Random Effects Multinomial Logit Model Research Articles

The role of forest status in households’ fuel choice in Uganda

In this study, we investigate how households' choice of energy source is influenced by the status of the local forest resource. We assume that households choose between clean fuels (e.g., kerosene, LPG, solar, and electricity), dirty biobased fuels (e.g., firewood, animal dung, crop residues, and charcoal), and mixed fuels. We integrate socioeconomic data with high-resolution satellite data on forest conditions from the Uganda National Panel Survey. The findings from a random-effects multinomial logit model indicate that households in vegetated areas are 6–7% less likely to rely solely on dirty biobased fuels, and 6–8% more likely to use mixed fuels, compared to those in non-vegetated areas. A larger forest stock is more strongly associated with lower use of firewood than charcoal. A possible explanation for the findings is the presence of policies for forest conservation and enhanced forest property rights, which improve forest conditions and limit opportunities to collect firewood. Given households' dependence on forest-based fuels, such policies could need to be modified to secure households’ access to these fuels.

Is the Bottom Line the Top Priority? A Choice Model of Revenue and Earnings Guidance

Based on the observation that managers provide almost 30% more revenue than earnings forecasts in recent periods, we jointly explore the decision to provide or forego revenue and/or earnings guidance. Our modeling innovation combines a random effects multinomial logit model, instrumental variables, and a correction for selection bias, which correctly classifies over 90% of observed guidance decisions. We find a stronger investor reaction to guidance surprises for firms whose guidance choice conforms to the predicted decision, suggesting that predictable guidance policies engender forecast credibility. The market response is driven more by managerial guidance than analyst revisions. Non-guiders quickly switch to guiding when the choice model suggests that guidance is the expected choice.

The Impact of Forest Status on Households’ Fuel Choice in Uganda

Forest firewood is a major source of household energy in Uganda. Firewood availability is strongly linked to forest status: degradation of the forest resource could force households to switch to other energy sources, which may be clean or dirty. The purpose of this paper is to investigate whether a change in forest resource status induces a switch to cleaner (e.g., kerosene, LPG, solar and electricity), dirtier (e.g., firewood, animal-dung, crop-residue and charcoal), or a mix of clean and dirty sources. To this end, we combine socio-economic data from the Uganda National Panel Survey with high-resolution satellite data on forest status and model the relationship between energy source choice and forest status using the random-effects multinomial logit model under a Generalized Structural Equation Modeling (GSEM) framework. The results confirm that forest status significantly affects fuel choice. The presence of better forests is associated with a 6-7% lower probability of relying on a basket of dirty fuels that comprise wood fuels (firewood and charcoal) and non-wood fuels (crop-residue and animal-dung) and 6-8% higher likelihood of using mixed energy relative to non-vegetated areas. The magnitudes of the effects are higher on firewood and lower on charcoal separately. Efforts on maximizing forest social benefit through participatory forest conservation interventions taking the fuel stacking behavior into account are important.

Comparisons in Cell Phone Use Rates between Talking and Manually Manipulating the Handheld Device

Cell phone use is among the most prevalent types of in-vehicle distractions. Studies have shown cell phones to have a negative impact on driving performance, with the magnitude of these impacts varying based on the level of engagement required for phone use, as well as the complexity of the driving environment. The development of effective policies and programs to mitigate distracted driving is contingent on an improved understanding of the nature of these relationships. To this end, this study investigates the rate of cell phone use under various settings through a statewide direct observation survey. Data were collected from 200 roadside locations throughout the State of Michigan and a random effects multinomial logit model was estimated to assess how the prevalence of cell phone use varies across drivers. While several studies have investigated the prevalence of cell phone use, this study distinguishes differences in the manner of cell phone use (i.e., talking on a handheld device versus manual manipulation of the device) across drivers while controlling for important contextual factors. The results show cell phone use is generally higher among female drivers and use rates consistently decrease with age. The prevalence of cell phone use is also shown to vary significantly depending on the manner of use. Use rates are shown to differ based on changes in the driving environment, suggesting potential risk compensation on the part of motorists. The results also show a significant correlation in use rates at the same sites over time, which is reflective of important unobserved site-specific factors.

Investigating influence factors of traffic violations at signalized intersections using data gathered from traffic enforcement camera.

To effectively reduce traffic violations that often cause severe crashes at signalized intersections, exploring their contributing factors seems hugely urgent and essential. This study attempted to investigate the influence factors of wrong-way driving (WWD), red-light-running (RLR), violating traffic markings (VTM), and driving in the inaccurate oriented lane (DIOL) at signalized intersections by using data collected from traffic enforcement camera in Hohhot, China. To this end, an ordinary multinomial logit model was developed. By considering the unobserved heterogeneity between observations, a random effects multinomial logit model was proposed as well. After that, the marginal effects of explanatory variables were computed. The outcomes showed that non-local vehicles were more likely to commit WWD and VTM than local vehicles. WWD and RLR frequently occurred in the daytime and evening (6:00–23:59), and on most days within a week. RLR and DIOL mainly happened in June and July. The left-turn lane ratio significantly increased RLR and DIOL. The cloudy, partly cloudy, and rainy days obviously increased WWD and VTM. The temperature from 21 to 30 degrees centigrade was apparently associated with the higher likelihoods of RLR and DIOL. According to the findings of this study, some intervention measures, targeting different vehicle types and considering temporal factors, road, and weather conditions, were recommended to reduce WWD, RLR, VTM, and DIOL at signalized intersections.

Intertemporal evaluation of household food security and its determinants: evidence from Rwanda

The food security literature has reiterated the importance of having an indicator that simultaneously captures the different dimensions of the food security definition. In this regard, this study builds on previous studies using a harmonized food security indicator within a relatively high frequency panel data from Rwanda to examine the dynamics of household food security. Summary statistics show the transitory nature of food security where households go either from complete state of food insecurity to being partially food insecure or from being partially food insecure to being completely food secure during the harvesting seasons and later fall back to their initial state during the sowing and growing seasons. This underscores the intertemporal dimension and transitory nature of food security over time. We also employed a correlated random effects multinomial logit model to further identify the determinants of households being in any of the identified states of food security. Results suggest that agricultural seasons of major food crops do determine households’ state of food security. Also, household socio-demographic characteristics, household assets, income diversification, and location of the household dwelling are some of the consistent factors that increase or decrease the likelihood of the household being food insecure.

Household Cooking Fuel Choice in India, 2004-2012: A Panel Multinomial Analysis

We use two waves of nationally representative India Human Development Survey to examine factors driving the cooking fuel choice in urban and rural India, separately. We utilize a random effects multinomial logit model that controls for unobserved household heterogeneity. We find that a clean-break with the use of traditional fuels is less likely in rural areas, but more probable in urban areas. The household characteristics (e.g. income, education) that are positively correlated with use of clean fuel also increases the probability of fuel stacking for rural households. We also find that access to paved road is an important determinant for rural household adopting clean fuel, and there exists evidence of social spillover effects in rural areas. Moreover, the bargaining power of women that is associated with economic status (e.g. education or economic freedom) is positively associated with the use of clean fuel. Finally, we find considerable impact of liquefied petroleum gas prices on the probability of use of clean fuel for urban households, but no significant impact for rural households.

Household Cooking Fuel Choice in India, 2004-2012: A Panel Multinomial Analysis

We use two waves of nationally representative India Human Development Survey to examine factors driving the cooking fuel choice in urban and rural India, separately. We utilize a random effects multinomial logit model that controls for unobserved household heterogeneity. We find that a clean-break with the use of traditional fuels is less likely in rural areas, but more probable in urban areas. The household characteristics (e.g. income, education) that are positively correlated with use of clean fuel also increases the probability of fuel stacking for rural households. We also find that access to paved road is an important determinant for rural household adopting clean fuel, and there exists evidence of social spillover effects in rural areas. Moreover, the bargaining power of women that is associated with economic status (e.g. education or economic freedom) is positively associated with the use of clean fuel. Finally, we find considerable impact of liquefied petroleum gas prices on the probability of use of clean fuel for urban households, but no significant impact for rural households.

A random effects multinomial logit model for the determinants of exit modes

PurposeThe purpose of this paper is to examine the factors associated with three modes of firms’ exit (voluntary liquidation, involuntary liquidation and acquisition) in a mutually exclusive environment. In particular, three modes of exit are treated as independent events given that different causes and consequences exist for each exit mode. The data set is a panel of 4,408 US manufacturing firms spanning over the period 1976–1995.Design/methodology/approachThe discrete choice model is used to establish a relationship between modes of exit and a set of explanatory variables, which are specific to the firm, industry and macroeconomic conditions. Use of panel data encourages us to estimate a random effects multinomial logistic regression model, which allows exit modes as mutually exclusive events and at the same time controls the firm-specific unobserved heterogeneity in the sample.FindingsThe analysis suggests that the determinants of voluntary liquidation are age, size, profitability, technology intensity and inflation level. The determinants of involuntary liquidation are size, leverage, profitability and inflation level. For acquisition, determinants are age, size, advertising intensity, Tobin’s q, GDP growth, inflation level and interest rate. The findings suggest that exit modes have a different set of determinants and the scale of effects of some common determinants such as age, size and profitability differs between exit modes.Research limitations/implicationsThe analysis presented in this study relies on data from US manufacturing firms only. Thus, there is a need to explore the determinants of exit modes in other countries as well using the proposed econometric model.Practical implicationsThe findings presented in this paper are useful for managers and policymakers to design strategies/actions for avoiding particular mode of exit.Originality/valueThis study provides empirical evidence on the differences in factors associated with exit modes and confirms the existence of mutually exclusive nature of exit modes. Findings suggest that for future empirical studies on firm exit, the exit modes must be treated as a heterogeneous event.

Modeling household cooking fuel choice: A panel multinomial logit approach

We use three rounds of a rich panel data set to investigate the determinants of household cooking fuel choice and energy transition in urban Ethiopia. We observe that the expected energy transition did not occur following economic growth in Ethiopia during the decade 2000–2009. Regression results from a random effects multinomial logit model, which controls for unobserved household heterogeneity, show that households' economic status, price of alternative energy sources, and education are important determinants of fuel choice in urban Ethiopia. The results also suggest the use of multiple fuels, or ‘fuel stacking behavior’. We argue that policy makers could target these policy levers to encourage transition to cleaner energy sources.

Party registration and party self-identification: Exploring the role of electoral institutions in attitudes and behaviors

How do electoral institutions affect self-identified partisanship? I hypothesize that party registration acts to anchor a person's party identification, tying a person to a political party even when their underlying preferences may align them with the other party. Estimating a random effects multinomial logit model, I find individuals registered with a party are more likely to self-identify with that party and away from the other party. Party registration also affects voting in presidential elections but not in House elections, leading to greater defection in the former where voters have more information about the candidates. These insights illuminate varying rates of electoral realignment, particularly among southern states, and the makeup of primary electorates in states with and without party registration.

Who Benefits from a Job Change

ABSTRACTIn this paper, we use panel data from the UK and Germany to investigate the effect of employer changes and in-firm job changes on year-to-year wage mobility of male full-time workers. Following segmentation theories and the job search theory, we study whether this effect differs for the low- and high-wage workers. As wage growth is endogenous to the decision of changing jobs, a two-stage Heckman selection approach is used. Specifically, we first estimate a random-effects multinomial logit model for the selection into a job transition and then a fixed-effects panel regression model for the wage growth. The findings suggest that both external and in-firm job changes result into substantial wage gains for the low-paid workers but not for the medium- or high-paid workers. However, the wage gain of low-paid workers due to an in-firm job change is only observed in the UK and is less pronounced than their gain by an external job change. In the German labour market, the later effect is insignificant. The results indicate that low-paid workers profit more from a voluntary change of employer in the coordinated German labour market and from a voluntary in-firm change in the liberal British labour market.

Predicting longitudinal trajectories of health probabilities with random‐effects multinomial logit regression

Researchers often encounter longitudinal health data characterized with three or more ordinal or nominal categories. Random-effects multinomial logit models are generally applied to account for potential lack of independence inherent in such clustered data. When parameter estimates are used to describe longitudinal processes, however, random effects, both between and within individuals, need to be retransformed for correctly predicting outcome probabilities. This study attempts to go beyond existing work by developing a retransformation method that derives longitudinal growth trajectories of unbiased health probabilities. We estimated variances of the predicted probabilities by using the delta method. Additionally, we transformed the covariates' regression coefficients on the multinomial logit function, not substantively meaningful, to the conditional effects on the predicted probabilities. The empirical illustration uses the longitudinal data from the Asset and Health Dynamics among the Oldest Old. Our analysis compared three sets of the predicted probabilities of three health states at six time points, obtained from, respectively, the retransformation method, the best linear unbiased prediction, and the fixed-effects approach. The results demonstrate that neglect of retransforming random errors in the random-effects multinomial logit model results in severely biased longitudinal trajectories of health probabilities as well as overestimated effects of covariates on the probabilities.

Determinants of Child Labour and School Attendance: The Role of Household Unobservables

The authors develop a semi-parametric latent class random effects multinomial logit model to distinguish between observed and unobserved household characteristics as determinants of child labor, school attendance and idleness. The authors find that much of the substitution between activities as a response to changes in covariates is between attending school and being idle, with work being rather resistant. Unobserved household heterogeneity is substantial and swamps observed income and wealth heterogeneity. A characterization of households into latent types reveals very different intrinsic propensities towards the three children's activities and those households with a high propensity to send their children to school are poorer and have less educated parents compared to households in the other classes.

A Random Effects Multinomial Logit Analysis of Using Medicare and VA Healthcare Among Veterans With Dementia

This study aimed to examine longitudinal patterns of VA-only use, dual VA and Medicare use, or Medicare-only use among veterans with dementia. Data on VA and Medicare use (1998–2001) were obtained from VA administrative datasets and Medicare claims for 2,137 male veterans with a formal diagnosis of Alzheimer's disease or vascular dementia enrolled in the National Longitudinal Caregiver Study. A random effects multinomial logit model accounting for unobserved individual heterogeneity was used to estimate the effects of patient and caregiver characteristics on use group over time. Compared to VA-only use, dual VA and Medicare use was associated with being white, married, higher education, having private insurance, Medicaid, low VA priority level, more functional limitations, and having lived in a nursing home or died in that year. Medicare-only use was associated with older age, being married, higher education, having private insurance, low VA priority level, living further from a VA Medical Center, having more comorbidities, functional limitations, and having lived in a nursing home or died. Veterans whose caregivers reported better health were more likely to be dual users, but those whose caregivers reported more comorbidities were more likely to use Medicare only. Different aspects of veterans' needs and caregiver characteristics have differential effect on where veterans seek care. Efforts to coordinate care between VA and Medicare providers are necessary to ensure patients receive high quality care.

Random effects models for operational patient pathways

Patient flow modeling is a growing field of interest in health services research. Several techniques have been applied to model movement of patients within and between health-care facilities. However, individual patient experience during the delivery of care has always been overlooked. In this work, a random effects model is introduced to patient flow modeling and applied to a London Hospital Neonatal unit data. In particular, a random effects multinomial logit model is used to capture individual patient trajectories in the process of care with patient frailties modeled as random effects. Intuitively, both operational and clinical patient flow are modeled, the former being physical and the latter latent. Two variants of the model are proposed, one based on mere patient pathways and the other based on patient characteristics. Our technique could identify interesting pathways such as those that result in high probability of death (survival), pathways incurring the least (highest) cost of care or pathways with the least (highest) length of stay. Patient-specific discharge probabilities from the health care system could also be predicted. These are of interest to health-care managers in planning the scarce resources needed to run health-care institutions.

A panel multinomial logit analysis of elderly living arrangements: Evidence from aging in Manitoba longitudinal data, Canada

Utilizing a unique longitudinal survey linked with home care use data, this paper analyzes the determinants of elderly living arrangements in Manitoba, Canada using a random effects multinomial logit model that accounts for unobserved individual heterogeneity. Because current home ownership is potentially endogenous in a living arrangements choice model, we use prior home ownership as an instrument. We also use prior home care use as an instrument for home care and use a random coefficient framework to account for unobserved health status. After controlling for relevant socio-demographic factors and accounting for unobserved individual heterogeneity, we find that home care and home ownership reduce the probability of living in a nursing home. Consistent with previous studies, we find that age is a strong predictor of nursing home entry. We also find that married people, those who have lived longer in the same community, and those who are healthy are more likely to live independently and less likely to be institutionalized or to cohabit with individuals other than their spouse.

Estimation of the multinomial logit model with random effects

In the present paper, it is shown how to perform estimation of the random effects multinomial logit model in the SAS program. Inference in the random effects multinomial logit model is complicated because it requires evaluation of multi-dimensional integrals. Presumably for this reason, no statistical software packages allow for direct estimation of the model. It is shown how modifications of the likelihood function enable the applied researcher to make inference using SAS. While the procedure works well for small samples/models, estimation time might become considerable for larger samples/models. A possible explanation for this is provided.

Determinants of Child Labor and School Attendance: The Role of Household Unobservables

The authors develop a semi-parametric latent class random effects multinomial logit model to distinguish between observed and unobserved household characteristics as determinants of child labor, school attendance and idleness. The authors find that much of the substitution between activities as a response to changes in covariates is between attending school and being idle, with work being rather resistant. Unobserved household heterogeneity is substantial and swamps observed income and wealth heterogeneity. A characterization of households into latent types reveals very different intrinsic propensities towards the three children's activities and those households with a high propensity to send their children to school are poorer and have less educated parents compared to households in the other classes.