Corporal punishment in Taiwanese schools: grade, gender, and location correlations

¶ … function of this study is to investigate the correlation between the frequency of corporal punishment and the students' grade level, gender, and rural vs. urban schools. Current literature examines the aforementioned correlations in other parts of the world, but no such literature exists for Taiwan. The outcomes of this study may become catalysts for policies and advocacy programs that diminish the corporal punishment of students by teachers in Taiwan.

The data necessary for this study has been gathered by various organizations in Taiwan over the years, but little has been done to analyze the data further so that a more comprehensive understanding of the corporal punishment of students can take place. The current study is based on secondary data analysis of available data from the Humanistic Education Foundation in Taipei, Taiwan. Similar to other data collection methods, secondary data analysis has certain inherent limitations. These limitations or problems include

Definitions- the definitions used by those that prepare secondary data may be different from the definitions intended for use by the researcher (Secondary Sources of information ). In addition, it often possible that certain definitions change over time and thus the analysis of data using outdated definitions may result in the development of erroneous conclusions (Secondary Sources of information ).

Measurement errors-In some cases standard deviations and standard errors are known to the researcher of the secondary data but are not published in the secondary source (Secondary Sources of information ). This can result in differences in the levels of accuracy that impact the outcome of a study or experiment. (Secondary Sources of information)

Source bias- this problem involves the ideal that those compiling information fro a secondary source may have ulterior motives and therefore paint a more pessimistic or optimistic view of the data being researched (Secondary Sources of information ). Secondary sources are often marred by the biases affiliated with the organization responsible for their publication (Secondary Sources of information ).

Reliability- problems involving reliability often occurs as a result of statistical changes over time which may not be know by the reader. For instance "geographical or administrative boundaries may be changed by government, or the basis for stratifying a sample may have altered (Secondary Sources of information )."

Time scale.- finally problems associated with the time scale may cause problems when using secondary data. For instance a census is taken every ten years so when secondary sources are publish such information may be inaccurate and out of date (Secondary Sources of information ). Thus the time period associated with the collection of the data will have an impact upon the very nature of the data (Secondary Sources of information ).

Perhaps the most serious limitation associated with the use of secondary data is that the data are only an approximation of the kinds of data that the investigator would like to employ in testing the hypothesis. In addition, there is often a foreseeable gap between the primary data the investigator would like to collect with specific research purposes in mind and the data already collected by previous investigators. Dissimilarities are likely to appear in the following areas; sample size and design, question wording and sequence, the interview schedule and method, and the overall structure of the experiment. Additionally, secondary analysis may cause other problems if the researcher has deficient information concerning how the data were collected. Such information is essential for determining probable sources of bias, errors, or other problems with internal or external validity. Based on all these abovementioned factors, the use of secondary data necessitates both the analyzing and scrutinizing of the available data in ways that are often beyond that of directly collected data.

The Humanistic Education Foundation in Taipei, Taiwan gathered the data used in the current study in 2004. The Humanistic Education Foundation is one of the grass-roots advocacy groups in Taiwan. The foundation is devoted to various educational issues, particularly school violence and the corporal punishment of students by teachers. They foundation has also conducted a series of surveys titled "Current Teachers' Corporal Punishment in Elementary and Junior High School in Taiwan" beginning in 1999.

The data for this study is based on one such survey conducted in 2004. There are several reason for not including data collected from 1999 to 2003. These reasons are as follows.

The questionnaire used in 1999 and 2000 was to pilot the study and that questionnaire has never been validated.

The questionnaire conducted in 2001 and 2002 was improved. However, the questionnaire was based on a qualitative research design, and as a result, there was very little quantitative information from the results.

Although the foundation revised several questions in their questionnaire for 2003, with the intent to create a more reliable and valid questionnaire, the 2003 data was collected only in Taipei and consisted of many missing value. Similar issues also existed in data conducted from 1999 to 2003.

In an effort to ensure improvement in the quality of data collected and broader representation of the data, the Foundation decided to make a significant investment in the 2004 survey. They enhanced the survey and overall approach of the project to address many of the issues confronting the previous versions of the survey. These improvements were successful and the data for 2004 covered a much broader geography, and specific effort was made to reduce missing values and increase response rate based on a validated and more reliable questionnaire. Thus the current research study utilizes the data from 2004.

Sampling Method

The study employs a national sample of 1,311 elementary and junior high school students in different areas of Taiwan. The sampling for the study is as follows:

(1)

Students from elementary schools

(2)

Students from junior high schools

(3)

Students from Northern, Central, and Southern regions in Taiwan

(4)

Schools from both urban and rural areas

(5)

Students represented from grade levels 1 to 9

A stratified random sampling method was implemented to depict a representative sample of students. Initially, investigators followed the Taiwan National Development Plan from the Taiwan Central Government and divided the country into three regions: North, Central and South. Each region has slight differences as it relates to culture, political and economic status, and natural environment. Next, within each region, one rural and one urban area was randomly selected. This consequently, results in a three by two stratification with urban areas in North, Central and South, and rural areas in North, Central, and South. The next step in this process was the implementation of a proportionally stratified random sample to represent elementary and junior high schools. As a result, 62 junior high schools and 162 elementary schools were selected to add to the distribution in different strata. In every school, disproportionate stratified sampling was used, with 6 students selected in each school in the sample frame. Because three elementary schools dropped out of the project, the total number of elementary schools was 159. Finally, 1344 students were selected with a response rate of 97.5%.

Strength and Limitation of Sampling Method

Primarily, researchers use stratified sampling to make certain that different groups of a population are represented adequately in the sample to increase the level of accuracy. Moreover, all other factors being equal, stratified sampling reduces considerably the cost of performing the research. The fundamental concept of stratified sampling is to make use of available information on the population to divide it into groups such that the elements present within each group are more akin to the elements present in the population as a whole. This results in forming a set of homogeneous samples based on the variables of interest. If a series of homogeneous groups can be sampled in such a manner that when the samples are combined, they constitute a sample of a larger and more heterogeneous population, the accuracy of population estimates will be increased. The procedure used for stratification does not breach the principle of random selection since a probability sample is subsequently drawn within each stratum or specific group.

Additionally, the elemental principle applied when dividing a sample into homogeneous strata is that the criteria upon which the division is based be related to the variable the researcher is studying. Another important consideration when stratified sampling is being utilized is when using these criteria, the ensuing number of subsamples does not, taken together, augment the total size of the sample beyond what would be necessary by a simple random sample. Nevertheless, if all these criteria were in fact used, the value of the stratified sample would be weakened because the number of sub-samples necessary would be enormous.

Sampling from the various strata can be either proportional or disproportional. For instance, if the number of sampling units taken from each stratum is of the same proportion within the total sample as the proportion of the stratum within the total population -- a uniform sampling fraction (n/N) -- we obtain a proportionate stratified sample. On the other hand, if proportion of the sampling units from each stratum included in the total sample is either above or below the proportion of the total numbers ( N ) in each stratum within the population -- that is, if the sampling fraction vary -- the sample is a disproportionate stratified sample. In other words, when the total number of people characterized by each variable (or stratum) oscillates within the population, to the researcher would choose the size of each sample for each stratum according to the research requirements. Such a choice is prejudiced by the probability of obtaining an adequate number of sampling units from each stratum within the final sample. As a rule, disproportionate stratified samples are used either to compare two or more particular strata or to analyze one stratum intensively (Creswell, 1994). Therefore, when researchers use a disproportionate stratified sample, we have to weight the estimates of the population's parameters by the number of units belonging to each stratum. In this sample, weighting strategies were not performed in the original data.

Once researchers have defined the population of interest, they draw a sample that adequately represents that population. The actual procedure involves selecting a sample from a sampling frame encompassing a complete listing of sampling units. Preferably, the sampling frame should consist of all the sampling units in the given population. However, in practice, such information is rarely available. For this reason, researchers usually use substitute lists that contains the same information but may not be comprehensive. In addition, there is a high degree of correspondence between a sampling frame and sampling population. The exactness of the sample depends, primarily, on the sampling frame, because every aspect of the sample design -- the population covered, the stages of sampling, and the actual selection process -- is influenced by the sampling frame. Prior to selecting a sample, the researcher must evaluate the sampling frame for potential problems. According to Kish, when inspecting the sampling frame, the characteristic problems found in sampling frames are as follows:

1. Incomplete frame. The problem of incomplete sampling frames occurs when sampling units present in the population are absent from the list. If the researcher finds that the sampling frame is incomplete, one option that might be available is the use of supplemental list. In this data, schools are sampled from the complete sampling frames, incomplete sampling frames does not exist in this sample.

2. clusters of elements, The second potential type of problem with sampling frame is clusters of elements. This problem occurs when the sampling units are listed in group rather individually. In this sample, this problem does not exist.

3. blank foreign elements. The third potential type of problem is blank foreign, which is quite common in studies. It occurs when some of the sampling units in the sampling frame are not part of the research population, such as the case where the research population is defined as eligible voters whereas the sampling frame includes individuals who are too young to vote. This problem often occurs when outdated lists are used as the sampling frame. In the current survey, this problem of blank foreign elements does not applied in this data.

Compared to a clinical, convenience or purposive sample, the sample for the current study is more generalizable by capturing a wide spectrum of data on teachers' corporal punishment of students in schools and students' responses with a range of victimization experiences. The study also reaches students who have not sought outside assistance for various reasons. Most importantly, it investigates many schools in rural areas, where there are less educational resources, compared to the area of Northern Taiwan which is more urbanized and rich in resources. Thus far, many studies have been conducted in the North but much fewer in Central or Southern regions. This data can aid us in understanding the current teachers' corporal punishment issues in these areas, which have more vulnerable educational systems.

Overall, such data is more representative nationally based on a stratified sampling method. The sampling approach also includes the students from rural or other areas in Taiwan that are often ignored by many researchers. Seemingly, using this sampling method, we can more accurately estimate the prevalence of teachers' employing corporal punishment on students. In addition, the victimization rate is more representative, and there are no prblems associated with sampling frames, which strengthens the confidence of this data. Since a high response rate (97%) is achieved in this data, it can be perceived as very representative of the overall population. However, because only elementary and junior high schools in Taiwan were sampled, the results of this study can only be generalized concerning the elementary and junior high schools in Taiwan. Senior high schools would not be part of the generalization, and the results may not be generalized to other countries in the world.

Data Analysis

All data cleansing and analysis were performed using SPSS for Windows (Version 13.0). The initial phase of the analysis involved calculating the completion (response) rate and examining various characteristics of the respondents, such as basic socio-demographic characteristics and background. Additionally, the completion rate was calculated by dividing the number of respondents who participated in the study by those prospective respondents who were contacted and met the sampling eligibility criteria. The respondents' characteristics were examined using descriptive statistics, such as frequencies, means, standard deviation, mode, and range.

The analytical methods implemented in testing the proposed hypotheses consisted primarily of the following procedures: 1) a descriptive analysis of the respondent's experiences with teachers' punishment; (2) estimating the prevalence of teachers' punishment, as well as a bivariate analysis of the prevalence by demographic and students' background; (3) bivariate analysis of experience of teachers' aggressive punishment.

The descriptive analysis of experience of teachers' punishment: Characteristics of violence experienced by the respondent were assessed by utililizing descriptive statistics. These analyses included

The specific types of punishment employed by teachers,

Witnessing of punishments

Students' academic performance

Students' behavioral and emotional response.

Content analysis was conducted on the responses to open-ended questions, such as the respondents' emotional and behavioral reactions to the teachers' punishment and the perceived influence of cultural values on teachers' response to teachers' aggressive punishment.

Estimating the prevalence of teachers' aggressive punishment: Basic statistical tools will be used to estimate teachers' punishment. Measurements of association were used to examine the association between the rates of teachers' punishment and students' basic demographics.

In order to test whether students' experiences of teachers' punishment vary by gender, the location of the schools, and students' grade levels, Chi-square, Kolmogorov-Smirnov test, and measurement of association was conducted (Creswell, 1994).

Reference

Creswell JW, (1994). Research design: qualitative and quantitative approaches.

Thousand Oaks, California, Sage Publications.

Kish, L. (1965). Survey Sampling. New York: John Wiley and Sons, Inc.

Secondary Sources of information . nd. http://www.fao.org/docrep/W3241E/w3241e03.htm#the%problems%20of%20secondary%20sources