中國在改革開放後，挾廉價勞工、充沛資源及廣大市場並與台灣同文同種的優勢，使許多台商西進投資，如今大陸已成為台灣對外投資的首選區域；而台灣對大陸貿易即使在沒有三通情況下，經由間接貿易對大陸進出口，也占台灣對外貿易主要地位。面對兩岸經貿深化情況，國人憂心會對國內經濟帶來負面效果，產生「產業外移大陸，負債徒留台灣」與「產業空洞化」的疑慮。 基於上述動機，本文透過海外直接投資相關理論，配合產業空洞化之相關指標來建立預期方向，利用經濟部對2006年所編製「製造業對外投資實況調查」之原始問卷資料，並採用 Ordered Probit模型，對於台灣製造業廠商特性、對外投資動機、對外投資地區及廠商產業別，進行廠商國內生產規模與就業人數決定因素之分析。研究發現廠商對中國等非新進國家投資可能產生減少國內投資擴廠及雇用人才的負面效果，但透過下列廠商特性與投資型態將可抵銷負面效果，並降低空洞化之產生。 (一) 對外投資地區：選擇到先進國家投資的台商，較選擇投資在非先進國家的廠商而言，更能夠擴大國內生產規模及提高國內人才雇用。 (二) 廠商特性：台商事業如果獲利與降低國外投資率將傾向於擴張國內生產規模與提高國內就業人數；台商對外投資若生產方式若以水平整合，則傾向於提高國內就業量，卻不傾向於擴大國內生產規模。 (三) 廠商對外投資動機：當廠商對外投資動機為「擴張市場」時，傾向擴大國內生產規模；但對於提高國內人才雇用卻不顯著。當廠商對外投資動機為「節省成本」時，將不傾向於提高國內人才雇用。 (四) 研發總額比：台商國內事業研發金額占國內外研發總額比率越高，越傾向擴大國內生產規模與提高國內就業人數。

臺灣和日本都面臨嚴重的高齡化問題，對於高齡化所導致的勞動力減少，最好的因應方式便是提升高齡者的勞動參與率，而日本的高齡者具有世界最高的勞動參與率，本研究藉由比較日本與臺灣的高齡者就業的影響因素，提出促進高齡者就業的方法，發現影響高齡者就業的因素可分為個人特質、經濟因素及家庭因素等。而日本樣本中本人外家庭收入和工作與否有顯著的正相關，而與子女同住及子女親族的奉養金則對高齡者就業的影響不顯著，相對於日本，子女對於臺灣高齡者就業有著顯著的影響，有子女奉養金的高齡者，就業的機會顯著的降低，若想提高高齡者就業，可以完善老人年金制度，讓臺灣的高齡者較不需依靠子女奉養，然而這卻也有可能因為財富效果反而造成就業率降低，最好的方法，無非是宣導高齡者就業的必要，改善社會對高齡者就業的偏見，並提供適宜高齡者就業的環境，如彈性的工時，輕鬆的工作等，並消除企業對年齡的歧視，如此才能活用高齡者的知識經驗，提高高齡者的就業率。

Despite the recognized importance of the bond rating industry, little academic work has been done to investigate the determinants of the standards these firms employ to assign credit ratings to individual firms. There is an ongoing debate in the literature arguing whether the decline in the percentage of highly rated firms is because rating standards have become more stringent over time or whether the credit quality of firms in the economy has declined. We investigate this question in this dissertation. Our first contribution is to address several empirical problems in prior literature. This study uses a combination of structural models of default and econometric model of ratings to study the determinants of rating standards and, by doing so, overcome the earlier methodological shortcomings. Our second contribution is to test new theory which hypothesizes that the standards of a rating agency are conditional upon the distribution of default risk in the economy at the time. The results are robust no matter which structural models of default we employ. The evidence suggests the standards are relative to the default risk distribution and there has been a secular increase in the stringency in the assignment of ratings over time. A third way we extend the literature is by examining the accuracy of the assignment of ratings. Theoretical models suggest rating agencies have incentives to purposefully add noise to the assignment of ratings. We conduct an empirical analysis of the classification errors using receiver operating characteristic analysis. The results suggest that error rates have decreased at the extreme ends of the rating spectrum (AAA vs. AA and below; B and below vs. BB and above) over time while it has increased in the middle rating categories. This error rate is directly related to the distribution of default risk across firms at any point in time. These findings not only strengthen our result that standards are relative and time varying, but also suggest there is more noise in the assignment of ratings at exactly the time when there is more uncertainty regarding the credit risk of firms in the economy – i.e., during a credit crisis.

Multivariate ordinal response data, such as severity of pain, degree of disability, and satisfaction with a healthcare provider, are prevalent in many areas of research including public health, biomedical, and social science research. Ignoring the multivariate features of the response variables, that is, by not taking the correlation between the errors across models into account, may lead to substantially biased estimates and inference. In addition, such multivariate ordinal outcomes frequently exhibit a high percentage of zeros (zero inflation) at the lower end of the ordinal scales, as compared to what is expected under a multivariate ordinal distribution. Thus, zero inflation coupled with the multivariate structure make it difficult to analyze such data and properly interpret the results. Methods that have been developed to address the zero-inflated data are limited to univariate-logit or univariate-probit model, and extension to bivariate (or multivariate) probit models has been very limited to date. In this research, a latent variable approach was used to develop a Mixture Bivariate Zero-Inflated Ordered Probit (MBZIOP) model. A Bayesian MCMC technique was used for parameter estimation. A simulation study was then conducted to compare the performances of the estimators of the proposed model with two existing models. The simulation study suggested that for data with at least a moderate proportion of zeros in bivariate responses, the proposed model performed better than the comparison models both in terms of lower bias and greater accuracy (RMSE). Finally, the proposed method was illustrated with a publicly-available drug-abuse dataset to identify highly probable predictors of: (i) being a user/nonuser of marijuana, cocaine, or both; and (ii), conditional on user status, the level of consumption of these drugs. The results from the analysis suggested that older individuals, smokers, and people with a prior criminal background have a higher risk of being a marijuana only user, or being the user of both drugs. However, cocaine only users were predicted on the basis of being younger and having been engaged in the criminal-justice system. Given that an individual is a user of marijuana only, or user of both drugs, age appears to have an inverse effect on the latent level of consumption of marijuana as well as cocaine. Similarly, given that a respondent is a user of cocaine only, all covariates--age, involvement in criminal activities, and being of black race--are strong predictors of the level of cocaine consumption. The finding of older age being associated with higher drug consumption may represent a survival bias whereby previous younger users with high consumption may have been at elevated risk of premature mortality. Finally, the analysis indicated that blacks are likely to use less marijuana, but have a higher latent level of cocaine given that they are user of both drugs.

Interaction and marginal effects are often an important concern, especially when variables are allowed to interact in a nonlinear model. In a linear model, the interaction term, representing the interaction effect, is the impact of a variable on the marginal effect of another variable. In a nonlinear model, however, the marginal effect of the interaction term is different from the interaction effect. This report provides a general derivation of both effects in a nonlinear model and a linear model to clearly illustrate the difference. These differences are then demonstrated with empirical data. The empirical study shows that the corrected interaction effect in an ordered logit or probit model is substantially different from the incorrect interaction effect produced by the margins command in Stata. Based on the correct formulas, this report verifies that the interaction effect is not the same as the marginal effect of the interaction term. Moreover, we must be careful when interpreting the nonlinear models with interaction terms in Stata or any other statistical software package. / text

This study assesses the role of crossing locations and light conditions in pedestrian injury severity through a multivariate regression analysis to control for many other factors that also may influence pedestrian injury severity. Crossing locations include midblock and intersections, and light conditions include daylight, dark with street lighting, and dark without street lighting. The study formulates a theoretical framework on the determinants of pedestrian injury severity, and specifies an empirical model accordingly. An ordered probit model is then applied to the KABCO severity scale of pedestrian injuries which occurred while attempting street crossing in the years 1986 to 2003 in Florida. In terms of crossing locations, the probability of a pedestrian dying when struck by a vehicle, is higher at midblock locations than at intersections for any light condition. In fact, the odds of sustaining a fatal injury is 49 percent lower at intersections than at midblock locations under daylight conditions, 24 percent lower under dark with street lighting conditions, and 5 percent lower under dark without street lighting conditions. Relative to dark conditions without street lighting, daylight reduces the odds of a fatal injury by 75 percent at midblock locations and by 83 percent at intersections, while street lighting reduces the odds by 42 percent at midblock locations and by 54 percent at intersections.

Obesity, the second leading cause of preventable death in the U.S., and related health problems increase people’s concerns about healthy food consumption. The increased prevalence of obesity is a major concern of societies both in developed and developing countries. Nutrition label usage has been increasing due to the link between diet and health. This study intends to provide a framework for describing profiles of consumers who are more likely to use nutrition labels in USA and Turkey, a developing country with increasing obesity rates in recent years. Empirical results present similarities and differences between consumers’ attributes for food label usage in two countries. The main contribution of this study is to investigate the relationship between the importance of serving size, while the number of expanded portion sized products in the market is increasing, and rising obesity rates. Ordered probit model analysis is used to identify the effects of demographics, health status and other components of the nutrition facts panel on selected dependent variables. Better understanding consumers’ responses to nutrition labels may guide consumers and manufacturers to broaden the communication channels through nutrition labels. The findings of this study can provide useful information to policy makers, agribusinesses, manufacturers and marketing professionals.

Food safety events receive substantial media coverage and can create devastating economics losses for agribusiness firms. It is unclear what factors influence consumers’ purchasing decisions before or after a food safety event occurs. The objectives of this study is to identify these factors that influence purchasing decisions, determine how consumers respond to hypothetical food safety events, and compare these findings across different products and geographical regions. The data for this research was obtained from two surveys. One survey concerned fresh produce while the second focused on meat products. The SPARTA model, based on the Theory of Planned Behavior, is used to determine the impact of probable factors that influence consumers’ purchasing decisions. The result of this research suggests that consumers have clearly-defined levels of trust regarding sources of food safety information. In general, a food safety event occurring in the fresh produce market seems to affect purchasing decisions more than the same event occurring in the meat market. Comparison of findings across geographical regions is less clear. Agribusiness firms can use these results to form a base strategic response plan for food safety events.

The three big credit rating agencies released their unsolicited ratings since 1996 and all of these unsolicited ratings are given to banks in Asia, especially in the emerging markets. This study aims to test whether there are differences between solicited and unsolicited bank ratings. We compare the financial profiles of solicited and unsolicited banks and investigate the factors that influence banks’ credit ratings. The empirical results show that unsolicited bank ratings are significantly lower than solicited ratings. It is seen that the financial variables of banks with solicited ratings are also better than those with unsolicited ratings. However, the profitability of banks with solicited ratings is significantly lower than those with unsolicited ratings. We see that listed and commercial banks tend to have lower credit ratings and it could be due to the fact that listed banks may face the volatility of their short-term stock prices, so their operating strategies are influenced by market noise, which leads to inferior performance. The reason why commercial banks tend to have lower credit ratings is that commercial banks face so fierce competition that their profitability is compressed. In the last section, we use an ordered probit model to examine the determinants of Fitch’s rating. We find that sovereign credit risk, solicited status, listed status, bank specialization, profitability and asset quality are the major factors influencing Fitch’s bank credit ratings.

Correlated ordinal data typically arises from multiple measurements on a collection of subjects. Motivated by an application in credit risk, where multiple credit rating agencies assess the creditworthiness of a firm on an ordinal scale, we consider multivariate ordinal regression models with a latent variable specification and correlated error terms. Two different link functions are employed, by assuming a multivariate normal and a multivariate logistic distribution for the latent variables underlying the ordinal outcomes. Composite likelihood methods, more specifically the pairwise and tripletwise likelihood approach, are applied for estimating the model parameters. Using simulated data sets with varying number of subjects, we investigate the performance of the pairwise likelihood estimates and find them to be robust for both link functions and reasonable sample size. The empirical application consists of an analysis of corporate credit ratings from the big three credit rating agencies (Standard & Poor's, Moody's and Fitch). Firm-level and stock price data for publicly traded US firms as well as an unbalanced panel of issuer credit ratings are collected and analyzed to illustrate the proposed framework.