Search

Traditionally Bayesian decision-theoretic design of experiments proceeds by choosing a design to minimise expectation of a given loss function over the space of all designs. The loss function encapsulates the aim of the experiment, and the expectation is taken with respect to the joint distribution of all unknown quantities implied by the statistical model that will be fitted to observed responses. In this paper, an extended framework is proposed whereby the expectation of the loss is taken with respect to a joint distribution implied by an alternative statistical model. Motivation for this includes promoting robustness, ensuring computational feasibility and for allowing realistic prior specification when deriving a design. To aid in exploring the new framework, an asymptotic approximation to the expected loss under an alternative model is derived, and the properties of different loss functions are established. The framework is then demonstrated on a linear regression versus full-treatment model scenario, on estimating parameters of a non-linear model under model discrepancy and a cubic spline model under an unknown number of basis functions., Supplementary material appears as an appendix

SYNOPSIS After the financial crisis of the late 2000s, concern about delayed credit-loss recognition under the incurred-loss method prompted the FASB and the IASB to develop expected-loss methods. We review the development of these methods, including through comment-letter analysis. Initially, the FASB recommended immediate full recognition of expected losses, including at day one, and the IASB recommended spreading the recognition of initially expected losses across time. After unsuccessful attempts to converge based on proposals that partly reflected initial recommendations of each board, the boards eventually adopted different methods. We report that U.S. respondents largely opposed the FASB's final method, which required day-one recognition of all expected losses, and that non-U.S. respondents largely supported the IASB's final method, which required day-one recognition of 12-month expected losses. Day-one loss was controversial and impeded convergence. Our comment-letter analysis suggests that a day-one-loss-free more forward-looking incurred-loss method might provide a route to a more converged solution. JEL Classifications: G21; M41.

This study proposed a method for constructing rating tools using logistic regression and linear discriminant analysis to determine the risk profile of SME portfolios. The objective, firstly, is to evaluate the impact of the crisis due to the Covid-19 by readjusting the profile of each company by using the expert opinion and, secondly, to evaluate the efficiency of the measures taken by the Moroccan state to support the companies during the period of the pandemic. The analysis in this paper showed that the performance of the logistic regression and linear discriminant analysis models is almost equivalent based on the ROC curve. However, it was revealed that the logistic regression model minimizes the risk cost represented in this study by the expected loss. For the support measures adopted by the Moroccan government, the study showed that the failure rate (critical situation) of the firms benefiting from the support is largely lower than that of the non-beneficiaries. © 2021 LLC CPC Business Perspectives. All rights reserved.

Gambling is legal in most countries. However, despite having some economic benefits, certain characteristics of gambling can have health consequences, rendering it a public health issue. The effects can be summarized according to the following three “laws” of ethical gambling: “Gambling Dynamics Law”: companies’ economic gains come directly from players’ losses; “Expected Loss Law”: the more one gambles, the greater the probability of losing; and “Addiction Law”: the more one gambles, the greater the need to play again, leading to further losses. Ludens is a gambling addiction prevention program that has four goals: inform participants about gambling and gambling addiction; sensitize participants to the risk of gambling for health, especially addiction; promote a change in attitudes toward gambling; and alert participants to risky behaviors that can lead to addiction. The prevention program was implemented during 2017 to 2019. Fourteen psychologists presented it to 2372 adolescents (48.8% females, 51.2% males) aged 14–19 years, none of whom were university students, recruited from 42 Spanish high schools in 132 groups taking different courses. The main dependent variables analyzed were the monthly frequencies of gambling, at-risk gambling, and gambling addiction (as measured by the National Opinion Research Center DSM-IV Screen for Gambling Problems, adapted to diagnose gambling disorder according to DSM-5, in which pathological gambling is considered an addictive disorder). Given that all of the gamblers were adolescents (most were minors), fulfilment of 1–3 the DSM-5 diagnostic criteria was considered to indicate a risk of problem gambling. After the administration of Ludens, statistically significant reductions were observed in the three variables of interest: monthly frequency of gambling, percentage of adolescents with risky gambling, and percentage of adolescents with gambling disorder. The results were analyzed according to sex and age (minors vs. adolescents between 18 and 19 years old). The results obtained after applying the prevention program indicate that Ludens is effective as a universal prevention program for gambling addiction.

PurposeThe purpose of this article is to derive formulas for lifetime expected credit loss of loans that are required for the calculation of loan loss reserves under IFRS 9. This is done both for fixed-rate and floating rate loans under different assumptions on LGD modeling, prepayment, and discount rates.Design/methodology/approachThis study provides exact formulas for lifetime expected credit loss derived analytically together with the mathematical proofs of each expression.FindingsThis articles shows that the formula most commonly applied in the literature for calculating lifetime expected credit loss is inconsistent with measuring expected loss based on expected discounted cash flows. Formulas based on discounted cash flows always lead to more conservative numbers.Practical implicationsFor banks reporting under IFRS 9, the implication of this research is a better understanding of the different approaches used for computing lifetime expected loss, how they are connected, and what assumptions are underlying each approach. This may lead to corrections in existing frameworks to make applications of risk management systems more consistent.Originality/valueWhile there is a lot of literature explaining IFRS 9 and evaluating its impact, none of the existing research has systematically analyzed the calculation of lifetime expected credit loss for this purpose and how the formula changes under different modeling assumptions. This gap is filled by this study.

PurposeCovid-19 pandemic is a unique and extraordinary situation for the globe, which has potentially disrupted almost all aspects of life. In this global crisis, the tourism and hospitality sector has collapsed in almost all parts of the world, and the same is true for India. Therefore, this paper aims to investigate the impact of Covid-19 on the Indian tourism industry.Design/methodology/approachThis study develops an appropriate model to forecast the expected loss of foreign tourist arrivals (FTAs) in India for 10 months. Since the FTAs follow a seasonal trend, seasonal autoregressive integrated moving average (SARIMA) method has been employed to forecast the expected FTAs in India from March 2020 to December 2020. The results of the proposed model are then compared with the ones obtained by Holt-Winter's (H-W) model to check the robustness of the proposed model.FindingsThe SARIMA model seeks to manifest the monthly arrival of foreign tourists and also elaborates on the progressing expected loss of foreign tourists arrive for the next three quarters is approximately 2 million, 2.3 million and 3.2 million, respectively. Thus, in the next three quarters, there will be an enormous downfall of FTAs, and there is a need to adopt appropriate measures. The comparison demonstrates that SARIMA is a better model than H-W model.Originality/valueSeveral studies have been reported on pandemic-affected tourism sectors using different techniques. The earlier pandemic outbreak was controlled and region-specific, but the Covid-19 eruption is a global threat having potential ramifications and strong spreading power. This work is one of the first attempts to study and analyse the impact of Covid-19 on FTAs in India.

Loss functions, commonly believed to be the cost of cognitive computing, are a key element in decision-making, and three-way decisions can be regarded as a cognitive computing method that seeks to minimize the overall risks involved in the decision-making process. Recently, many studies on loss functions have been conducted based on fuzzy sets, intuitionistic fuzzy sets, and interval intuitionistic fuzzy sets. However, most of these studies draw conclusions based on two descriptions, which may fail to capture the whole picture of decision-making. In this paper, in order to improve the accuracy of decision-making, we propose loss functions based on three descriptions, adding a hesitation description to the Pythagorean fuzzy environment. Then, we redefine the expected loss functions, which allow people to make a decision with more uncertainty. Subsequently, on the basis of the Bayesian minimum risk decision theory, four strategies for dealing with expected losses are proposed, and three-way decision models are established. Finally, group decision models are discussed. Three-way decision models of real value loss functions and Pythagorean fuzzy loss functions based on three descriptions are proposed, and data analyses of different parameters show the feasibility of the three-way decision models.

In recommender systems, supervised information is usually obtained from the historical data of users. For example, if a user watched a movie, then the user-movie pair will be marked as positive. On the other hand, the user-movie pairs did not appear in the historical data could be either positive or negative. This phenomenon motivates us to formalize the recommender task as a Positive Unlabeled learning problem. As the model trained on the biased historical data may not generalize well on future data, we propose an active learning approach to improve the model by querying further labels from the unlabeled data pool. With the target of querying as few instances as possible, an active selection strategy is proposed to minimize the expected loss and match the distribution between labeled and unlabeled data. Experiments are performed on both classification datasets and movie recommendation dataset. Results demonstrate that the proposed approach can significantly reduce the labeling cost while achieving superior performance regarding multiple criteria.

PurposeThis paper provides an objective approach based on available market information capable of reducing subjectivity, inherently present in the process of expected loss provisioning under the IFRS 9.Design/methodology/approachThis paper develops the two-step methodology. Calibrating the Credit Default Swap (CDS)-implied default probabilities to the through-the-cycle default frequencies provides average weights of default component in the spread for each forward term. Then, the impairment provisions are calculated for a sample of investment grade and high yield obligors by distilling their pure default-risk term-structures from the respective term-structures of spreads. This research demonstrates how to estimate credit impairment allowances compliant with IFRS 9 framework.FindingsThis study finds that for both investment grade and high yield exposures, the weights of default component in the credit spreads always remain inferior to 33%. The research's outcomes contrast with several previous results stating that the default risk premium accounts at least for 40% of CDS spreads. The proposed methodology is applied to calculate IFRS 9 compliant provisions for a sample of investment grade and high yield obligors.Research limitations/implicationsMany issuers are not covered by individual Bloomberg valuation curves. However, the way to overcome this limitation is proposed.Practical implicationsThe proposed approach offers a clue for a better alignment of accounting practices, financial regulation and credit risk management, using expected loss metrics across diverse silos inside organizations. It encourages adopting the proposed methodology, illustrating its application to a set of bond exposures.Originality/valueNo previous research addresses impairment provisioning employing Bloomberg valuation curves. The study fills this gap.

PURPOSE: The purpose of this paper is to show that the Value at Risk (VaR) method can be used to estimate the death rate from Covid-19 infection., DESIGN/METHODOLOGY/APPROACH: The VaR method allows for risk measurements and estimations of the highest expected loss on a portfolio at an assumed confidence level over a specified time horizon. The most important assumption affecting the calculation method is that price changes in financial markets follow a normal distribution., FINDINGS: It appears that by appropriately re-defining the concepts of assets and portfolio rates of return, we can describe the volatility in the numbers of deaths caused by Covid-19. We also confirmed using the Shapiro-Wilk and Skewness and Kurtosis tests that the rates of return distribution for the death numbers follow a normal distribution., PRACTICAL IMPLICATIONS: The VaR method allows to estimate the number of deaths based on current trends which can be utilised to better manage available resources in order to reduce casualties. We use the data regarding the number of deaths in the Visegrad Group (V4) countries as a case study to test the effectiveness and accuracy of the VaR method in a different, non-financial domain., ORIGINALITY/VALUE: The theory we used in this paper is currently mainly applied to financial investments. We use this theory to describe social phenomenon which is the number of deaths, our approach has not been seen in the literature so far., peer-reviewed

We propose a new assessed value approach to control for the amount of persistent unobserved quality. We apply our approach to a well-established two-stage framework developed by Genesove and Mayer (GM, 2001), who test the effect of an expected loss on final transaction prices in the housing market. We show that our assessed value model effectively mitigates the omitted variable bias and produces similar results as GM when the first-stage residual is included. Importantly, our model does not rely on repeat sales and therefore provides a powerful new tool for estimating market value. Results are robust to alternative specifications, to controlling for loan-to-value ratios, to replacing final sale price with listing price, to alternative fixed effects, to subperiods, to different holding periods, to simulated quality, to excluding flippers, and to controlling improvements between sales.

The consumer goods market is characterized by strong competition. Thus, to bind customers to the company, a higher priority needs to be attributed to customer retention measures. Such measures include the loyalty card, but the use of the physical card is declining. To counteract this decline, mobile loyalty cards were developed. The basis for the use of mobile loyalty cards is sufficient consumer acceptance. This work is expected to contribute to the explanation of acceptance in the form of usage behavior. Based on the Technology-Acceptance-Model-2 (TAM2) and the literature, hypotheses were derived and a research model was developed. For model testing, a dataset of 255 participants was generated through an online survey and analyzed using partial least squares structural equation modeling (PLS-SEM). The results show that in addition to financial benefits, convenience benefits and psychological factors also have an influence on acceptance. Furthermore, the usage behavior is not negatively influenced by the expected loss of control over personal data. Based on the findings, indications for marketing implementation are given for the confirmed factors.

Phase I studies of a cytotoxic agent often aim to identify the dose that provides an investigator specified target dose-limiting toxicity (DLT) probability. In practice, an initial cohort receives a dose with a putative low DLT probability, and subsequent dosing follows by consecutively deciding whether to retain the current dose, escalate to the adjacent higher dose, or de-escalate to the adjacent lower dose. This article proposes a Phase I design derived using a Bayesian decision-theoretic approach to this sequential decision-making process. The design consecutively chooses the action that minimizes posterior expected loss where the loss reflects the distance on the log-odds scale between the target and the DLT probability of the dose that would be given to the next cohort under the corresponding action. A logistic model is assumed for the log odds of a DLT at the current dose with a weakly informative t-distribution prior centered at the target. The key design parameters are the pre-specified odds ratios for the DLT probabilities at the adjacent higher and lower doses. Dosing rules may be pre-tabulated, as these only depend on the outcomes at the current dose, which greatly facilitates implementation. The recommended default version of the proposed design improves dose selection relative to many established designs across a variety of scenarios.

Accounting for credit losses under U.S. GAAP is transitioning from an incurred to an expected loss model. The model change was motivated by concerns that reporting only incurred losses does not provide investors with sufficient and timely information about banks’ credit risk. In this paper, I develop a measure of lifetime expected credit losses using vintage analysis to examine whether stock prices reflect information about unrecognized expected credit losses in an incurred loss regime. Consistent with investors being able to obtain information about expected losses that are not recognized in the financial statements, I find that unrecognized expected credit losses are negatively associated with bank stock prices. The pricing of these losses is stronger for larger banks, consistent with lower costs of obtaining this information for banks with better information environments. I also find that recorded allowances were less than estimated expected losses, on average, consistent with concerns that implementing the expected loss model will adversely impact regulatory capital adequacy.

PurposeThe purpose of this paper is to develop and operationalize a process for prioritizing supply chain risks that is capable of capturing the value at risk (VaR), the maximum loss expected at a given confidence level for a specified timeframe associated with risks within a network setting.Design/methodology/approachThe proposed “Worst Expected Best” method is theoretically grounded in the framework of Bayesian Belief Networks (BBNs), which is considered an effective technique for modeling interdependency across uncertain variables. An algorithm is developed to operationalize the proposed method, which is demonstrated using a simulation model.FindingsPoint estimate-based methods used for aggregating the network expected loss for a given supply chain risk network are unable to project the realistic risk exposure associated with a supply chain. The proposed method helps in establishing the expected network-wide loss for a given confidence level. The vulnerability and resilience-based risk prioritization schemes for the model considered in this paper have a very weak correlation.Originality/valueThis paper introduces a new “Worst Expected Best” method to the literature on supply chain risk management that helps in assessing the probabilistic network expected VaR for a given supply chain risk network. Further, new risk metrics are proposed to prioritize risks relative to a specific VaR that reflects the decision-maker's risk appetite.

According to the current Intergovernmental Panel of Climate Change report, climate change will increase the probability of occurrence of droughts in some areas. Recent contributions at the international level indicate that trade is expected to act as an efficient tool to mitigate the adverse effect of future climate conditions on agriculture. However, no contribution has focused on the similar capacity of trade within any country yet. The U.S. is an obvious choice given that many climate impact studies focus on its agriculture and around 90% of the U.S. crop trade is domestic. Combining a recent state‐to‐state trade flow dataset with detailed drought records at a fine spatial and temporal resolution, this paper highlights first that trade increases as the destination state experiences more drought and inversely in the origin state. As a result, crop growers' profits depend on both local and trade partners' weather conditions. Projections based on future weather data convert the crop grower's expected loss without trade into expected profit. As such, this paper challenges the estimates of the current climate impact literature and concludes that trade is expected to act as a $14.5 billion mitigation tool in the near future.

In game theory and statistical decision theory, a random (i.e., mixed) decision strategy often outperforms a deterministic strategy in minimax expected loss. As experimental design can be viewed as a game pitting the Statistician against Nature, the use of a random strategy to choose a design will often be beneficial. However, the topic of minimax-efficient random strategies for design selection is mostly unexplored, with consideration limited to Fisherian randomization of the allocation of a predetermined set of treatments to experimental units. Here, for the first time, novel and more flexible random design strategies are shown to have better properties than their deterministic counterparts in linear model estimation and prediction, including stronger bounds on both the expectation and survivor function of the loss distribution. Design strategies are considered for three important statistical problems: (i) parameter estimation in linear potential outcomes models, (ii) point prediction from a correct linear model, and (iii) global prediction from a linear model taking into account an L 2-class of possible model discrepancy functions. The new random design strategies proposed for (iii) give a finite bound on the expected loss, a dramatic improvement compared to existing deterministic exact designs for which the expected loss is unbounded. Supplementary materials for this article are available online.

Due to climatic hazards and extreme weather events, the pricing of property insurance products is increasingly attracting the attention of policyholders, insurance companies, and governments. Pricing based on market-oriented methods has to consider the affecting factors from policyholders’ perceived value. Pricing strategy design generates the need for natural hazards risk assessments. A natural hazard risk assessment is closely related to the human factors of a disaster-bearing body. In response to this need, we design an extension of the expected utility that is inconsistent with the additive expected utility, considering the human factors of policyholders, which is referred to as the expected loss utility (ELU). The ELU presents two improvements of the currently used utility. First, subjective probability, which is derived from individual predictions over acts, is applied to the ELU function to overcome the disadvantage that objective probability attaches to uncertainty does not reflect the uncertainty of human factors. Policyholders’ risk attitudes are reflected by the interpretation of interactions among uncertain events. Second, the hesitant fuzzy linguistic preference relation (HFLPR) is employed as the assessment of individual loss evaluation to reflect a policyholder’s hesitation. We apply the techniques of fuzzy linguistic term aggregation and perform a comparison to simplify our loss utility function. A detailed process of expected loss assessment is proposed due to variations in natural environment factors, local social characteristics, and disaster-bearing body factors. An illustrative example is given to perform a comparison with cumulative prospect theory to show the merits of the ELU. This study quantifies policyholder’s cognition of uncertain event and the cognition’s influence on risk assessment which can guide pricing strategy of property insurance products.

We develop a conceptual framework to model the seasonality in the probability of catastrophe bonds being triggered. This seasonality causes strong seasonal fluctuations in spreads. For example, the spread on a hurricane bond is highest at the start of the hurricane season and declines as time goes by without a hurricane. The spread is lowest at the end of the hurricane season assuming the bond was not triggered, and then gradually increases as the next hurricane season approaches. The model also implies that the magnitude of the seasonality effect increases with the expected loss and the approaching maturity of the bond. The model is supported by an empirical analysis that indicates that up to 47% of market fluctuations in the yield spreads on single-peril hurricane bonds can be explained by seasonality. In addition, we provide a method to obtain market-implied distributions of arrival frequencies from secondary market spreads.

This paper examines banks’ disclosures and loss recognition in the financial crisis and identifies several core issues for the link between accounting and financial stability. Our analysis suggests that, going into the financial crisis, banks’ disclosures about relevant risk exposures were relatively sparse. Such disclosures came later after major concerns about banks’ exposures had arisen in markets. Similarly, the recognition of loan losses was relatively slow and delayed relative to prevailing market expectations. Among the possible explanations for this evidence, our analysis suggests that banks’ reporting incentives played a key role, which has important implications for bank supervision and the new expected loss model for loan accounting. We also provide evidence that shielding regulatory capital from accounting losses through prudential filters can dampen banks’ incentives for corrective actions. Overall, our analysis reveals several important challenges if accounting and financial reporting are to contribute to financial stability. Keywords: Banks, Financial crisis, Financial stability, Disclosure, Loan loss accounting, Expected credit losses, Incurred loss model, Prudential filter, Fair value accounting

Model-based reinforcement learning is expected to be a method that can safely acquire the optimal policy under real-world conditions by using a stochastic dynamics model for planning. Since the stochastic dynamics model of the real world is generally unknown, a method for learning from state transition data is necessary. However, model learning suffers from the problem of bias-variance trade-off. Conventional model learning can be formulated as a minimization problem of expected loss. Failure to consider higher-order statistics for loss would lead to fatal errors in long-term model prediction. Although various methods have been proposed to explicitly handle bias and variance, this paper first formulates a new loss function, especially for sequential training of the deep neural networks. To explicitly consider the bias-variance trade-off, a new multi-objective optimization problem with the augmented weighted Tchebycheff scalarization, is proposed. In this problem, the bias-variance trade-off can be balanced by adjusting a weight hyperparameter, although its optimal value is task-dependent and unknown. We additionally propose a general-purpose and efficient meta-optimization method for hyperparameter(s). According to the validation result on each epoch, the proposed meta-optimization can adjust the hyperparameter(s) towards the preferred solution simultaneously with model learning. In our case, the proposed meta-optimization enables the bias-variance trade-off to be balanced for maximizing the long-term prediction ability. Actually, the proposed method was applied to two simulation environments with uncertainty, and the numerical results showed that the well-balanced bias and variance of the stochastic model suitable for the long-term prediction can be achieved.

This paper provides an analysis of how the benefits of information segmentation can assist an organization to derive the appropriate amount to invest in cybersecurity from a cost-benefit perspective. An analytical model based on the framework of the Gordon-Loeb Model ([1]) is presented that provides a set of sufficient conditions for information segmentation to lower the total investments in cybersecurity and the expected loss from cybersecurity breaches. A numerical example illustrating the insights gained from the model is also presented.

With the introduction of performance-based earthquake engineering (PBEE), engineers have strived to relate building performance to different seismic hazard levels. Expected annual loss (EAL) and collapse safety described by mean annual frequency of collapse (MAFC) have become employed more frequently, but tend to be limited to seismic assessment rather than design. This article outlines an integrated performance-based seismic design (IPBSD) method that uses EAL and MAFC as design parameters. With these, as opposed to conventional intensity-based strength and/or drift requirements, IPBSD limits expected monetary losses and maintains a sufficient and quantifiable level of collapse safety in buildings. Through simple procedures, it directly identifies feasible structural solutions without the need for detailed calculations and numerical analysis. This article outlines its implementation alongside other contemporary risk-targeted and code-based approaches. Several case study reinforced concrete frame structures are evaluated using these approaches and the results appraised via verification analysis. The agreement and consistency of the design solutions and the intended targets are evaluated to demonstrate the suitability of each method. The proposed framework is viewed as a stepping stone for seismic design with advanced performance objectives in line with modern PBEE requirements.

This article is devoted to the methodology and analysis of the results of economic assessment and forecasting of the consequences of global climate change in the form of permafrost thawing and degradation for the housing sector in eight regions of the Russian Arctic. Changes in the state of permafrost soils during the implementation of the most negative (scenario RCP 8.5) of the IPCC forecast options as the most appropriate to the conditions of the Russian Arctic were taken as a physio-geographic basis for the assessment. It is shown that, under a conservative scenario of the housing sector development in this macroregion of Russia in 2020–2050, the annual average cost of maintenance and restoration of the lost housing stock will exceed ₽30 bln. With the implementation of the modernization scenario, the cost above will increase to ₽36 bln. The maximum expected loss is predicted in the Yamalo-Nenets Autonomous Okrug and Krasnoyarsk krai, and the minimum, in the Chukotka and Khanty–Mansi autonomous okrugs.