Iranian Journal of Economic Research

Abstract Achieving the correct insight into the structure of connectedness and the spillover of volatilities between different stock exchange industries plays an important role in risk management and forming an optimal stock portfolio. Also, the analysis of inter-sectoral connectedness helps policy makers in designing policies that stimulate economic growth and implementing preventive measures to curb the propagation of systemic risk. In this regard, this article tries to use the data of 3370 trading days during the period of 1388/07/01 to 1402/06/31, encompassing 20 stock market industries (which constitute more than 80% of the Iranian stock market) and applying the connectedness approach based on the vector autoregression model with time-varying parameters (TVP-VAR), to estimate the systemic risk and volatility connectedness of the stock market network. In addition, we implement the minimum connectedness approach in the optimal stock portfolio and compared its performance with two other conventional approaches. The findings reveal that, first; the systemic risk in Iranian stock market is significant and has reached unprecedented figures of 80% in the last three years. Second, the four major export industries (petrochemicals, metals, mining and refining) experience the strongest pairwise connectedness, and among them, base metals appear as one of the most important transmitters of volatilities to the entire stock network. Thirdly, the stock portfolio based on the minimum connectedness method, compared to the minimum variance and minimum correlation methods, shows a better performance based on the criteria of cumulative return and hedge ratio efficiency.

Abstract The current study employed a comparative analytical framework to examine credit-default prediction. It relied on a comprehensive dataset of 56,965 loan contracts issued between 2019 and 2024 across the northern branches of Bank Melli Iran. Three modeling approaches were evaluated: traditional logistic regression and two ensemble machine learning methods—random forest (RF) and extreme gradient boosting (XGBoost). The analysis incorporated 29 predictive features categorized into three conceptual groups: loan contract characteristics (e.g., principal amount, repayment tenure, collateral type), borrower attributes (e.g., age, occupational profile, credit history), and institutional factors (e.g., branch location, branch type). Data preprocessing included outlier removal, text categorization, and the extraction of variables such as age and grace period. The models were evaluated under both baseline and optimized (hyperparameter-tuned) settings. The results showed that the machine learning models substantially outperformed the conventional logistic regression model. XGBoost delivered the highest discriminatory power (ROC-AUC = 99.73%), followed closely by RF (99.68%), whereas logistic regression lagged significantly (75.34%). On average, the AUC difference between the machine learning models and logistic regression was approximately 0.243, and statistical tests with 95% confidence intervals confirmed the significance of this gap. Overall, the findings provided strong evidence for the superior reliability of machine learning approaches in forecasting loan default.

Introduction

Although traditional econometric models such as logistic regression have long served as the foundation of credit scoring systems, their reliance on linearity assumptions and error independence limits their ability to capture the complex, nonlinear patterns typical of financial data. These limitations are further compounded by sensitivity to multicollinearity and distributional assumptions that are frequently inconsistent with real-world conditions. The present research aimed to address these shortcomings by conducting a rigorous comparative analysis of predictive methodologies within Iran’s banking sector—a context in which machine learning applications remain relatively underutilized despite widespread global adoption of artificial intelligence in finance. Specifically, the study intended to compare the performance of two ensemble learning techniques ( i.e., random forest and extreme gradient boosting or XGBoost), with that of conventional logistic regression in forecasting loan defaults using extensive real-world data from Bank Melli Iran. The methodological advantages of machine learning approaches arise from their ability to model complex nonlinear relationships without requiring predefined functional forms, to automatically capture variable interactions through hierarchical partitioning, to maintain robustness in the presence of outliers and non-normal distributions, and to detect subtle patterns in high-dimensional data that escape parametric detection. By systematically evaluating these capabilities, the current study tried to offer empirical evidence to support financial institutions in adopting more advanced and reliable risk modeling frameworks.

Materials and Methods

The selection of predictive models in this study is informed by theoretical foundations, empirical literature, and practical forecasting capabilities. Three distinct modeling approaches—random forest (RF), extreme gradient boosting (XGBoost), and logistic regression (LR)—were employed to evaluate their effectiveness in predicting loan defaults. As a widely used ensemble learning algorithm, random forest (RF) builds multiple decision trees using bootstrap aggregating and random subsets of observations and features. Each tree is trained independently, and final predictions are obtained through majority voting (classification) or averaging (regression). This structure reduces overfitting and improves generalization compared to single decision trees. XGBoost is an advanced gradient boosting algorithm known for its efficiency and high predictive accuracy. XGBoost constructs trees sequentially, with each new tree reducing the residual errors of the ensemble through gradient descent optimization. Rooted in the logistic function and formalized in modern choice modeling, logistic regression improves on linear probability models by mapping predictions to the [0,1] interval via a sigmoid transformation. Although valued for its interpretability, conventional econometric models such as logistic regression suffer from a series limitations, including linearity assumptions, limited interaction detection, multicollinearity sensitivity, and distributional constraints. These methodological constraints potentially compromise predictive performance in complex, non-linear domains such as credit risk assessment.

Results and Discussion

The machine learning models were evaluated under two configurations: a baseline setting using default parameters and an optimized setting using hyperparameter tuning. Hyperparameters—settings external to the model that are not learned from data—strongly influence predictive accuracy, computational efficiency, and generalization. Suboptimal hyperparameter selection can lead to underfitting or overfitting, thereby compromising model performance. Common optimization strategies include grid search, random search, and Bayesian optimization. Empirical evidence shows that random search is often more efficient in high-dimensional spaces (Bergstra & Bengio, 2012). Although default parameters may yield reasonable baseline performance, they rarely yield optimal performance (Probst et al., 2019). Prior research suggests that systematic tuning can increase accuracy by 10–20% (Hutter et al., 2019) and improve generalization (Liao et al., 2018). In this study, hyperparameters were optimized to maximize the area under the curve (AUC), a standard practice in credit risk modeling (Feurer et al., 2015). This approach can reduce prediction errors and enhance model stability in ensemble methods. The empirical results revealed substantial performance improvements through hyperparameter optimization. For the RF model, accuracy increased from 96% in the untuned configuration to 99% after tuning, with a notable reduction in false negatives and improved precision, albeit with a slight decline in recall for the default class. The optimized XGBoost model—using 375 trees, a maximum depth of 12, and a learning rate of 0.03—achieved the lowest false-negative and false-positive rates, offering an optimal balance between learning capacity and predictive accuracy. In contrast, logistic regression showed limited discriminatory power, with a recall of 0.16 and a ROC-AUC of 0.75, indicating inherent limitations in capturing the complex patterns associated with default events.




 
Random Forest Model (With Hyperparameter Tuning)


 
Random Forest Model (Without Hyperparameter Tuning)




 
(XGBoost) Model (With Hyperparameter Tuning)


 
(XGBoost) Model (Without    Hyperparameter Tuning)




 
Logistic Regression Model




Source: Research Results
Summary of Model Results




Model


State


ACCURACY


Precision (Bad)


Precision (Good)


Recall (Bad)


Recall (Good)


F1-Score (Bad)


F1-Score (Good)


ROC-AUC






RF


Unoptimized


97%


94/0


98/0


83/0


99/0


88/0


98/0


935/0




RF


Optimized


99%


97/0


99/0


94/0


99/0


95/0


99/0


9968/0




XGBOOST


Unoptimized


98%


96/0


99/0


85/0


99/0


90/0


99/0


9966/0




XGBOOST


Optimized


99%


97/0


99/0


88/0


99/0


92/0


99/0


9973/0




LR


-


96%


90/0


96/0


16/0


98/0


27/0


98/0


7534/0




Source: Research Results

Conclusion

The empirical results of this study demonstrates the superior predictive capabilities of machine learning methods—particularly XGBoost)—compared with conventional econometric approaches for estimating the probability of default (PD) in Bank Melli Iran’s loan portfolio. This performance gap primarily arises from machine learning algorithms’ ability to capture nonlinear relationships and latent structural patterns among default determinants—features that linear parametric models are unable to detect. Model precision was evaluated using several metrics, including confusion matrix analysis, total accuracy, and area under the ROC Curve (AUC). The findings indicated that machine learning models deliver substantially higher predictive precision and improved default detection rates. The optimized XGBoost model achieved outstanding performance (accuracy = 99%, AUC = 0.9973), far surpassing the logistic regression model’s ability to identify default cases (recall = 0.16). This distinct performance disparity strongly supports the research hypothesis regarding the comparative advantage of machine learning in PD estimation. Despite their superior predictive performance, the operational deployment of advanced machine learning techniques in financial institutions remains constrained by two key challenges: the computational complexity of hyperparameter optimization and the interpretability limitations inherent in black-box models. These limitations highlight the practical importance of developing hybrid frameworks that integrate the interpretive transparency of traditional methods with the predictive power of machine learning approaches. This research provided evidence of a paradigm shift in credit risk analytics, moving away from the long-standing reliance on conventional statistical models (such as logistic regression and linear probability models) toward machine learning methodologies. While prior studies using traditional techniques achieved moderate success, their limitations in handling imbalanced distributions and complex interaction effects have become increasingly apparent. The present findings align with international research trends and offer novel empirical evidence from Iran’s banking sector—demonstrating that well-tuned machine learning algorithms can achieve unprecedented levels of accuracy (99% accuracy compared with a 16% default identification rate for logistic regression).

Abstract Iran’s economy has encountered significant challenges in recent years, with the government debt to contractors emerging as one of the most urgent issues. This situation has negatively impacted Iran’s monetary and banking system, leading to several adverse consequences such as increased funding costs for banks, higher loan interest rates, excessive money supply, and a reduced capacity for banks to provide loans. A proposed solution is based on credit easing and endogenous money, which involves settling the government debt to contractors by making adjustments on the asset side of the Central Bank’s balance sheet. However, the practical implementation of this policy depends on the use of Central Bank resources, which raises concerns about a sudden increase in the money supply and potential negative effects on other economic variables, especially inflation. This uncertainty has led to doubts about the feasibility of such a strategy. The present research aimed to examine the fundamental principles and prerequisites for adopting a credit easing policy in Iran. The study also used stock-flow consistent models to evaluate the potential outcomes of implementation of the policy. The findings indicate that settling the government debt to banks by using the Central Bank resources results in an expansion of the monetary base and money supply, an increase in real GDP, and a reduction in both inflation and interest rates compared to the baseline scenario.
1.Introduction
Iran’s economy has been facing various problems in recent years. One significant issue is the government debt to contractors, which has adversely affected Iran’s economy, particularly the monetary and banking system. The government’s failure to settle its debts with contractors results in contractors being unable to repay loans taken from banks, leading to an increase in the banks’ non-performing loans. This predicament has precipitated several adverse consequences, including higher funding costs for banks, increased interest rates on loans, an uncontrolled surge in the money supply, and a diminished capacity for banks to provide loans. To address this challenge, some economists, emphasizing endogenous money, look to the quantitative policies applied by central banks in advanced countries like Japan and the United States. They have proposed a solution grounded in credit easing, which involves settling the government debt to contractors by making adjustments on the asset side of the Central Bank’s balance sheet.
2.Materials and Methods
In this method, the government issues bonds to settle its debt with contractors and provides these bonds to the contractors. The Central Bank then purchases these bonds by increasing the bank’s reserves. Since the Central Bank does not directly transact with individuals, it uses commercial banks as intermediaries to facilitate the payments. Consequently, the money supply and the monetary base increase immediately. However, if the contractors owe money to the banks, according to the law of reflux, the newly created money will quickly disappear. This method is largely similar to the second type of treasury bonds used by the Iranian government in recent years. Implementing this policy can reduce non-performing loans, curb the growth of the money supply, and prevent the recognition of illusory profits. It can also lower the level of overdue loans and improve banks’ balance sheets. Additionally, it can reduce the banks’ debt to the Central Bank, thereby lowering the cost of money and reducing loan interest rates. Moreover, the reduction in interest rates can lead to increased loan demand and, consequently, future growth in the money supply.
It should be noted that this policy leads to a change in the composition of the Central Bank’s assets, but it does not necessarily result in the growth of monetary base and money supply. However, since the policy relies on the use of Central Bank resources, concerns about a sharp increase in the money supply and potential adverse effects on macroeconomic variables, such as inflation, have always hindered its adoption. The present study used Stock-Flow Consistent (SFC) models to evaluate the effects of these policies on Iran’s macroeconomy. Having gained prominence since the 2007–2008 financial crisis, SFC models aim to integrate the real and financial sectors of the economy within a single framework. They help predict endogenous crises in the economy and enable modeling of the economy based on endogenous money. Therefore, SFC models were used to determine the effects of policies similar to credit easing to settle the government debt with contractors. The focus is on various economic variables, including the monetary base, money supply, and banks’ balance sheets in the monetary sector, as well as real GDP, economic growth, real consumption, inflation, and interest rates in the real sector of the economy.
3.Results and Discussion
The results indicate that settling the government debt to banks by using the Central Bank resources leads to an expansion in the monetary base and money supply, as well as an increase in real GDP and real consumption compared to the baseline scenario. However, the effect of this policy on economic growth completely dissipates after eight periods following its implementation, with the growth rate difference eventually tending towards zero. The graph below illustrates the difference in economic growth between the baseline scenario and the scenario where the government debt to contractors is settled using the Central Bank resources.
 
 
Figure 1. Difference in Economic Growth: The Baseline Scenario and the Government Debt Settlement Scenario
 
Source: The research analysis
According to the model’s results, implementing this policy leads to a long-term decrease in inflation by 0.23 percentage points.
Figure 2.  Difference in Inflation: The Baseline Scenario and the Government Debt Settlement Scenario
 
Source: The research estimations
Additionally, the results indicate that the policy can lead to a 0.53 percentage point decrease in the interest rate.
 
 
Figure 3. Difference in the Interest Rate: The Baseline Scenario and the Government Debt Settlement Scenario
 
Source: The research estimations
4.Conclusion
The model’s results indicated that the policy, despite increasing the money supply compared to the base scenario, leads to improved economic growth, reduced inflation and interest rates, enhanced bank balance sheets, and increased household welfare (via higher real consumption) compared to the baseline scenario.
However, the method is recommended only to address the current problem in the present situation. The research results showed that the proposed policies guide the economy onto a better path than its current trajectory, but they are not a prescription for the government’s indiscriminate use of the monetary base. To improve conditions in the long term, the government needs a program to control its budget deficit and stop borrowing from banks and the Central Bank. According to the findings, borrowing from the Central Bank to settle outstanding debts with contractors is preferable to leaving these debts unpaid. However, the optimal approach is for the government to avoid needing to borrow from the Central Bank altogether.

Abstract Extensive applications of asset pricing in the fields of finance and economics lead to an increasing importance of this issue, which has attracted more attentions of researchers in theoretical and empirical aspects. Due to this issue, the main purpose of this paper is to compare two asset pricing methods i.e. “Beta” and “stochastic discount factor” in Iran Stock Exchange market. Using the monthly data of Tehran Stock Exchange index return and return of shares of the companies listed in the stock exchange market of Iran during 1379(1) to 1398(6), we have formed 5*5 baskets-called 25 portfolios of Fama and French- to evaluate the efficiency and stability of one factor model (capital asset pricing model) and multi-factors model (Fama and French’s 3 factors model) using Generalized Method of Moments (GMM) estimation method. The results show that the aforementioned methods are not completely superior to each other. In fact, for CAPM model, stochastic discount factor method is more efficient and less stable than Beta method and vice versa for Fama and French’s 3 factors model.

Abstract In this article, we show that the application of LQ methods for estimation of RIOT in Iran requires two types of residuals. To tackle with this problem, a new mixed FLQ-RAS method is proposed. This method maintains the official data of regional accounts that has been provided by the Statistical Centre of Iran and therefore, the sectoral export is used as a residual. Domestic National, Gilan IOTs and regional accounts for the year 2002 have been used. FLQ and FLQ-RAS methods have used to estimate the RIOTs of Gilan. For the evaluation, we have used five conventional statistical methods for error measurment. The results are twofolds.The minimum adjustment is 0.9% for agriculture and the maximum adjustment is 55% for mining. Second the degree of accuracy between the two methods reveals that the proposed method outperforms than the FLQ method. The application of the proposed method has four advantages in Iran: one- flexibility for covering more sectors, two- extension to other regions, three- its complete consistency with the basic data of the country and four- its flexibility in considering exogenous or superior data at the regional level.
 

Abstract Since the introduction of supply-use input-output model by the United Nations in its 1968 SNA, there has been a controversy on choosing the most appropriate technology assumption for estimating symmetric product by product input-output table. These arguments have focused on two technology assumptions; the product technology assumption (PTA) and the activity technology assumption (ATA). The PTA states that each product has a unique input structure that is independent of producing industry. In contrast, the ATA is defined as each activity has its own specific way of production, irrespective of its product mix. Each assumption has its own advantages and disadvantages. Because of ATA’s inconsistency with some fundamental economic theories, “Product Technology” assumption is more widely applied for calculating symmetric product-by-product input-output table. In this paper, we show that only PTA fulfills the four desirable properties (material balance, financial balance, scale invariance and price invariance) which are introduced by Jansen and ten Raa (1990) but ATA fulfils only one of them. This result can be used by compliers and users in choosing the appropriate economic assumptions for deriving symmetric input-output tables.