Role of Statistics in Politics - Research Paper Example

Can statistics tell us anything important about politics? Introduction Politics is defined as the process by which groups of people make decisions. It is clearly known that today’s politics is corrupted with wrong information or inadequate information which is vulnerable to manipulation by superior political powers. Hence there always arises a need to validate these decisions with strong evidence and examination. Social scientists employ critical analysis to study objects of enquiry they regard as inconsistent with the scientific approach. In other words, hermeneutic methods are adopted to extract the pure knowledge from the available information. It has been proved widely through all fields that statistics is the only tool that converts real-world information into knowledge. Politics is referred to the national accomplishment of the system of governance at the federal, state and local level. Evidences have been found that statistics had its origin in politics. Historically, politics required the collection, compilation and analyses of quantitative information gathered from the official records or surveys conducted by the government. Politicians highly considered the role of statistics and the use of statistical methods to make decisions of national significance. The last five decades has noticed the inevitable usage of statistical or quantitative methodologies by social scientists for their research. This trend is found to be on the increase with more complex quantitative research on social sciences appearing on popular journals. Statistics enables the scientist or researcher to study enormous qualitative or quantitative data on the human society in a simple manner. Several tests can be performed to examine the validity and reliability of the data. With this approach, the scientist is able to provide concrete conclusions on unclear political or social concepts. Measuring political concepts There are several methods available in statistics to measure political situations, concepts or issues. The media plays a major role in providing all kinds of information, relevant or irrelevant to a political situation. Analyzing this information to obtain the valid and reliable knowledge is a crucial task. The political concepts can be qualitative, pertaining to immeasurable data or quantitative, pertaining to numerical data and these can be studied by applying suitable statistical methods. Based on the issue of study, the variables or factors contributing to the issue are recognized. The best construct or definition which explains the issue is framed. The expected conclusion or understanding of the issue is predefined. It is this conclusion or hypothesis that is tested using statistical methods. When the test appears positive based on statistical criteria, the hypothesis is accepted else rejected. These test also provide the probability (measure of chance) at which the conclusion remains true. A few political concepts that have been analyzed using statistics are measuring the probability of winning an election, measuring the relationship between gender of voters to the nationality of the elector and so on. Standards of measurement The political surveys conducted by the government or the media to obtain public opinion on topics of national interest contain huge amount of data which have to be gathered, organized and analyzed in the most efficient manner. Sampling is a methodology used to study about a million people through just 1000 (Stephen Ansolabehere 2003). Statistics provides various samplings methods, such as Simple Random Sampling, Stratified Random Sampling, Cluster Sampling, to gather such maximum data at the least cost and time. The methods applied to obtain valuable information from the huge dataset can be the simplest descriptive statistics consisting of average and range, correlation and cross-tabulation to complex forecasting models. Choosing the statistical method for analysis depends completely on the political situation, defined hypothesis and the measured variables or factors under study. It is a common practice of most politicians in developing countries to promise during election time to eradicate poverty and hunger to attract votes. The facts and figures given by these politicians on the extent of hunger have always been a topic of scrutinization for social scientists and political researchers. For this purpose, past data relating to hunger can be obtained from local agricultural departments which administer the food stamp program. More data can be obtained from social organizations, church groups and child welfare foundations as they provide food to the deprived. More useful information can also be obtained from food manufacturers and distributors. Secondary data on the studies and surveys conducted on hunger can be obtained from periodicals, journals and newspapers. This information can be analyzed to compare all previous studies and surveys to obtain the truth. In most cases, the studies and surveys do not provide the same results. One popular statistical methodology used in the analysis of survey is the Probability theory which provides a measure of uncertainty (Leonard 1972). The Bayesian Probability theory is widely used in the measurement of politics. These tools provide politicians the chances of winning the election and so on. The possibilities of acceptance or rejection of a government policy among the legislative members can also be studied in advance. Given below is an example that explains the application of statistics in political environment at the time of election. The following example explains the application of the Bayesian probability theory to predict with high probability that a randomly chosen person is a councilor or political activist (Paul Whiteley 1976). It is assumed that probability of finding a councilor and elector are equal. It is in such ways that statistics is used to analyze political situation. Consider the example of the procedure for revising probabilities. Given: P (C) The prior probability that an individual is a councilor P (E) The prior probability that an individual is a elector P (M/C) The Likelihood that an individual is male given that he is a councilor P (M/E) The Likelihood that an individual is male given that he is an elector Given the probabilities, the following result is obtained: The result shows a high probability of political activism among males which is due to the naïve prior probabilities. To obtain such sophisticated prior probabilities, severe research on different resources is required to and the statistician can validate the reliability of the data. Usage of Statistical models Statistics is a science dealing with small and large sets of numbers. The best way to handle these numbers in a concise manner is to frame equations that describe them. A statistical model is simply one or more mathematical equations consisting of random variables their associated probability distributions thereby describing the behavior of the object of study. The fundamental idea of statistical modeling is to consider the data as observed values of random variables. It is these random variables carrying numerical information that need to be studied and tested to reach specific conclusions. The variable to the left of the “=” sign is called the response or outcome variable. The variables to the right of the “=” sign are called explanatory variables. These variables form an arithmetic expression describing the relationship between them. A statistical model is as given below: Response variable = f (explanatory variables, random noise, parameters). The parameters are estimated from the data. The statistical model is then used to obtain information and/or prediction. For this purpose, few statistical methods used are linear regression, logistic regression, analysis of variance and so on. Apart from knowing the values of the parameters, the models also depend on the distribution of the random noise. In such case, it is called a non-parametric model. When the distribution of the random error is known or assumed, it is called a parametric model. A statistical model tries to measure the relationship between the explanatory variables and the response variable thereby considering the random or natural disturbances. The measurement of this random disturbance plays a crucial role in determining the exact relationship between the variables of study. The model also is sensitive to the nature of the variables when being continuous, discrete, categorical and binary and so on. The distributions applied in these models follow certain rules or assumptions. The most commonly used distribution is the normal distribution which follows the principle that most values are located around the average. The following example uses a statistical model to study the influence of certain factors on whether or not a political candidate wins an election.  Here, the outcome (response) variable is binary (0/1); win or lose.  Hence a logistic regression model is used (UCLA). The variables or factors of interest are: the amount of money spent on the campaign, the amount of time spent campaigning negatively and whether or not the candidate is an incumbent. Cross-tabulation method of analysis and chi-square analysis are commonly used to measure the relationship between variables of study. The suitable statistical method that handles binary variables is adopted as the response variable is binary. The next step of the research is to measure the effect of one factor on another and all factors on the response variable. Different statistical techniques are available to measure such effects, for example: the effect of police on crime, effect of electoral rules on the translation of votes into legislative seats. Despite all the information obtained from the statistical models, it is unfortunate that these models have become the bible of statisticians. The models have to be framed to find solutions for problems (King, G. R. Keohane and S. Verba, 1994). However, the more attention is given to only framing models therein forgetting the problem. This affects the indigenous nature of the statistical model. The level of significance of above 5%, obtained after testing the model is considered as the most reliable result and conclusions are made accordingly. Several studies have been performed and concluded only based on this level of significance which could be false. To avoid such ambiguity, post-hoc tests must be conducted to obtain the right solution to the problem under study. Likelihood Models and likelihood are very important to data analysis and modern statistics. Literally, likelihood and probability have the same meaning. But statistically they are different. It is comprehendible that likelihood is the reverse of conditional probability. That is, probability enables us to measure and predict unknown outcomes from known parameters. But likelihood enables us to estimate unknown parameters from known outcomes. A likelihood function is described as a conditional probability function considered as a function of the second argument (parameter), with the first argument (outcome) held fixed. The probability model of approach depends on the level of significance based on a test statistic assuming a normally-distributed and homogenous error term (Edwards 1972). But the likelihood approach compares the strength of two models based on the available data and chooses the best model (Cary Institute of Ecosystem Studies 2009). However the assumption on the error term is maintained. Thus a difference between likelihood and explanation is noticed. Correlation and Causation Correlation is the most useful statistic when studying two or more variables. It is a measure of the strength of the relationship between two variables. Only when the relationship between two variables is noticed, a statistical model can be framed to describe this relationship. The relationship can be linear or non-linear. That is when variables or events are correlated; a systematic pattern can be noticed. However, correlation only means that two variables are related. The coefficient determining the strength of this relationship is between -1 and +1. Strong relation between the variables exists when the coefficient is 1. A positive correlation means that as the value of one variable increases the value of the other variable also increases. However, it does not mean that the value of one variable increases because of the other variable. That is, correlation does not mean causation in any way. Making conclusions only based on correlation does not make sense and would lead to false conclusions. However, strong correlation does often demand further investigation to determine causation. It should also be noted that the correlation coefficient obtained from the data set under study does not represent the entire population. Hence proper care must be taken while making conclusions. Correlation in political scenarios has been noticed several times that determine the relationship between two or more factors. It can be found whether the dressing style of a politician has an impact on the election results. Similarly, the correlation between regime type and political discrimination can also be studied. This can be used to test the hypothesis that the conflict potentiality of socioeconomic inequalities increases with the level of political elimination of minority groups in a country (Gudrun 2006). Causation is defined as causing or producing an effect (Cambridge 2000 memos 2001). The misassumption of correlation meaning causation has caused worst afflictions in the British chattering classes including the BBC. To prove causation, the mechanism of the cause should be known and verified experimentally. The surveys conducted are never examined for the validity and the results are taken for granted. The surveys that notice a correlation between real-life factors should not declare that one factor is caused due to the other or changing one factor would improve the other factor. For example, a survey found that red cars were twice likely to meet with accidents than blue cars. A relation between accidents and colour of car is noted. If it means that changing all red cars to blue colour would prevent accidents then it is causation. Correlation only means the importance of two variables in the occurrence of an event while causation means the sufficiency of the two variables for the occurrence of an event. Proving the causation is harder and more expensive than correlation. Hence, conclusions made based on correlation only should be done very carefully. The theory of modernization can be explained as the causal relationship between economic and political development (Bert Kritzer 2006). The theory states that variables such as per capita income, urbanization, high levels of industrialization and education lead to increase the level of democratic participation. Modernization in a country is distinct with industrialization; industrialization creates a need for labour in urban and developed areas. This causes the ruralites to migrate to urban areas. And the urban class finds itself in an environment that is more conducive to political participation. Industrial employment itself creates shared interests for workers as a class, often leading to the formation of trade unions to bargain for better working conditions. Thus it is understood that causal relationship exists between economic development and political development. Who can use statistics? The usage of statistics is far and wide penetrating to all fields, industries and businesses (Andrew Winthorp 2007). It has become the standardized unit of measurement to present data in a meaningful and useful format. Statistics simplifies the handling of huge amount of data, enables easy understanding through simple terminologies and enables prediction of future performance based on past records. The insights got into the performance of a business unit based on the data helps the business management to make changes in the organization goals and objective to improve performance. This ensures proper planning and execution thereby leading to a successful business management. Similarly, statistics is exploited by governments also while formulating policies and laws. Academic organizations use statistics to improve the performance of the students and gain reputation and respect. Advertising agencies also use statistics to study the target markets and organize campaigns and promotions.   The use of statistics also expands to politics to measure degree of uncertainty at the time of election (Steve Gillman 2009). Though ample amount of political data is available, relying on this information completely is always a question. Political scientists and researchers then apply statistical methodologies to extract information from these data. Statistics is believed Actual information exposure is therefore more widespread in the population than would seem to be revealed by knowledge-based measures of information retention. Conclusion Despite the overwhelming power of statistics to make clear information is evident, the high possibility of misusing the tool is prevalent in politics. The interpretations and understanding of the results of any statistical analysis of political data can vary from person to person. This inconsistency in reporting allows manipulation of information of personal requirements. This problem arises with statistics presented by news channels, newspapers etc. Election reporting is a perfect platform to demonstrate lying with statistics. It can be noted that the vote percentages of the significant parties always add up to 100%. But this is possible only when no votes have been gained by other candidates. In reality, other parties often get several percent of the votes. However, the news agencies only report the vote percentages of the parties they consider is important. The participation of several other parties has been ignored and results have been manipulated to highlight the political significance of certain powers. Being aware of such irregularity is the only way to obtain the accurate information. References 1. Andrew Winthorp, 2007. Using Statistics to Improve and Measure Business Performance. [Online] Available at http://ezinearticles.com/?Using-Statistics-To-Improve-And-Measure-Business-Performance&id=744164. [Accessed 12 January 2010] 2. Ben Kritzer, 2006. The Relationship between Political Development and Modernization. UW-Madison. [Online] Available at http://users.polisci.wisc.edu/kritzer/Teaching/ps551/example2.pdf [Accessed 12 January 2010] 3. Cambridge 2000 memos, 2001. Correlation and Causation. [Online] Available at http://www.cambridge2000.com/memos/correlation.html [Accessed 16 January 2010]. 4. Cary Institute of Ecosystem Studies, 2009. Likelihood and Information Theoretic Methods in Forest Ecology – Models as Hypothesis. [Online] Available at http://www.sortie-nd.org/lme/Course_Schedule_2009/Day_2/Lecture_3_Models_as_Hypotheses.ppt [Accessed 16 January 2010]. 5. Davison A. C., 2003. Statistical models. Cambridge University Press. [Online] Available at http://books.google.co.in [Accessed 17 January 2010]. 6. Edwards A. W. F., 1972. Likelihood. Cambridge University Press. [Online] Available at http://books.google.co.in [Accessed 17 January 2010]. 7. Field A., 2009 Discovering Statistics Using SPSS, Third Edition. Sage Publications Ltd. Available at http://books.google.co.in [Accessed 16 January 2010] 8. Gudrun Østby, 2006. Horizontal Inequalities, Political Environment and Civil Conflict: Evidence from 55 Developing Countries. University of Oxford. [Online] Available at http://www.crise.ox.ac.uk/pubs/workingpaper28.pdf [Accessed 15 January 2010]. 9. UCLA: Academic Technology Services, Statistical Consulting Group. Introduction to SAS. [Online] Available at http://www.ats.ucla.edu/stat/sas/notes2/ [Accessed 12 January 2010]. 10. King, G. R. Keohane and S. Verba, 1994. Designing Social Inquiry: Scientific Inference in Qualitative Research. Princeton University Press. [Online] Available at http://press.princeton.edu/chapters/s5458.html [Accessed 11 January 2010] 11. Leonard J Savage, 1972. The Foundations of Statistics. Dover Publications. [Online] Available at http://books.google.co.in [Accessed 16 January 2010] 12. Michael M. Bechtel and Dirk Leuffen. Forecasting in the Study of European Union Politics: A Note on How to Derive Out-of-sample Forecasts in Political Time Series Analysis. [Online] Available at http://www.ib.ethz.ch/people/mbechtel/box_feeder/eu_forecasting.pdf [Accessed 15 January 2010] 13. Paul Whiteley, 1976. Bayesian Statistics and Political Recruitment: A Comment. Cambridge University Press. [Online] Available at http://www.jstor.org/pss/193370 [Accessed 13 January 2010] 14. Stephen Ansolabehere, 2003. Introduction to Statistics for Political Science: Introduction. Massuchusetts Institute of Technology. [Online] Available at http://ocw.mit.edu/NR/rdonlyres/Political-Science/17-872Spring2004/EAA12458-C8FE-4925-B847-D84FEA331F1A/0/intro_sts_p12003.pdf [Accessed 15 January 2010] 15. Steve Gillman, 2009. Lying with Statistics – Politics as Usual. [Online] Available at http://www.articlesbase.com/politics-articles/lying-with-statistics-politics-as-usual-835879.html [Accessed on 13 January 2010] Read More