Suchergebnisse

Purpose
The purpose of this paper is to analyse the determinants of actual evacuation decisions of victims of the unprecedented 2014 year-end flood disaster which wreaked havoc across two east-coast states in Malaysia.


Design/methodology/approach
The target population of this study is the group of victims affected by the December 2014 flood in the Malaysian east-coast states of Kelantan and Pahang. Sampling frames of the flood victims were obtained from the National Security Council offices of the two states. The empirical analysis of this paper is based on a unique data set obtained from a questionnaire survey of the flood victims. The final working sample consists of 372 respondents.


Findings
Important findings from this study are: victims who were given evacuation notices were five times more likely to evacuate, victims who participated in flood awareness programmes were less likely to move to evacuation centres, the further away victims' homes were from the evacuation centres the more likely they were to evacuate, older victims were less likely to evacuate, larger households were more likely to evacuate, and victims with tertiary education were also less likely to evacuate.


Originality/value
This paper is unique because previous studies of Malaysian flood-related disasters are confined to floods of regular magnitude. This paper is also unique because it uses a semi-parametric estimation approach to obtain the marginal effects of the explanatory variables on evacuation decisions.

'Die in den Sozialwissenschaften, den Wirtschaftswissenschaften und der Biometrie bekanntesten Modelle für binäre abhängige Variablen sind das Probit- und Logitmodell. Als Verteilung der Fehlervariablen verwendet man dabei die Normalverteilung beziehungsweise die Logistische Verteilung. Beide Modelle liefern in der Regel ähnliche Schätzungen. Ist die Verteilung der Fehlervariablen schief, so führen Tests schnell zur Ablehnung beider Modelle. Beim Gallant-Nychka Ansatz wird die Verteilung geeignet approximiert. Neben den Parametern des Modells sind simultan die Parameter der Verteilung zu schätzen. Simulationen zeigen, daß die Schätzung bei normalverteilten Fehlervariablen fast genauso effizient wie im Probitmodell, jedoch viel besser bei Abweichungen von der Normalverteilung ist.' (Autorenreferat)

In many economic models, objects of interest are functions which satisfy conditional moment restrictions. Economics does not restrict the functional form of these models, motivating nonparametric methods. In this paper we review identification results and describe a simple nonparametric instrumental variables (NPIV) estimator. We also consider a simple method of inference. In addition we show how the ability to uncover nonlinearities with conditional moment restrictions is related to the strength of the instruments. We point to applications where important nonlinearities can be found with NPIV and applications where they cannot.

Kernel estimation is a common nonparametric method for data based estimation of densities or regression functions. Although one may consider nonparametric estimation as a procedure that does not involve parameters, one has to estimate bandwidth parameters. The difference to parameter based estimation of e.g. density functions is that no specific form of the nonparametric density has to be assumed. This makes nonparametric estimation methods more flexible. This thesis compromises three parts. The first part covers bandwidth selection in kernel density estimation, which is a common tool for empirical studies in many research areas. The discussion about finding the optimal bandwidth based on the data has been going on over three decades. The typical aim of empirical studies in the past was mostly to show that a new method outperforms existing ones. Review articles on comparing methods are very rare and were written a long time ago. Hence, the first part of this thesis is an update review of existing methods comparing them on a set of different designs. The second part is on bandwidth selection in nonparametric kernel regression. The aim is similar to the first part: reviewing and comparing existing methods on a set of designs. In part one and two, smooth densities of a random variable X were assumed, therefore global bandwidth selection is adequate for the kernel estimation. In contrast to the first two parts we assume a density of X with a sharp peak and smooth areas in the third part. Usually local bandwidth selection is used in this case. However, we want to apply global bandwidth selection methods and hence, it is tested if good results can be obtained by a prior transformation. Therefore, part three covers a comparison between using a transformation and estimating the global bandwidth without a transformation. The main question is whether an improvement with respect to the typical error criteria in nonparametric regression can be made by using a prior transformation.