly Muslim countries.
- Regulatory and macroprudential considerations:
  - Consider features of Islamic banking in broader policy design:
    - Equity-like profit and loss sharing saving accounts that share profits and losses with depositors.
    - Asset-backed financing that may limit excessive borrowing and debt overhang.
  - Recognize interaction between religious branding and market discipline: depositor loyalty can provide resilience but may also mask underlying vulnerabilities if governance and monitoring are weak.
- Future regulatory focus:
  - Ensure development of shariah-compliant liquidity instruments and money market infrastructure to mitigate IBIs’ liquidity management constraints.
  - Monitor displaced commercial risk and adequacy of profit equalization reserves (PER) and investment risk reserves (IRR) as smoothing mechanisms.

### Limitations and suggested avenues for future research
- Scope limitations:
  - Analysis focused on banking sector resilience; does not cover non-deposit-taking or non-banking financial institutions.
- Research agenda:
  - More theoretical and empirical work is needed on how pursuit of social or ethical objectives by financial actors affects financial stability and real economy transmission.
  - Investigate broader financial-sector-wide implications and cross-country comparisons.

### Key statistics and dates (for reference)
- Panic window used in regressions: 27-Sep-2008 to 14-Nov-2008.
- Aggregate deposit loss during panic: over Rs. 130 billion or 4 percent (reported).
- Alternative loan data period (monthly data availability): 30-Sep-2008 to 30-Nov-2008.
- Exchange-rate contextual note: 1 USD ~ PKR 79 in October 2008.

*Source: _wp1541 (extracted content supplied).*

### 1. Introduction....................................................................................... 3

### 1. Introduction

### Context and research question
- Maturity transformation—the conversion of short term liabilities into long term assets—is a core function of banks and creates liquidity mismatch risk that can lead to bank runs and insolvency (Diamond and Dybvig (1983)).
- The paper examines whether a financial panic affected deposit and lending behavior of Islamic and conventional banks differently in Pakistan, where no explicit deposit insurance exists.
- Two-step approach:
  - Determine whether rumor-driven media coverage in September–October 2008 produced deposit withdrawals of different magnitudes across banks, creating heterogeneous liquidity shocks.
  - Assess whether the change in deposits affected banks’ lending behavior, and whether that effect differed for Islamic versus conventional banks, controlling for banks’ financial strength and using fixed effects (bank fixed effects for deposit changes; borrower (firm) fixed effects for lending).

### Event: the September–October 2008 financial panic in Pakistan
- Panic timeline and magnitude:
  - Demand deposits in the banking sector fell continuously over a period of seven weeks (from September 27, 2008 to November 14, 2008).
  - In three weeks (from September 27, 2008 through October 18, 2008), demand deposits declined by 4 percent or 131 billion PKR (Pakistani Rupee).
  - The panic was largely contained within about two months through central bank interventions aimed at restoring liquidity in the banking sector.

### Data and empirical strategy
- Data sources:
  - Banks’ weekly liquidity positions from the State Bank of Pakistan (SBP) to measure deposit movements during the crisis period.
  - Loan-level information from the Pakistani credit registry (Credit Information Bureau at the SBP); loan-level data refer to new granted loans (intensive margin of bank lending).
- Empirical identification:
  - Bank fixed effects used in specifications for deposit changes, including comparisons of Islamic and conventional operations of the same bank.
  - Borrower (firm) fixed effects used in lending specifications to control for credit demand shifts and to assess how bank credit to the same firm was affected by Islamic versus conventional banks.

### Key findings (as reported in the introduction)
- Deposit behavior:
  - Islamic banks (particularly Islamic bank branches and subsidiaries) experienced less deposit withdrawals than conventional banks during the panic.
  - Some Islamic bank branches and subsidiaries recorded increases in deposits during the panic, leading to an average inflow of deposits into Islamic branches and subsidiaries over the period.
  - This result holds across specifications with and without bank characteristics and suggests a role for “religious branding”.
- Lending behavior:
  - Islamic banks were better able to maintain the supply of credit to the real economy during the financial panic.
  - Lending by Islamic banks is less sensitive to the change in deposits than it is for conventional banks.
- Robustness and scope:
  - Results are robust to specifications that control for banks’ financial strength.
  - The loan data focus on the intensive margin (new loans), so the analysis pertains to newly granted credit.

### Relation to existing literature and contribution
- Builds on cross-country and multi-country findings:
  - Hasan and Dridi (2010): different business model helped IBIs shield profitability in early crisis stages; weak risk management affected profitability later; IBIs posted higher credit and asset growth.
  - Beck, Demirgüç-Kunt and Merrouche (2013): during the global financial crisis, Islamic banks had a higher intermediation ratio, higher asset quality, and were better capitalized.
  - Cihák and Hesse (2010): Islamic banks are financially stronger when small, but lose relative strength as they grow, reflecting credit risk management challenges.
  - Cihák and Hesse (2007): cooperative banks’ ability to use customer surplus improves stability relative to commercial banks.
  - Studies on transmission of financial shocks to real economy: Kapan and Minoiu (2013), de Haas and van Horen (2013), Giannetti and Laeven (2012), Ivashina and Scharfstein (2010), Khwaja and Mian (2008).
- Unique contributions of this paper:
  - Combines weekly deposit data and loan-level credit-registry data to gauge the impact of a rumor-driven financial panic on deposits and the consequent effects on lending for Islamic and conventional banks in an emerging market context.
  - Examines banks of all sizes and controls for bank fundamentals when assessing differential impacts.

### Policy relevance and interpretation
- Potential mechanisms:
  - Islamic banks could be more vulnerable because Islamic money markets are less developed and, in some countries, there is no lender of last resort facility; sharing profit and loss on investment deposits may increase depositor withdrawal risk (Khan and Ahmed (2001)).
  - Offsetting mechanism: religion-oriented depositors may be less sensitive to rate of return and to perceptions of bank financial strength, producing deposit stickiness for Islamic banks.
- Policy implication:
  - If Islamic banks are more resilient during periods of financial stress, then greater financial inclusion through Islamic banking could dampen the transmission of financial shocks to the real economy.

*Source: 1. Introduction (from the supplied PDF content).*

### 2. Background on Islamic Banking in Pakistan

### 2. Background on Islamic Banking in Pakistan

### Overview and structure of the banking system
- Pakistan has a bank-centric financial system where conventional and Islamic banks co-exist.
- Islamic banking development began in the 1980s and was spurred by favorable regulation in the late 1990s.
- The financial industry was permitted to set up:
  - regular conventional banks (CBs),
  - full-fledged Islamic banks (IBs),
  - Islamic banking subsidiaries or standalone Islamic banking branches (ISs) of existing CBs.
- Current counts: 13 CBs, 5 full-fledged IBs, and 15 CBs with Islamic branches (representing about 7 percent of total banking sector assets).
- For exposition, IBs and ISs are referred to collectively as Islamic banking institutions (IBIs).

### Key features and contractual differences of IBIs vs CBs
- IBIs cannot engage in interest-based lending or borrowing.
- Ideal financing form: profit and loss sharing (PLS) contracts, but in practice IBIs predominantly use asset-backed fixed return arrangements:
  - deferred payment sales (Murabahah),
  - operating leases (Ijarah),
  - diminishing musharakah.
- In Pakistan, Murabahah, Ijarah and diminishing musharakah form over 80 percent of total financing provided by Islamic banks; the share of strictly PLS contracts remains small.
- Asset-backed arrangements render IBIs relatively less susceptible to financing crises.

### Liquidity management and regulatory treatment
- Challenges for Islamic banks:
  - scarcity of shariah-compliant liquidity management instruments,
  - small and underdeveloped Islamic money markets,
  - lack of lender of last resort facility in some countries (absence of an alternative to the discount rate).
- These constraints lead IBIs to maintain higher cash reserve buffers (cost-inefficient; equivalent to a tax on financial intermediation).
- To offset this disadvantage, the State Bank of Pakistan (SBP) requires Islamic banks to keep a lower level of Statutory Liquidity Reserves (SLR) than CBs.
- As of May 22, 2008:
  - Cash Reserve Requirement (CRR) for all banks was 9 percent of demand and time liabilities (with less than 1 year tenor).
  - Statutory Liquidity Reserves (SLR) were 9 percent and 19 percent of total demand and time liabilities for Islamic and conventional banks respectively.
- Empirical observation: banks in Pakistan, especially Islamic banks, hold significantly lower excess reserves (in percent of total deposits) than in other jurisdictions.

### Liability structure: current accounts, PSIAs, and implications
- Liability composition of IBIs mainly consists of:
  - current accounts (transaction deposits) — withdrawable on demand and raised on interest-free debt contract (Qard),
  - profit & loss sharing saving and investment accounts (PSIAs),
  - equity.
- IBIs must maintain a certain fraction of demand deposits as reserves in cash or shariah-compliant government bonds (sukuk); thus current accounts can be affected by financial panics similarly to CBs.
- PSIAs:
  - Typically mobilized via mudarabah contracts with mudarabah account holders (MAHs).
  - MAHs share profit or loss in proportion to investment; neither return nor principal is guaranteed.
  - MAHs are not considered creditors and are not first claimants in insolvency; MAHs’ funds are invested for a fixed period and withdrawals may be allowed subject to contract terms — PSIAs can be viewed as limited-duration equity investments or puttable equity instruments.
  - Depositors (MAHs) face risk of capital loss but PSIAs provide an extra line of protection to Islamic banks when losses occur.
- Agency and governance issues:
  - MAHs lack shareholder voting rights and board access; monitoring depends on shareholders or the ability to divest.
  - Two factors can mitigate moral hazard:
    - redeemable nature of PSIAs creates commercial pressure on IBIs to pay competitive returns to retain funds,
    - alignment of interests between MAHs and shareholders when funds are commingled.
- Displaced commercial risk:
  - If current profits are insufficient to pay market-based returns to MAHs, IBIs may shift losses to equity holders, exposing them to displaced commercial risk.
  - To manage this, IBIs maintain profit equalization reserves (PER) and/or investment risk reserves (IRR) to smooth returns to MAHs.
  - In PER, both IBIs and MAHs contribute some share of their profits; in IRR only the profit share of MAHs is set aside.

### Asset-side features and stability considerations
- Islamic contracts limit speculative activities: restrictions on sale of debt, short sales, excessive uncertainty (gharar), and gambling (qimar).
- IBIs focus on credit for purchase of real goods and services; asset-backed financing links financial transactions to the real sector.
- These features promote market discipline and may make IBIs more stable:
  - depositors have stronger incentives to monitor banks,
  - reduced propensity for excessive risk-taking.
- Empirical and literature-based indicators suggest:
  - Depositors of Islamic banks may not be in “search of yield” and may be less sensitive to profit rates on deposits.
  - Overall, IBIs may be less likely to face bank runs than CBs due to stronger liquidity positions, better asset quality, and the ban on excessive speculation and certain derivatives.

### Comparative bank strength (key ratios reported)
- IBIs are on average younger than CBs but exhibit stronger balance-sheet metrics:
  - capital to asset ratios: IB 23%, IS 12%, CB 7%.
  - reliance on non-deposit funding: IBIs rely less than CBs (Islamic subsidiaries rely more).
  - non-performing loan (NPL) ratios: IB and IS 1.27 and 1.26 percent respectively; CBs 13.37.
  - liquid-to-total-asset ratios: both IBs and ISs have lower liquid-to-total asset ratios than CBs (contextual note: conventional banks in Pakistan maintained high liquid asset to total asset ratio because of attractive rates on government securities).

### Behavioral factors and empirical question
- Client motivations (branch proximity, product suitability) can dominate choice between IBIs and CBs; in such cases, deposit withdrawal behavior in a panic may not systematically differ by bank type.
- Religious preference may increase propensity to not abandon IBIs during stress; evidence from Malaysia: over 60 percent of Muslim customers of IBIs stated they would not withdraw deposits even if their bank did not pay any return.
- Conclusion: whether IBIs experienced greater deposit declines than CBs during financial panics is ultimately an empirical question to be examined via regression analysis.

---

### 3. The 2008 Financial Panic (summary of events and policy response)
- Timing and nature of the panic:
  - Occurred during the last week of September 2008; deposit falls occurred over a total period of seven weeks.
  - Induced largely by widespread rumors in media about potential failure of several financial institutions; rumors included allegations of bankruptcy of a few banks in early October 2008.
- Aggregate deposit impact:
  - Demand deposits fell from a pre-panic maximum of 3,397 bn PKR to a low of 3,233 bn PKR, or by 4.8 percent.
- Heterogeneity across banks:
  - Some banks experienced deposit outflows up to 46 percent, others experienced inflows up to 58 percent.
  - The shock was positive for 18 banks/branches and negative for 34 banks/branches.
  - CBs mainly experienced negative shocks; IBIs (in particular ISs) mainly experienced positive shocks.
- Market disruption:
  - Several banks failed to meet weekly cash reserve requirements.
  - Annualized overnight interbank lending rate reached 48 percent in October 2008.
  - Rumor mongering was assessed as a primary intensifier of liquidity pressure; other contributing factors cited included seasonal withdrawals for Eid and decline in foreign currency inflows.
- Central bank and policy actions:
  - SBP declared the banking sector “stable, safe and sound” and publicly communicated that the most affected banks had strong balance sheet positions and robust financial ratios.
  - SBP decreased the reserve requirement gradually from 9 percent to 5 percent in October 2008, and exempted short-term liabilities from the cash reserve requirement.
  - Reductions in CRR and SLR infused almost PKR 270 billion worth of liquidity into the banking system.
  - SBP conducted thirteen open market operations (OMO) to inject temporary liquidity in excess of PKR 300 billion.
  - Several banks borrowed through the discount window facility.
  - It took more than two months for deposits to reach the pre-crisis level and for confidence to be restored.

*Source: _wp1541 - 2. Background on Islamic Banking in Pakistan*

### 4. Data Analysis and Econometric Specifications

### 4. Data Analysis and Econometric Specifications

### Data set and financial panic period
- Weekly deposit data for all banks operating in Pakistan between July 11, 2008 and January 2, 2009.
- “Financial panic” period defined as the period of large deposit outflows spanning seven weeks starting on September 27, 2008 and ending on November 14, 2008.
- Other bank-level variables available and discussed below.
- 1 USD ~ PKR (Pakistani Rupee or Rs.) 79 in October 2008.

### Deposit-change regression (Specification (1))
- Objective: evaluate impact of the financial panic on deposits of Islamic and conventional banks.
- Dependent variable:
  - Δ log(deposits) over the seven-week liquidity crisis period for bank i, branch type j (difference in log-deposits between November 14, 2008 and September 27, 2008).
- Key regressors:
  - IB_i: indicator variable for Islamic Banks.
  - IS_ij: indicator variable for Islamic Banking Subsidiaries/Branches of mixed CBs.
  - Omitted category: CB (conventional banks).
  - X_ij: matrix of bank characteristics (see list below).
- Bank characteristics (measured as of June 30, 2008 to avoid endogeneity with the run):
  - CAMEL-type fundamentals (CAMEL = Capital Adequacy, Asset Quality, Management, Earnings, Liquidity).
  - Size (log of total assets).
  - Bank reach (log of number of branches).
  - Bank age (the number of quarters since start of operations).
  - Credit rating.
  - Non-deposit funding to total funding (proxy for funding-structure differences).
- Fixed effects:
  - Some specifications include bank fixed effects to compare Islamic vs. conventional operations within the same mixed bank.
- Reported summary observation:
  - Only ISs gained deposits during the financial panic, while IBs and CBs lost deposits during this period.
- Estimation strategy:
  - First estimate specification (1) with an intercept and dummies for IBs and ISs to compare with CBs, then add covariates X_ij to control for bank characteristics.

### Loan-level lending regression (Specification (2))
- Objective: check whether the liquidity shock was absorbed by banks or transmitted to the real economy via a credit crunch by analyzing newly granted loans.
- Data augmentation:
  - Bank balance-sheet data augmented with granular, loan-level data from the Pakistani credit registry (matched by bank).
  - Focus on new granted loans (intensive margin of bank lending).
- Dependent variable:
  - ln(new loan_l) granted by bank i, branch type j, to borrowing firm k during the financial panic.
- Key regressors:
  - Δ log(deposits): change in log of deposits between the beginning and the end of the seven-week liquidity crisis.
  - IBI_ij: dummy for IBIs that is either IB_i or IS_ij.
  - X_ij, F_k, L_l: bank, borrower, and loan characteristics, respectively.
- Empirical challenge:
  - Need to disentangle credit supply effects (bank-driven) from credit demand effects (firm-driven), since events causing liquidity crunch often affect credit demand.

### Identification strategy and borrower fixed effects (Specification (3))
- Identification approach follows Khwaja and Mian (2008):
  - Use borrower (firm) fixed effects to test how bank lending to the same firm—borrowing from different banks (Islamic vs. conventional, more vs. less affected by the shock)—was affected during the liquidity crisis.
- Specification (3):
  - Identical to specification (2) except it adds a full set of borrower (firm) fixed effects to isolate credit supply effects from demand effects.

### Additional contextual and institutional details
- Credit rating agencies in Pakistan:
  - Pakistan Credit Rating Agency (PACRA, sponsored by Fitch Ratings among others).
  - JCR-VIS (co-sponsored by Japan Credit Rating Agency).
  - Both agencies rate banks on rating scales ranging from AAA (highest credit worthiness) to D (default); rating scales are domestically calibrated and are not internationally comparable.
  - AAA rating definition quoted: ‘Highest credit quality. The risk factors are negligible, being only slightly more than for risk-free Government of Pakistan’s debt’.
  - All banks operating in Pakistan are required to be rated by one of the two credit rating agencies.
- Reference to loan-margin literature:
  - For analyses of the extensive margin of lending, see Jimenez et al. (2014, 2012).

*Source: _wp1541 - 4. Data Analysis and Econometric Specifications*

### 5. Results

### 5. Results

### 5.1. Islamic vs. conventional banks
- Parsimonious specification (column I) regressing change in log-deposits during the crisis on IB and IS dummies:
  - Constant term is negative and statistically significant, suggesting that CBs experienced a large withdrawal of demand deposits during September-October 2008 (to the tune of 7 percent).
  - Coefficient for full fledged IBs is statistically insignificant, implying demand deposit outflows from IBs were similar to those out of CBs.
  - Estimated coefficient on ISs is positive and statistically significant, indicating an increase in demand deposits during the period to the tune of 19 percent for the average ISs.
  - With mean demand deposits at PKR 64 million, 19.8 percent is economically significant.
  - Conclusion: Islamic nature per se is not necessarily the main factor in the bank run; other bank characteristics may be important.

### 5.2. Do bank fundamentals matter?
- Extended specification (Table 3, column II) adds pre-crisis bank characteristics:
  - Capital to asset ratio: positive but statistically insignificant.
  - Log of number of branches: statistically insignificant — branch network size does not affect the decision to withdraw demand deposits.
  - Bank age (number of quarters in operation): appears irrelevant for change in deposits.
  - CAMEL-related controls (asset quality NPLs, earnings return on average assets, liquidity liquid-to-total assets, share of non-deposit funding, bank size): none of these individual controls matter for the extent of the bank run in this cross-section; small sample size may drive insignificance.
  - Even after controls, coefficient on ISs remains statistically significant at the 10 percent level, suggesting Islamic banking branches/subsidiaries experienced deposit inflows conditional on pre-crisis bank characteristics.

### 5.3. Does independent information help banks?
- Add bank credit rating (Table 3, column III):
  - Credit rating constructed as numerical rating 1 to 10 (1 = lowest BBB- equivalent, 10 = highest AAA equivalent).
  - Bank credit rating is positively related to bank-specific change in deposits: a one-notch improvement in the bank’s credit rating results in a deposit withdrawal smaller by 2.1 percent.
  - When controlling for credit rating, the estimated coefficient on ISs remains statistically and economically significant.
  - Interpretation: credit rating captures “flight to quality” effects; residual IS effect suggests possible “religious branding” effects associated with Islamic bank branches and subsidiaries.

### 5.4. Bank or her (religious) depositor?
- Bank fixed effects within mixed banks (Table 3, column IV):
  - Positive and statistically significant coefficient on ISs persists, indicating Islamic operations performed better than conventional operations of the same bank during the panic.
  - Interpretation: “religious branding”, rather than other observable bank characteristics, might have helped ISs fare better.
- Survey evidence (Pakistan):
  - 86 per cent of business customers choose IBIs because “owners of the company are religiously motivated”.
  - 72 per cent IBI customers believe “Islamic banking practices are in line with the religious and moral philosophy of the company”.
  - 62 per cent were willing to pay more for Islamic banking services because these were shariah compliant.
  - 98.4 per cent of respondents: idea that “Islamic products/services offer religious satisfaction and comfort” was an important reason to switch to Islamic banking.
  - Willingness-to-absorb-loss survey results:
    - “Are you willing to deposit money into a Islamic savings account even if there is a chance of loss?” (Strongly Agree/Agree 62%, Strongly Disagree/Disagree 21%, Neutral 17%).
    - “Would you withdraw your money from an Islamic savings account if the Bank announces loss?” (Strongly disagree/Disagree 55%, Strongly Agree/Agree 21%, Neutral 24%).
    - “In the event of loss would you withdraw your money from an Islamic savings account and deposit in a fixed return account with a conventional bank?” (Strongly disagree/Disagree 57%, Strongly Agree/Agree 15%, Neutral 28%).
  - Conclusion: a devoted and loyal depositor base likely makes IBIs less prone to bank runs and more resilient during panics; evidence indicates this was the case during September/October 2008.

### 5.5. Did the liquidity shocks affect the supply of credit?
- Estimating specifications (2) and (3) reported in Table 4 (columns I-IV), with clustered standard errors by bank where degrees allow:
  - Parsimonious model (column I): IBIs on average extended 77 percent more credit than their conventional counterparts during the liquidity crisis.
  - Column II: change in deposits positively and significantly supports new lending (statistically significant at the 1 percent level).
  - Interaction of IBI dummy with change in deposits: for IBIs, decision to grant new loans is less sensitive to change in deposits — among banks subject to the same deposit withdrawals, IBIs reduce supply of new loans less than CBs.
  - Column III (controls added): after controlling for bank, borrower, and loan characteristics:
    - IBIs still appear to grant more new loans during the crisis compared to CBs.
    - Change in deposits effect is no longer statistically significant.
    - Bank characteristics positively affecting new loan volume: capital adequacy, size of branch network, age, liquidity.
    - Bank characteristics negatively related to loan volume: NPLs, profitability, reliance on non-deposit funding.
    - Borrower side: larger borrowers receive more loans.
  - Column IV (split IBI into ISs and IBs): both ISs and IBs granted more new loans than CBs during the financial panic; effect holds after controls.
  - Column V (bank fixed effects): within the same bank, Islamic operations extended larger loan volumes during the financial panic compared to conventional operations.
  - Addressing demand-side effects: add borrower fixed effects and interaction term between ISs and change in deposits.
    - Table 4, column VI: after controlling for unobservable bank characteristics and credit demand:
      - Banks that experienced an increase in deposits during the financial panic raised the supply of loans (coefficient statistically significant at the 10 percent level).
      - Loan supply by ISs does not appear more sensitive to change in deposits during the liquidity crunch: linear combination of estimated coefficients on deposit change and interaction with IS dummy yields 0.022 with a standard error of 0.021.
    - Interpretation: IBIs can help shield the real sector from financial sector shocks; IBIs may operate less efficiently (charging a “religiosity premium” via higher financing rates and lower deposit rates) but this may allow continued credit provision during liquidity crunches.
    - Regulatory note: SBP sets a floor rate of interest for conventional deposits, which by definition is not applicable on profit and loss sharing Islamic deposits.

### 5.6. Robustness
- Alternative unit of analysis: replace dependent variable “Log of New Loans granted during Financial Panic” with “Change in Log of Lending to a Firm during Financial Panic”.
  - Define a loan as a Firm-Bank pair (aggregate multiple loans of a firm from the same bank into a single loan) following Khwaja and Mian (2008).
  - Sample comprises 17,606 loans to 15,224 borrowers.
  - Equation estimated (and variant with firm fixed effects) where:
    - dependent variable is natural log of change in loan to firm k granted by bank i, branch type j during the crisis period;
    - change in log of deposits between beginning and end of seven-week liquidity crisis;
    - IBIij is dummy for IBIs (IBi or ISij); Xij are bank characteristics.
  - Results presented in Table 5: coefficients differ in size compared to previous treatment, but sign and significance of coefficients of prime interest remain largely the same and support earlier results.

*Source: _wp1541 - 5. Results.*

### 6. Conclusions

### 6. Conclusions

### Main findings
- The study uses detailed data on bank balance sheets and their granted loans from Pakistan to examine differential behavior of Islamic and conventional banks during a financial panic that occurred during September-October 2008.
- Islamic branches of mixed banks are less prone to the risk of deposit withdrawals during panics, both unconditionally and conditional on bank characteristics.
- The Islamic operations of the mixed banks on average attracted deposits during the panic compared to the banks’ conventional operations, which suggests a role for religious branding.
- Islamic banks were more likely to grant new loans during the liquidity crisis and their lending decisions were less sensitive to changes in deposits relative to their conventional counterparts.
- The transmission of financial shocks to the real economy may be partially dampened if faith-based financial institutions are in operation (in a muslim majority country in this case).

### Mechanisms and interpretation
- Religious branding: Islamic operations attracting deposits during the panic point to a branding effect tied to religiosity and affiliation between bank and customer.
- Customer behavior and contract features: IBIs’ customers’ willingness to share in losses and their higher degree of affiliation with IBIs (because bank and customer share the same set of values) are highlighted as possible reasons for greater resilience.
- Financial fundamentals: Findings are consistent with evidence that stronger bank fundamentals (better capital, liquidity) inhibit transmission of financial sector shocks to the real economy.
- Interaction with parent bank fragility: Results indicate ISs (Islamic subsidiaries/branches) appear to fare better than their conventional parent, which contrasts with literature documenting a negative role of parent bank fragility on subsidiary lending.
- Contractual characteristics: Equity-like profit and loss sharing saving accounts provide an additional cushion to banks’ capital and share profits and losses with depositors. Asset-backed financing may place a natural cap on excessive borrowing and ensuing debt overhang.

### Policy implications
- Greater financial inclusion of faith-based groups, for instance through Islamic banking, may enhance banking system stability.
- Some features of Islamic banking may be considered for adaption and adoption in conventional banking, including:
  - Equity-like profit and loss sharing saving accounts that share profits and losses with depositors.
  - Asset-backed financing that may limit excessive borrowing and debt overhang.
- Marcropudential policy and banking regulation: Findings have important implications for macroprudential policy and banking regulations given differential resilience properties observed.

### Limitations and suggestions for future research
- The analysis in this paper is limited to the resilience of the banking sector; future research can include the overall financial sector by incorporating non-deposit taking or non-banking financial institutions.
- More theoretical and empirical research is needed on the effect of pursuit of social or ethical objectives by players/decision-makers on the financial sector and real economy.

### Key statistics and dates referenced in conclusions and related estimations
- Financial panic period analyzed: September-October 2008.
- Liquidity crisis period used in regressions: 27-Sep-2008 to 14-Nov-2008.
- Banking sector in Pakistan lost over Rs. 130 billion or 4 per cent of deposits during the panic due to rumors regarding viability of some financial institutions.
- Alternative loan data period noted where applicable: 30-Sep-2008 to 30-Nov-2008 (loan data available only on monthly frequency).

*Source: _wp1541 - 6. Conclusions*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp1541.pdf_