Factor Model Approach Research Articles

A two-step dynamic factor modelling approach for forecasting inflation in small open economies

We build a dynamic factor model to forecast inflation in a small open economy. The model is estimated with both market and survey data, and a unique two-step methodology to incorporate exogenous factors. Estimations with market data provide a better fit for in-sample and out-of-sample values of inflation. More importantly, our model outperforms univariate and estimated DSGE models, the more common approaches to inflation forecasting that perform well for advanced economies. Our findings, therefore, suggest that a dynamic factor modelling approach for a small open economy such as Jamaica can be a good alternative to the preferred methods for forecasting inflation in advanced economies.

CP factor model for dynamic tensors

Abstract Observations in various applications are frequently represented as a time series of multidimensional arrays, called tensor time series, preserving the inherent multidimensional structure. In this paper, we present a factor model approach, in a form similar to tensor CANDECOMP/PARAFAC (CP) decomposition, to the analysis of high-dimensional dynamic tensor time series. As the loading vectors are uniquely defined but not necessarily orthogonal, it is significantly different from the existing tensor factor models based on Tucker-type tensor decomposition. The model structure allows for a set of uncorrelated one-dimensional latent dynamic factor processes, making it much more convenient to study the underlying dynamics of the time series. A new high-order projection estimator is proposed for such a factor model, utilizing the special structure and the idea of the higher order orthogonal iteration procedures commonly used in Tucker-type tensor factor model and general tensor CP decomposition procedures. Theoretical investigation provides statistical error bounds for the proposed methods, which shows the significant advantage of utilizing the special model structure. Simulation study is conducted to further demonstrate the finite sample properties of the estimators. Real data application is used to illustrate the model and its interpretations.

Investigating the dependency of in vitro benchmark concentrations on exposure time in transcriptomics experiments

There is increasing interest to employ in vitro transcriptomics experiments in toxicological testing, for example to determine a point-of-departure (PoD) for chemical safety assessment. However current practices to derive PoD tend to utilise a single exposure time despite the importance of exposure time on the manifestation of toxicity caused by a chemical. Therefore it is important to investigate both concentration and exposure time to determine how these factors affect biological responses, and as a consequence, the derivation of PoDs. In this study, metabolically competent HepaRG cells were exposed to five known toxicants over a range of concentrations and time points for subsequent gene expression analysis, using a targeted RNA expression assay (TempO-Seq). A non-parametric factor-modelling approach was used to model the collective response of all significant genes, which exploited the interdependence of differentially expressed gene responses. This in turn allowed the determination of an isobenchmark response (isoBMR) curve for each chemical in a reproducible manner. For 2 of the 5 chemicals tested, the PoD was observed to vary by 0.5–1 log-order within the 48-h timeframe of the experiment. The approach and findings presented here clearly demonstrate the need to take both concentration and exposure time into account when designing in vitro toxicogenomics experiments to determine PoD. Doing so also provides a means to use concentration-time-response modelling as a basis to extrapolate a PoD from shorter to longer exposure durations, and to identify chemicals of concern that can cause cumulative effects over time.

A National Study Exploring Factors Promoting Adolescent College Readiness in Math and Science (STEM-CR)

Numerous national and state endeavors have advocated for approaches, funding, and programs focused on expanding the science, technology, engineering, and mathematics (STEM) workforce of the nation and investing in education to cultivate a more diverse and impactful cohort of students who pursue STEM pathways. Educators, support personnel, and policymakers are in a position to engage in discussions about expanding STEM college readiness (STEM-CR) and participation. However, few are cognizant that STEM-CR is a progression that students strive for in developing skills, behaviors, and attitudes that spans over time. The current study established and validated a measurement model of student STEM-CR in mathematics and science utilizing the High School Longitudinal Study of 2009. The model was created based on a multidimensional and theoretical perspective of college readiness using a confirmatory factor analysis and modeling approach that accounted for measurement invariance. The sample ( N = 16,044) comes from a racialized/ethnoracial and socioeconomically diverse high school population in the United States. The findings confirmed that STEM-CR involves four related yet distinct dimensions of Think, Know, Act, and Go. Results also demonstrated soundness of these STEM-CR dimensions by race and gender (key learning skills and techniques/Act). Academic self-efficacy was the strongest dimension of our STEM-CR model and strongly predicted academic achievement and college enrollment. Research and practice implications are discussed.

Social and dimensional comparison effects in general and domain-specific test anxiety: a nested factor modeling approach

The generalized internal/external frame of reference (GI/E) model assumes social and dimensional achievement comparisons to form self-perceptions. These domain-specific comparisons have been shown to shape two facets of test anxiety (i.e., worry and emotionality) both directly and indirectly through academic self-concepts. However, examinations of such domain-specific relations have rarely integrated general components, although the hierarchical nature of both test anxiety and academic self-concept is well-known. Thus, the present study implemented a nested factor modeling approach. We examined social and dimensional comparison effects on worry and emotionality as well as mediation effects of academic self-concepts in the math and verbal domains while controlling for general components. We contrasted this approach with the conventionally used first-order factor model where general components were not considered. Data from N = 348 German secondary school students (Mage = 15.3 years, Grades 9–10) were analyzed using structural equation models. Direct negative within-domain and positive cross-domain achievement-anxiety relations emerged, yet, the pattern of cross-domain relations changed across modeling approaches. Only the nested factor model showed indirect cross-domain mediation relations. Our findings suggest the importance of structural representations of hierarchical constructs. The nested factor model approach enhanced predictions within the GI/E model, particularly those related to dimensional comparisons.

Statistical Inference for the Factor Model Approach to Estimate Causal Effects in Quasi-Experimental Settings

Causal inference using quasi-experimental data is of great interest to marketers. The factor model approach to estimate treatment effects accommodates a large number of control units and can easily handle a large number of treatment units while flexibly allowing for cases where the treatment is outside the range of the control units. However, the factor model method lacks formal inference theory, instead relying on bootstrap or permutation procedures with strong assumptions. Specifically, the extant Xu (2017) bootstrap procedure requires that the treatment and control error variances are equal. In this research the authors establish that when this assumption is violated, the bootstrap procedure results in biased coverage intervals. The authors develop a formal inference theory for the factor model approach to estimate the average treatment effects on the treated. The approach enables formal quantification of uncertainty through hypothesis testing and confidence intervals. The inference method is applicable to both stationary and nonstationary data. More importantly, the inference theory accommodates treatment and control unit outcomes with different distributions, which includes different error variances as a special case. The authors show the performance of the inference theory with simulated data. Finally, they apply the method to empirically quantify the uncertainty in the effect of legalizing recreational marijuana on the beer market and the sales effect of a digitally native online brand opening a physical showroom.

On the efficient synthesis of short financial time series: A Dynamic Factor Model approach

In this paper, we present a fully data-driven statistical approach to building a synthetic index based on intrinsic information of the considered ecosystem, namely the financial one. Among the several methods made available in the literature, we propose the employment of a Dynamic Factor Model approach which allows us to compare observations at hand in space and time. We contribute to the research field by offering a statistically sound methodology which goes beyond state-of-the-art techniques on dimension reduction, mainly based on Principal Component Analysis. We adopt a country-by-country fitting strategy to elicit the inner country-specific characteristics and then we combine results together by means of a Vector Autoregressive and Kalman filter approach. To this aim, we analyze a set of 17 Financial Soundness Indicators provided by the International Monetary Fund ranging from 2010 to 2020 for 116 countries that span the globe, including both strong and developing economies. Results show that our index is able to identify banking and debt crisis and the contribution of the latent variables can isolate countries that experienced crisis, representing a valid aid to policy makers and institutions in understanding countries movements, reactions and suffering periods.

Factor models for high‐dimensional functional time series II: Estimation and forecasting

This article is the second one in a set of two laying the theoretical foundations for a high‐dimensional functional factor model approach in the analysis of large cross‐sections (panels) of functional time series (FTS). Part I establishes a representation result by which, under mild assumptions on the covariance operator of the cross‐section, any FTS admits a canonical representation as the sum of a common and an idiosyncratic component; common components are driven by a finite and typically small number of scalar factors loaded via functional loadings, while idiosyncratic components are only mildly cross‐correlated. Building on that representation result, Part II is dealing with the identification of the number of factors, their estimation, the estimation of their loadings and the common components, and the resulting forecasts. We provide a family of information criteria for identifying the number of factors, and prove their consistency. We provide average error bounds for the estimators of the factors, loadings, and common components; our results encompass the scalar case, for which they reproduce and extend, under weaker conditions, well‐established similar results. Under slightly stronger assumptions, we also provide uniform bounds for the estimators of factors, loadings, and common components, thus extending existing scalar results. Our consistency results in the asymptotic regime where the number of series and the number of time points diverge thus extend to the functional context the ‘blessing of dimensionality’ that explains the success of factor models in the analysis of high‐dimensional (scalar) time series. We provide numerical illustrations that corroborate the convergence rates predicted by the theory, and provide a finer understanding of the interplay between and for estimation purposes. We conclude with an application to forecasting mortality curves, where our approach outperforms existing methods.

Factor models for high‐dimensional functional time series I: Representation results

In this article, which consists of two parts (Part I: representation results; Part II: estimation and forecasting methods), we set up the theoretical foundations for a high‐dimensional functional factor model approach in the analysis of large cross‐sections (panels) of functional time series (FTS). In Part I, we establish a representation result stating that, under mild assumptions on the covariance operator of the cross‐section, we can represent each FTS as the sum of a common component driven by scalar factors loaded via functional loadings, and a mildly cross‐correlated idiosyncratic component. Our model and theory are developed in a general Hilbert space setting that allows for mixed panels of functional and scalar time series. We then turn, in Part II, to the identification of the number of factors, and the estimation of the factors, their loadings, and the common components. We provide a family of information criteria for identifying the number of factors, and prove their consistency. We provide average error bounds for the estimators of the factors, loadings, and common components; our results encompass the scalar case, for which they reproduce and extend, under weaker conditions, well‐established similar results. Under slightly stronger assumptions, we also provide uniform bounds for the estimators of factors, loadings, and common components, thus extending existing scalar results. Our consistency results in the asymptotic regime where the number of series and the number of time observations diverge thus extend to the functional context the ‘blessing of dimensionality’ that explains the success of factor models in the analysis of high‐dimensional (scalar) time series. We provide numerical illustrations that corroborate the convergence rates predicted by the theory, and provide a finer understanding of the interplay between and for estimation purposes. We conclude with an application to forecasting mortality curves, where we demonstrate that our approach outperforms existing methods.

The Performance of Shrinkage Estimator for Stock Portfolio Selection in Case of High Dimensionality

Harry Markowitz introduced the Modern Portfolio Theory (MPT) for the first time in 1952 which has been applied widely for optimal portfolio selection until now. However, the theory still has some limitations that come from the instability of covariance matrix input. This leads the selected portfolio from MPT model to change the status continuously and to suffer the high cost of transaction. The traditional estimator of the covariance matrix has not solved this limitation yet, especially when the dimensionality of the portfolio soars. Therefore, in this paper, we conduct a practical discussion on the feasible application of the shrinkage estimator of the covariance matrix, which is expected to encourage the investors focusing on the shrinkage–based framework for their portfolio selection. The empirical study on the Vietnam stock market in the period of 2011–2021 shows that the shrinkage approach has much better performance than other traditional methods on the primary portfolio evaluation criteria such as return, level of risk, Sharpe ratio, maximum loss, and Alpla coefficient, especially the superiority is even more evident when the dimension of covariance matrix increases. The shrinkage approach tends to create more stable and secure portfolios than other estimators, as demonstrated by the average volatility and maximum loss criteria with the lowest values. Meanwhile, the factor model approach is able to generate portfolios with higher average returns and lower portfolio turnover; and the traditional approach gives good results in the case of low—dimensionality. Besides, the shrinkage method also shows effectiveness when beating the tough market benchmarks such as VN-Index and 1/N portfolio strategy on almost performance metrics in all scenarios.

Measuring dynamic pandemic-related policy effects: A time-varying parameter multi-level dynamic factor model approach

This study evaluates the dynamic impact of various policies adopted by U.S. states, including social distancing, financial assistance, and vaccination policies. We propose a time-varying parameter multilevel dynamic factor model (TVP-MDFM) to improve the model’s accuracy for evaluating the dynamic policy effect. The estimation is based on the Bayesian shrinkage method jointly with the Markov chain Monte Carlo (MCMC) algorithm that combines model selection and parameter estimation into the same iterative sampling process. The advantages and reliability of the TVP-MDFM are explored using simulation studies and robustness tests. The main empirical results highlight that the direct causal effect of the social distancing policy is more significant than the indirect effect mediated through human behavior. We also find income heterogeneity in financial assistance policies. Moreover, we provide evidence that banning vaccination certification by legislation is a stronger driver of the new case rate than executive orders during the Omicron dominance.

FINANCIAL LIBERALIZATION, DEMOCRATIC INSTITUTIONS, AND INCOME INEQUALITY: NEW INSIGHTS FROM PAKISTAN

This study investigates the relationship between financial liberalization, democratic institutions, and income inequality in Pakistan by using a two-state Markov switching (MS) methodology for the period 1972–2017. The dynamic factor modelling approach is used to construct the financial liberalization index. The impact of the Global Financial Crisis (GFC) of 2007–2008 on financial liberalization–institutions–inequality nexus was checked using the Chow breakpoint test. The result reveals no significant effect of GFC on Pakistan’s economy. Results from the MS model divulge that financial liberalization lessens income inequality in the high-income inequality regime as well as in the low-income inequality regime. The result further discloses that democratic institutions enhance income inequality in high-income inequality regimes while reduce income inequality in the low-inequality regime. While studying the joint role of democratic institutions and financial liberalization, it is observed that the weak role of democratic institutions offsets the benefits of financial liberalization in the high-income inequality regime; however, institutions and financial liberalization are supportive in reducing income inequality in the low-inequality regime. Besides, the results also confirm the existence of Kuznets hypothesis in Pakistan. Therefore, there is a need for further reforms to revamp institutions to reduce income inequality in Pakistan.

Switching latent factor value-at-risk models for conditionally heteroskedastic portfolios: A comparative approach

In this paper, a computationally efficient Monte Carlo-based latent factor modeling approach for portfolio Value-at-Risk (VaR) estimation is introduced. We examine whether the inclusion of conditional heteroskedasticity in financial returns, and taking into account for possible hidden (Markovian) “regime changes” in the latent correlation structure of the portfolio can enhance the accuracy of VaR forecasts. Practical details of training such models with the expectation-maximization algorithm are also discussed. In conjunction with an approximated version of the Kalman filter, we show how to calculate maximum likelihood estimates of the model parameters, and to yield inferences about the unobservable path of the common factors, their volatilities and the hidden state sequence of the Markov process. The methodology is illustrated by an example using data from the Tunisian foreign exchange market, over the period of the Tunisian revolution from January 02, 2010 to December 30, 2012. We found that this new specification exhibits a good fit to the data, improves the accuracy of VaR predictions of the Tunisian foreign public debt portfolio and reduces the number and average size of back-testing breaches when a financial crisis occurs.

A flexible moderated factor analysis approach to test for measurement invariance across a continuous variable.

Effort has been devoted to the development of moderated factor models in which the traditional factor model parameters are allowed to differ across a moderator variable. These models are valuable as they enable tests on measurement invariance across a continuous background variable. However, moderated factor models require the specification of a parametric functional form between the factor model parameters and the moderator variable while, in some situations, it is unclear what functional form to assume. Therefore, in the present article, a semiparametric moderated factor modeling approach is presented in which no assumption concerning the functional form between the moderator and the model parameters is imposed. In a simulation study, the semiparametric moderated factor model is shown to be viable in terms of parameter recovery and the power to distinguish the different models for measurement invariance. In addition, the model is applied to a real dataset pertaining to intelligence. (PsycInfo Database Record (c) 2022 APA, all rights reserved).

Co-movement of Banking Stress Across Emerging and Developed Economies: A Dynamic Factor Model and Wavelet Coherence Approach

This study develops a robust banking stability indicator for emerging and advanced economies and examines the linkage of banking stress across economies. We contribute by including interbank borrowings and banking sector volatility to measure contagious risk besides the traditional variables. Second, we use aggregate banking sector data for five countries (China, India, Japan, the UK and the USA) from 1998 to 2015 and employ dynamic factor model to develop the index. Results show higher stress levels in the UK and China, and all economies witness increased stress during the 2007–2009 crises, but the recovery phase varies. Finally, we use wavelet coherence analysis and find evidence of stress transmission from emerging economies to other emerging and advanced economies.

Factor-Based Framework for Multivariate and Multi-step-ahead Forecasting of Large Scale Time Series.

State-of-the-art multivariate forecasting methods are restricted to low dimensional tasks, linear dependencies and short horizons. The technological advances (notably the Big data revolution) are instead shifting the focus to problems characterized by a large number of variables, non-linear dependencies and long forecasting horizons. In the last few years, the majority of the best performing techniques for multivariate forecasting have been based on deep-learning models. However, such models are characterized by high requirements in terms of data availability and computational resources and suffer from a lack of interpretability. To cope with the limitations of these methods, we propose an extension to the DFML framework, a hybrid forecasting technique inspired by the Dynamic Factor Model (DFM) approach, a successful forecasting methodology in econometrics. This extension improves the capabilities of the DFM approach, by implementing and assessing both linear and non-linear factor estimation techniques as well as model-driven and data-driven factor forecasting techniques. We assess several method integrations within the DFML, and we show that the proposed technique provides competitive results both in terms of forecasting accuracy and computational efficiency on multiple very large-scale (>102 variables and > 103 samples) real forecasting tasks.

Assessment of the oral cancer control activities through care seeking behavior of hospital attendees and their level of awareness

Oral cancer is the most common cancer among males in the Sri Lankan population. The aim of this study was to assess the health seeking behaviors and associated factors for OPMD.A hospital based descriptive cross-sectional study was conducted among Out Patients Department attendees of the Institute of Oral Health, Maharagama, from July 1, 2018 to December 31, 2019 by administering a questionnaire followed by an intra oral examination.The total number of 110 OPMD/oral cancer patients were identified. Awareness on OPMD was that 40 ​%. Majority were unaware of the danger of areca nut and alcohol use. Out of the identified cases during oral examination, 47.3 ​% were incidental findings while only 7.3 ​% were referred by the General Medical Practitioners (GMP). None were referred by the Primary Health Care (PHC) staff.The level of awareness on OPMD and its risk factors were very low. Incidental findings during the dental screening being the main mode of detection of OPMD, highlights the importance of conducting opportunistic screening at dental settings. Implementation of the Risk Factor Model (RFM) approach for the early detection of oral cancer through PHC staff and GMP is validated to encourage more early referrals.

Examining Cross-border Comovements of REITs Around the World

We examine the amount of the comovement among real estate investment trusts (REITs) around the world by estimating implied (risk-adjusted) comovements using a factor model approach. Our investigation uses a sample of individual REITs from 21 developed economies from 1990 to 2015 to examine whether cross-border comovements are observable. Our findings provide evidence of comovements after controlling for common risk factors and relevant firm characteristics. Furthermore, we show that the comovements increase over time and are more common for larger and more liquid REITs. The results indicate that country-level factors such as the local market conditions, financial development, domestic exchange rates, credit constraints, and governance significantly affect the amount of comovements.

Factor Models for High-Dimensional Tensor Time Series

Large tensor (multi-dimensional array) data routinely appear nowadays in a wide range of applications, due to modern data collection capabilities. Often such observations are taken over time, forming tensor time series. In this article we present a factor model approach to the analysis of high-dimensional dynamic tensor time series and multi-category dynamic transport networks. This article presents two estimation procedures along with their theoretical properties and simulation results. We present two applications to illustrate the model and its interpretations.

Forecasting value-at-risk and expected shortfall in large portfolios: A general dynamic factor model approach

Beyond their importance from the regulatory policy point of view, Value-at-Risk (VaR) and Expected Shortfall (ES) play an important role in risk management, portfolio allocation, capital level requirements, trading systems, and hedging strategies. However, due to the curse of dimensionality, their accurate estimation and forecast in large portfolios is quite a challenge. To tackle this problem, two procedures are proposed. The first one is based on a filtered historical simulation method in which high-dimensional conditional covariance matrices are estimated via a general dynamic factor model with infinite-dimensional factor space and conditionally heteroscedastic factors; the other one is based on a residual-based bootstrap scheme. The two procedures are applied to a panel with concentration ratio close to one. Backtesting and scoring results indicate that both VaR and ES are accurately estimated under both methods, which both outperform the existing alternatives.