Pith. sign in

REVIEW 5 major objections 6 minor 8 references

The Determinants of Net Interest Margin in the Turkish Banking Sector: Does Bank Ownership Matter?

T0 review · 5 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Bank ownership changes which factors set Turkish banks' net interest margins, a dynamic panel study of 23 banks finds.

desk verdict Valuable Turkish bank dataset and a plausible full-sample GMM, but the ownership-heterogeneity claim rests on missing Chow tests and a subsample GMM that fails its own diagnostics. read the letter →

arxiv 2506.04384 v2 pith:MJWIK5DN submitted 2025-06-04 econ.GN q-fin.EC

Fatih Kansoy This is my paper
classification econ.GNq-fin.EC
keywords netinterestmarginTurkishbankingbankownershipdynamicpanelGMMcreditriskinflationcompetitionfinancialintermediation
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper asks whether a bank's ownership—foreign, state-controlled, or private—changes how the usual determinants of net interest margin work in Turkey, using quarterly data on 23 commercial banks from 2001 to 2012. It argues that they do: credit risk, bank size, market concentration, and inflation have ownership-specific effects, while implicit interest payment, operating diversity, and operating cost move margins the same way in all three groups. If the result holds, bank managers and regulators cannot rely on a single pooled model of margins; they must know who owns the bank. The study also finds that better-managed banks charge lower margins and that inflation raises margins, which makes price stability a policy lever for financial intermediation costs.

What carries the argument

The load-bearing machinery is a dynamic panel model of net interest margin in which the lagged margin appears as a regressor, estimated with a system generalized method of moments that instruments the lagged dependent variable with its own deeper lags in levels and differences. The model is applied to the full 23-bank sample and then to three ownership subsamples, with the dealership model of interest margins supplying the variable set: risk aversion, credit risk, operating cost, size, liquidity, management quality, implicit interest payment, deposit growth, operation diversity, concentration, GDP growth, and inflation.

What would settle it

Re-estimate the state-owned and foreign-owned subsamples with a collapsed or otherwise valid instrument set and check whether credit risk, size, and inflation keep their ownership-specific signs and significance; if the state-bank credit-risk coefficient loses its positive sign or the foreign-bank inflation effect disappears under valid instruments, the paper's central heterogeneity claim is falsified.

Watch

Extended reading notes

Core claim

The paper's central discovery is that the net interest margin in Turkish commercial banking is not governed by one uniform equation. Estimating a dynamic panel model separately for foreign-owned, state-controlled, and private banks shows that credit risk (non-performing loans over total loans) lowers margins for foreign and private banks but raises them for state banks; bank size and capitalization raise margins only for foreign banks; market concentration raises margins only for private banks; and inflation raises margins only for foreign banks. By contrast, implicit interest payment, operating diversity, and operating cost have statistically significant coefficients with the same sign across all ownership groups. The overall-sample results also show that management quality lowers margins, inflation raises them, and credit risk lowers them, with operating cost and implicit interest payment loading positively.

Load-bearing premise

The ownership comparison stands on the system GMM estimates for each subsample being consistent, which requires valid instruments and no serial correlation; the state-bank subsample fails the instrument-validity test and the foreign-bank subsample lacks the expected first-order autocorrelation, so those two groups' coefficients may not be trustworthy.

Editorial extensions

If this is right

  • If ownership-specific coefficients are real, pooling all Turkish banks into one margin equation can mislead: a determinant that matters for private banks may appear irrelevant or wrong-signed in the aggregate.
  • Management quality is a reliable margin reducer in every ownership group, so policies that improve bank efficiency should lower intermediation costs across the sector.
  • Inflation raises margins, and in the ownership estimates it is significant only for foreign banks; this makes price stability a targeted policy tool.
  • Operating cost and implicit interest payment push margins up uniformly, so banks' cost structure and fee practices have the same margin consequences regardless of owner.
  • Bank size matters only among foreign banks in this period, suggesting foreign banks use scale to set higher margins rather than to exploit scale economies.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension would re-run the same ownership-split design on post-2012 Turkish data, where inflation targeting, digital banking, and foreign-bank entry have changed the competitive landscape; the stable homogeneity of operating cost and diversity effects could be checked against the new period.
  • The heterogeneity result implies that cross-country studies that pool banks of different ownership structures may average away opposing signs; ownership interactions deserve a place in international margin regressions.
  • If state banks respond to credit risk with higher margins while private banks accept lower margins to gain market share, then policy responses to non-performing loans should differ by ownership type—for example, recapitalising state banks versus monitoring private banks' risk appetite.
  • The paper's instrument-validity failure for state banks suggests its strongest ownership comparisons are the foreign/private ones; a direct replication with a collapsed instrument set would show whether the state-bank results survive.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

5 major / 6 minor

Summary. The paper studies the determinants of the net interest margin (NIM) for the Turkish commercial banking sector using quarterly bank-level data from 2001Q4 to 2012Q1. It estimates static panel models and a dynamic system GMM model, and then splits the sample by ownership (foreign, state, private) to ask whether the determinants of NIM differ across ownership groups. The main reported findings are that credit risk, bank size, market concentration, and inflation have ownership-specific effects, while implicit interest payment, operation diversity, and operating cost are claimed to be homogeneous across ownership groups. The paper also includes several robustness checks and policy recommendations based on the estimated coefficients.

Significance. If the ownership-heterogeneity result were well established, it would be relevant to Turkish bank managers and regulators: it would imply that ownership status must be taken into account when predicting or steering net interest margins, and that policies affecting credit risk, size, concentration, or inflation would have different margin consequences in foreign, state, and private banks. The paper has useful raw ingredients: a quarterly dataset covering nearly all Turkish commercial banks over a decade, a clear dynamic panel setup, and reporting of specification tests and robustness exercises. However, the central claim that coefficients differ by ownership is not supported by the evidence actually presented: the Chow tests referenced in the text are not in the paper, the state-bank subsample fails the Sargan test at 5%, and the ownership subsamples do not partition the stated full sample. As a result, the paper's main contribution is currently an assertion rather than an established empirical result.

major comments (5)
  1. [Section 4.2, Tables 10-12] The ownership-heterogeneity conclusion rests on visual comparison of subsample coefficients in Table 5, but the Chow tests that the text says are reported in Tables 10-12 do not exist in the paper: those appendix tables are literature reviews, not coefficient-equality tests. No F-test, Wald test, or interaction-term test of cross-group coefficient equality is reported anywhere. Therefore the statement in Section 4.2 that 'according to Chow test results the coefficients do not have the same affect for the three ownership groups' is not statistically substantiated.
  2. [Table 5, state-owned subsample] For the state-owned subsample (N=3 banks, 120 observations), the Sargan test p-value is 0.0278, so the overidentifying restrictions are rejected at the 5% level. System GMM estimates with invalid instruments are inconsistent, which means the state-bank coefficients, including the positive credit-risk effect on RBD, cannot be trusted. Since the credit-risk coefficient is one of the variables claimed to vary by ownership, this directly undermines a headline result.
  3. [Table 5, foreign-owned subsample] For the foreign-owned subsample, the Arellano-Bond AR(1) test p-value is 0.1284. In a correctly specified first-differenced dynamic panel with i.i.d. level errors, one expects strong negative first-order autocorrelation in the differenced residuals, so a non-rejection at this p-value is a specification red flag, indicating possible instrument weakness or misspecification. Consequently, the foreign-specific results, including the significant positive LOGTA coefficient, are not reliable evidence of ownership heterogeneity.
  4. [Section 3.1 and Table 5] The ownership subsamples do not add up to the full sample. Table 5 reports 10 foreign, 3 state, and 16 private banks with 288, 120, and 498 observations, respectively; these sum to 29 banks and 906 observations, whereas Section 3.1 and Table 4 state that the full sample contains 23 commercial banks and 920 observations. Either the ownership classification or the sample definition is inconsistent, and this must be resolved before any subsample comparison can be interpreted.
  5. [Abstract and Section 4.2] The abstract claims that the impacts of implicit interest payment, operation diversity, and operating cost are homogeneous across all banks, but Table 5 contradicts this. IIP is statistically significant for foreign banks (0.376***) and private banks (0.425***) but insignificant for state banks (0.0308, s.e. 0.104). Operating cost is insignificant in all three ownership subsamples while significant in the full-sample GMM, so the meaning of 'homogeneous' is not established by the reported estimates.
minor comments (6)
  1. [Section 4.1 and Table 6] The text states that the Breusch-Pagan LM test 'rejects the null hypothesis' that POLS is appropriate, but Table 6 reports chibar2(01)=0.00 with Prob>chi2=1.0000, which fails to reject the null. This internal inconsistency in the model-selection narrative should be corrected.
  2. [Section 4.1, RBD discussion] The text describes the credit-risk result as 'A positive and significant relationship' between credit risk and NIM, but Table 4 reports RBD = -0.0343***. The coefficient is negative, and the surrounding discussion should be aligned with the reported sign.
  3. [Section 4.2, LQR discussion] The text says 'the effect of liquidity risk on NIM is positive for foreign banks,' but Table 5 reports LQR = -0.00827** for foreign banks; the positive effect appears for state banks (0.00276***). The narrative does not match the table.
  4. [Table 5, LOGTA private row] The standard error for LOGTA in the private-bank column is printed as (-0.175); this appears to be a typographical error and should be the positive value (0.175).
  5. [Section 4.1, equations (11)-(12)] The Fisher-equation decomposition would be clearer if the symbols gamma_L, gamma_D, and pi were defined in the text and if the derivation explicitly stated that these are real rates on loans and deposits and the inflation rate, respectively.
  6. [Throughout] There are several typographical and consistency issues, such as 'Turkey;s' in the Introduction, the citation style of 'Times, Financial' in the references, and the phrase 'This result verifies our hypothesis' where 'supports' would be more precise.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper is a standard empirical regression exercise whose coefficients come from data, not from definitions or self-citation.

full rationale

This is an empirical econometrics paper, not a derivation. The headline findings (which determinants are significant and whether their effects differ by ownership) come from estimating the dynamic panel model of Eq. (3) by system GMM on bank-level data. The explanatory variables are not defined in terms of the outcome in any way that forces the estimated coefficients: NIM is net interest income over total assets, while variables such as implicit interest payment (net non-interest income over total assets), credit risk (non-performing loans over total loans), and operating cost (operating expenses over total assets) are distinct accounting ratios, and the regression does not impose an identity linking them. The Section 4.1 Fisher-equation manipulation, Eqs. (11)-(12), is an arithmetic identity used to interpret a positive inflation coefficient; it does not generate the coefficient and no fitted value is derived from it. There are no load-bearing self-citations: the reference list contains no self-citations by Kansoy, no prior-work uniqueness theorem is invoked, and no functional form or ansatz is smuggled in by citation. The ownership-heterogeneity claim rests on separate GMM columns in Table 5 and on the paper's assertion that 'Chow Test results' in Tables 10-12 support coefficient differences; those appendix tables are in fact literature-review tables, and the state-bank subsample has a Sargan p-value of 0.0278. Those are specification and evidence problems, not circularity: the reported coefficients are not equal to their inputs by construction and no prediction is a renamed fit. Therefore no circular step is identified.

Assumptions & free parameters 6 free parameters · 4 assumptions · 0 invented entities

The paper is an empirical panel study, so the main parameters are the estimated regression coefficients, all fitted to data. It does not introduce new theoretical entities. The econometric assumptions listed are standard for dynamic panels but are partially contradicted by the reported diagnostics for the subsamples.

free parameters (6)
  • L.NIM coefficient = 0.228
    Estimated persistence coefficient in the full-sample system GMM, fitted to data.
  • RBD coefficient = -0.0343
    Fitted effect of credit risk (non-performing loans / total loans) on NIM.
  • IIP coefficient = 0.369
    Fitted effect of implicit interest payment on NIM.
  • DVRSTY coefficient = -0.379
    Fitted effect of operation diversity on NIM.
  • INF coefficient = 0.0295
    Fitted effect of inflation on NIM.
  • Ownership-specific coefficients (e.g., LOGTA for foreign, HHI for private, RBD for state) = LOGTA foreign 0.417; HHI private 0.0692; RBD state 0.0103
    Fitted subsample parameters that directly support the ownership-heterogeneity claim.
assumptions (4)
  • domain assumption The system GMM moment conditions hold: instruments are valid and errors are serially uncorrelated.
    Required for consistent estimates. Violated for the state-owned subsample, where the Sargan test p-value is 0.0278, and questionable for foreign banks, where AR(1) is insignificant.
  • domain assumption Explanatory variables are weakly exogenous.
    Standard GMM assumption; not directly tested in the paper.
  • domain assumption Bank ownership classification is fixed over 2001-2012.
    Turkish bank ownership changed during this period (e.g., transfers to the Savings Deposit Insurance Fund and foreign acquisitions); the paper splits the sample by ownership without accounting for time variation.
  • domain assumption Linear functional form relating NIM to its determinants.
    The model imposes linearity; ownership heterogeneity is captured only through sample splits, not interaction terms or nonlinear transformations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of The Determinants of Net Interest Margin in the Turkish Banking Sector: Does Bank Ownership Matter?." pith.science (2026). https://pith.science/paper/MJWIK5DN

@misc{pith2026250604384,
  author       = {Pith},
  title        = {Pith review of: The Determinants of Net Interest Margin in the Turkish Banking Sector: Does Bank Ownership Matter?},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MJWIK5DN}},
  note         = {Machine review of arXiv:2506.04384}
}
read the original abstract

This research presented an empirical investigation of the determinants of the net interest margin in Turkish Banking sector with a particular emphasis on the bank ownership structure. This study employed a unique bank-level dataset covering Turkey`s commercial banking sector for the 2001-2012. Our main results are as follows. Operation diversity, credit risk and operating costs are important determinants of margin in Turkey. More efficient banks exhibit lower margin and also price stability contributes to lower margin. The effect of principal determinants such as credit risk, bank size, market concentration and inflation vary across foreign-owned, state-controlled and private banks. At the same time, the impacts of implicit interest payment, operation diversity and operating cost are homogeneous across all banks

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

8 extracted references · 8 canonical work pages

  1. [1]

    Commercial bank net interest margins, default risk, interest-rate risk, and off-balance sheet banking

    Angbazo, Lazarus. 1997 . “Commercial bank net interest margins, default risk, interest-rate risk, and off-balance sheet banking. ”Journal of Banking & Finance21 (1): 55–87 . Aysan, Ahmet, Cagri Husnu Dalgic, and Murat Demirci

    work page 1997

  2. [8]

    Investors Drawn to Turkey;s Banking Sector

    “Investors Drawn to Turkey;s Banking Sector. ” Valverde, Santiago Carbó, and Francisco Rodríguez Fernández. 2007 . “The determinants of bank margins in European banking. ”Journal of Banking & Finance31 (7): 2043–2063. 31 Williams, Barry. 2007 . “Factors determining net interest margins in Australia: domestic and foreign banks. ”Financial Markets, Institut...

    work page 2007

  3. [1981]

    The determinants of bank interest margins: theory and empirical evidence

    “The determinants of bank interest margins: theory and empirical evidence. ”Journal of Financial and Quantitative analysis16 (4): 581–600. Horvath, Roman et al. 2009 . “The determinants of the interest rate margins of Czech banks. ”Czech Journal of Economics and Finance (Finance a uver)59 (2): 128–136. Kasman, Adnan, Gokce Tunc, Gulin Vardar, and Berna Okan

    work page 2009

  4. [2000]

    The determinants of bank interest rate margins: an international study

    “The determinants of bank interest rate margins: an international study. ”Journal of international Money and Finance19 (6): 813–832. Costa da Silva, G, J Oreiro, and L De Paula. 2007 . “Macroeconomic Determinants of Bank Spread in Brazil: An Empirical Evaluation. ”Federal University of Parana Working Paper(0066). Tan, Ms Tatum Blaise Pua. 2012.Determinant...

    work page 2007

  5. [2003]

    Assessing the success of reform in transition banking 10 years later: an interest margins analysis

    “ Assessing the success of reform in transition banking 10 years later: an interest margins analysis. ”Journal of Policy Modeling25 (3): 309–317 . Flamini, Valentina, Miss Liliana Schumacher, and Mr Calvin A McDonald. 2009 . “The determinants of commercial bank profitability in Sub-Saharan Africa. ”IMF Working Paper(9-15). Fungáčová, Zuzana, and Tigran Poghosyan

    work page 2009

  6. [2004]

    Factors explaining the interest margin in the banking sectors of the European Union

    “Factors explaining the interest margin in the banking sectors of the European Union. ”Journal of Banking & Finance28 (9): 2259–2281. Maudos, Joaquin, and Liliana Solis. 2009 . “The determinants of net interest income in the Mexican banking system: An integrated model. ”Journal of Banking & Finance33 (10): 1920–1931. Mody, Ashoka, and Maria Soledad Martin...

    work page 2009

  7. [2008]

    Determinants of bank interest margins in Central and Eastern Europe: A comparison with the West

    “Determinants of bank interest margins in Central and Eastern Europe: A comparison with the West. ”Economic Systems32 (2): 197–216. Dabla-Norris, Era, and Holger Floerkemeier. 2007 .Bank efficiency and market structure: what deter- mines banking spreads in Armenia?.: International Monetary Fund. Demirgüç-Kunt, Ash, and Harry Huizinga. 1999 . “Determinants...

    work page 2007

  8. [2010]

    Consolidation and commercial bank net interest margins: Evidence from the old and new European Union members and candidate countries

    “Consolidation and commercial bank net interest margins: Evidence from the old and new European Union members and candidate countries. ”Economic Modelling27 (3): 648–655. Laeven, Luc, and Ross Levine. 2007 . “Is there a diversification discount in financial conglomerates?” Journal of Financial Economics85 (2): 331–367 . Lerner, Eugene

    work page 2007

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.