Dynamic Discrete Choice Model Research Articles

Estimating a Dynamic Discrete Choice Model with Partial Observability for Household Mortgage Default and Prepayment Behaviors

Households are forward-looking when deciding whether to default on or refinance their mortgages. There are two types of default generated by two mechanisms: illiquidity-triggered default and strategic default. However, researchers can observe only whether households default but not whether the default is illiquidity-triggered or strategic. Moreover, typically researchers can observe only whether households prepay but not whether the prepayment is due to refinancing or moving. This paper extends the conditional choice probability (CCP) method to estimate a dynamic discrete choice model with partially observable outcomes. Exclusion restrictions and identification at infinity arguments provide the identification conditions. Counterfactual analyses for foreclosure-mitigating loan modification policies show that writing down the principal can reduce both illiquidity-triggered default and strategic default, that interest reduction can reduce illiquidity-triggered default but cannot effectively reduce strategic default, and that term extension can reduce illiquidity-triggered default but will increase strategic default.

Dynamic treatment effects

This paper develops robust models for estimating and interpreting treatment effects arising from both ordered and unordered multi-stage decision problems. Identification is secured through instrumental variables and/or conditional independence (matching) assumptions. We decompose treatment effects into direct effects and continuation values associated with moving to the next stage of a decision problem. Using our framework, we decompose the IV estimator, showing that IV generally does not estimate economically interpretable or policy-relevant parameters in prototypical dynamic discrete choice models, unless policy variables are instruments. Continuation values are an empirically important component of estimated total treatment effects of education. We use our analysis to estimate the components of what LATE estimates in a dynamic discrete choice model.

Counterfactual Policy Predictions Using Dynamic Discrete Choice Models Without Normalization of Per Period Utilities

This paper studies the nonparametric identification and estimation of the structural parameters, including the per period utility functions, discount factors, and state transition laws, of general dynamic programming discrete choice (DPDC) models. I show an equivalence between the identification of the DPDC model and the identification of a linear GMM system. Using such an equivalence, I simplify both the identification analysis and the estimation practice of DPDC model. First, I prove a series of identification results for the DPDC model by using rank conditions. Previous identification results in the literature are based on normalizing the per period utility functions of one alternative. Such normalization could severely bias the estimates of counterfactual policy effects. I show that the structural parameters can be nonparametrically identified without the normalization. Second, I propose a closed form nonparametric estimator for the per period utility functions, of which computation involves only least square. The existing estimation procedures rely on assuming the dynamic programming (DP) problem is stationary or on solving the DP problem numerically with the aid of terminal conditions. Neither my identification nor the estimation requires terminal conditions, the DPDC model to be stationary, or a sample that covers the entire decision horizon.

Multinational versus National Firms on Capital Adjustment Costs: A Structural Approach

Abstract This paper provides an alternative perspective on the firm-level empirical analysis of the relation between foreign ownership and capital demand adjustment in host countries. The author estimates a dynamic structural model of investment on a sample of 4672 Belgian firms for the period 2003-2010, permitting him to distinguish the ‘ownership status’ of firms. He considers a dynamic discrete choice model of a general specification of adjustment costs including convex and non-convex components. The author uses the method of simulated moments in order to estimate the structural parameters. His results indicate that multinationals’ affiliates face lower capital adjustment costs than national firms.

Crime, Employment and Social Welfare: An Individual-Level Study on Disadvantaged Males

We seek evidence for economic and social mechanisms that aim to explain the relationship between employment and crime. We use the distinctive features of social welfare for identification. We consider a sample of disadvantaged males from The Netherlands who are observed between ages 18 and 32 on a monthly time scale. We simultaneously model the offending, employment and social welfare variables using a dynamic discrete choice model, where we allow for state dependence, reciprocal effects and time-varying unobserved heterogeneity. We find significant negative bi-directional structural effects between employment and property crime. Robustness checks show that only regular employment is able to significantly reduce the offending probability. Further, a significant uni-directional effect is found for the public assistance category of social welfare on property offending. The results highlight the importance of economic incentives for explaining the relationship between employment and crime for disadvantaged individuals. For these individuals the crime reducing effects from the public assistance category of social welfare are statistically equivalent to those from employment, which suggests the importance of financial gains. Further, the results suggest that stigmatizing effects from offending severely reduce future employment probabilities.

Suppliers Dynamic Approach to Invest in R&D with Sunk Costs in Indian Contexts

In this paper we test for the presence of sunk costs in suppliers’ R&D activities by analyzing the persistence of these activities using supplier level panel data. We develop and estimate a dynamic discrete choice model where each supplier’s current R&D expenditure is a function, among other factors, of its previous experience in performing R&D activities. The data used is a panel data of Indian suppliers, for the period 2003 - 2013. We find that prior R&D experience significantly affects the current decision to invest in R&D, and that, although important, the effect of prior R&D experience depreciates fairly quickly over time.

Is there a demand for multi-year crop insurance?

Purpose – The purpose of this paper is to examine the aggregate demand for single- and multi-year crop insurance contracts and to discuss market potential for multi-year crop insurances. Design/methodology/approach – In this paper the authors develop a dynamic discrete choice model of insurance alternatives, in which single- and multi-year insurance contracts are offered to heterogeneous risk averse farmers. The farmers determine their insurances choices based on inter-temporal utilities. Findings – The results show that in a competitive insurance market with heterogeneous risk averse farmers, there is simultaneous demand for both insurance contracts. Moreover, the introduction of multi-year contracts enhances the market penetration of insurance products. Research limitations/implications – The effect of introducing multi-year crop insurance is moderate when applying the model to US corn production. In practice, however, the increase of insurance demand could be more pronounced because we did not consider marketing and administrative costs and thus ignore this cost reduction potential of multi-year insurance. Originality/value – This study adds to the literature analyzing the feasibility of multi-year crop insurance and also shows that there is market potential for multi-year crop insurance.

Estimating Dynamic Discrete Choice Models with Hyperbolic Discounting, with an Application to Mammography Decisions

We extend the semi-parametric estimation method for dynamic discrete choice models using Hotz and Miller’s (1993) conditional choice probability (CCP) approach to the setting where individuals may have hyperbolic discounting time preferences and may be naive about their time inconsistency. We illustrate the proposed estimation method with an empirical application of adult women’s decisions to undertake mammography to evaluate the importance of present bias and naivety in the under-utilization of this preventive health care. Our results show evidence for both present bias and naivety.

ESTIMATING DYNAMIC DISCRETE CHOICE MODELS WITH HYPERBOLIC DISCOUNTING, WITH AN APPLICATION TO MAMMOGRAPHY DECISIONS

We extend the semiparametric estimation method for dynamic discrete choice models using Hotz and Miller's (Review of Economic Studies 60 (1993), 497–529) conditional choice probability approach to the setting where individuals may have hyperbolic discounting time preferences and may be naive about their time inconsistency. We illustrate the proposed identification and estimation method with an empirical application of adult women's decisions to undertake mammography to evaluate the importance of present bias and naivety in the underutilization of this preventive health care. Our results show evidence for both present bias and naivety.

ESTIMATION OF DYNAMIC DISCRETE CHOICE MODELS BY MAXIMUM LIKELIHOOD AND THE SIMULATED METHOD OF MOMENTS

We compare the performance of maximum likelihood (ML) and simulated method of moments (SMM) estimation for dynamic discrete choice models. We construct and estimate a simplified dynamic structural model of education that captures some basic features of educational choices in the United States in the 1980s and early 1990s. We use estimates from our model to simulate a synthetic dataset and assess the ability of ML and SMM to recover the model parameters on this sample. We investigate the performance of alternative tuning parameters for SMM.

A Bounded Rationality Model of Information Search and Choice in Preference Measurement

It is becoming increasingly easier for researchers and practitioners to collect eye-tracking data during online preference measurement tasks. The authors develop a dynamic discrete choice model of information search and choice under bounded rationality, which they calibrate using a combination of eye-tracking and choice data. Their model extends Gabaix et al.'s (2006) directed cognition model by capturing fatigue, proximity effects, and imperfect memory encoding and by estimating individual-level parameters and partworths within a likelihood-based hierarchical Bayesian framework. The authors show that modeling eye movements as the outcome of forward-looking utility maximization improves out-of-sample predictions, enables researchers and practitioners to use shorter questionnaires, and allows better discrimination between attributes.

Multinational versus National Firms on Labour Adjustment Costs: A Structural Approach

This paper provides a different perspective on the firm-level empirical analysis of the relation between foreign ownership and labour demand adjustment in host countries. We estimate a dynamic structural model of employment on a sample of 8215 Belgian firms observed between 2003 and 2010 that permits to distinguish the ‘ownership status’ of firms. We consider a dynamic discrete choice model of a general specification of adjustment costs including convex, non-convex and “disruption of production” components. We use a method of simulated moments procedure to estimate the structural parameters. Our results indicate that multinational’s affiliates face lower convex and fixed adjustment costs but higher “disruption of production” adjustment costs than national firms.

Estimation of dynamic discrete models from time aggregated data

An important component in dynamic discrete choice models and dynamic discrete games is the transition density of state variables from the current period to the next period. Most empirical dynamic discrete choice models identify the theoretical time interval in the behavioral model with that observed in the data set. However, many empirical data sets are time aggregated. In this paper, we show that when the time interval in the behavioral theory model differs from that in the observed data, difficulties with nonparametric identification and specification arise. In addition, we study the properties of parametric maximum likelihood estimators and flexible semiparametric estimators of the transition density in dynamic discrete models with time aggregated data sets.

Learning Large-Scale Dynamic Discrete Choice Models of Spatio-Temporal Preferences with Application to Migratory Pastoralism in East Africa

Understanding spatio-temporal resource preferences is paramount in the design of policies for sustainable development. Unfortunately, resource preferences are often unknown to policy-makers and have to be inferred from data. In this paper we consider the problem of inferring agents' preferences from observed movement trajectories, and formulate it as an Inverse Reinforcement Learning (IRL) problem . With the goal of informing policy-making, we take a probabilistic approach and consider generative models that can be used to simulate behavior under new circumstances such as changes in resource availability, access policies, or climate. We study the Dynamic Discrete Choice (DDC) models from econometrics and prove that they generalize the Max-Entropy IRL model, a widely used probabilistic approach from the machine learning literature. Furthermore, we develop SPL-GD, a new learning algorithm for DDC models that is considerably faster than the state of the art and scales to very large datasets. We consider an application in the context of pastoralism in the arid and semi-arid regions of Africa, where migratory pastoralists face regular risks due to resource availability, droughts, and resource degradation from climate change and development. We show how our approach based on satellite and survey data can accurately model migratory pastoralism in East Africa and that it considerably outperforms other approaches on a large-scale real-world dataset of pastoralists' movements in Ethiopia collected over 3 years.

DYNAMIC WAGE AND EMPLOYMENT EFFECTS OF ELDER PARENT CARE

This article formulates and estimates a dynamic discrete choice model of elder parent care and work to analyze how caregiving affects a woman's current and future labor force participation and wages. The model incorporates parental health changes, human capital accumulation, and job offer availability. The estimates indicate that women face low probabilities of returning to work or increasing work hours after a caregiving spell. I use the estimated model to simulate the caregiving, employment, and welfare effects of a longer unpaid work leave than currently available under the Family and Medical Leave Act, a paid leave, and a caregiver allowance.

A DYNAMIC MODEL OF CLEANUP: ESTIMATING SUNK COSTS IN OIL AND GAS PRODUCTION

The environmental remediation required to permanently decommission most industrial projects is an expensive, irreversible investment. Real options literature shows that temporary closure has value under uncertainty. However, even if there is no intention to restart operations, there is an incentive to label a closure as “temporary” to avoid having to remediate ongoing or future environmental externalities. I estimate a dynamic discrete choice model of closure under price and quantity uncertainty to evaluate the likelihood of reactivation. The model reveals that the option to temporarily close is being widely used to avoid environmental remediation of oil and gas wells in Canada.

Utilization of Reproducing Kernel Hilbert Spaces in Dynamic Discrete Choice Models: An Application to the High-speed Railway Timetabling Problem

This paper introduces a generalized nested logit model that results from combining discrete and continuous response variables. Reproducing Kernel Hilbert Spaces are used to define the (dynamic) systematic utilities, allowing correlations between alternatives close together on the continuous spectrum, and reconciliation mechanisms between both types of response variables are established. The seminal motivation of this model is the passenger-centric train timetabling problem. For this reason, the discussion in this paper focuses on a high-speed railway (HSR) demand-forecasting model.The model proposes a maximum likelihood approach to estimating the parameters, and a Monte Carlo simulation study is conducted to test the proposed methodology.

Modeling the Effects of Grade Retention in High School

A dynamic discrete choice model is set up to estimate the effects of grade retention in high school, both in the short- (end-of-year evaluation) and long-run (drop-out and delay). In contrast to regression discontinuity designs, this approach captures treatment heterogeneity and controls for grade-varying unobservable determinants. A method is proposed to deal with initial conditions and with partial observability of the track choices at the start of high school. Forced track downgrading is considered as an alternative remedial measure. In the long-run, grade retention and its alternative have adverse effects on schooling outcomes and, more so, for less able pupils.

Reconciling User Benefit and Time-Geography-Based Individual Accessibility Measures

This paper presents a dynamic discrete choice model of activity scheduling that features classic time-geography properties within a microeconomic framework. We present results that show how the model can produce accessibilities that form space-time prisms, while retaining the properties of traditional measures based on consumer surplus in the form of logsums. The main features of the model are that it handles time-space constraints, travel time uncertainty, and endogenous trip chaining in one consistent framework. The resulting accessibility respects the individual's time budget and fixed activities. The dynamic discrete choice framework makes possible estimation of behavioural parameters using well-known methods. Some of the remaining computational challenges are discussed. The final section provides some examples of the policy analysis possibilities provided by a model of this kind.

Enrollment and Degree Completion in Higher Education Without Ex Ante Admission Standards

Many countries organize their higher education system with limited or no ex ante admission standards. They instead rely more heavily on an ex post selection mechanism, based on the students' performance during higher education. We analyze how a system with ex post selection affects initial enrollment and final degree completion, using a rich dataset for Belgium (region of Flanders). We develop a dynamic discrete choice model of college/university and major choice, where the outcome of the enrollment decision is uncertain. Upon observing past performance, students may decide to continue, reorient to another major, or drop out. We find that ex post student selection is very strong: less than half of the students successfully complete their course work in the first year. Unsuccessful students mainly switch from university to college majors, or from college majors to drop-out. We use the estimates of our model to evaluate the effects of alternative, ex ante admission policies. We find that a suitably designed ex ante screening system (with moderate admission thresholds) can considerably increase degree completion in higher education. A discriminatory screening system for universities only, can raise total degree completion even more, though it implies a shift from university to college degrees.