milestone / REVISION

Description/ Introduction

I would like establish whether economic events affect the outcome of elections in U.S. The dependent variable for this study would be the percentage of votes garnered by incumbent. The explanatory variables would be interest rate change and unemployment rate change. In this study, the research question would be, “Do economic events affect the outcome of elections?” This particular research question is relevant since it seeks to establish the relationship between economic events and political outcomes. In this case, this study will of both academic and political interest. The outcome of this study will help to fill the knowledge gap existing in the field of political economy whereby no appropriate study linking economic events and political outcomes has been conducted in recent times (Bateman, 2011).

The target audiences for this study would be both technical and non-technical. These audiences would include political decision makers, economic policy makers and the general public. Political decision makers would be interested in understanding what affects the outcome of elections. This means that the outcome of this study will be of great help to the political decision makers since it will help them to come up with appropriate campaign manifesto in the hope of garnering more votes. However, it is imperative to note that political decision makers need not to exclusively use the results of this study when making their political decisions. Other factors also need to be considered. For example, in the event that it is established that lower unemployment rates favor election outcome, then an expansionary monetary policy may not be the best solution since it has some negative consequences on the economy. An expansionary monetary policy may cause the country’s exchange rate to decline and hence it is not an appropriate tool for lowering unemployment rates. Thus political decision makers need to be cautious when applying the results of this study.

Economic policy makers would be interested in knowing whether the economic policies they formulate are appropriate for the electorate. For example, a policy to lower interest rates may be beneficial to individuals as it will create an incentive to borrow and invest. Policy makers are appointed by the government and hence they will be interested on the outcome of this study so as formulate appropriate policies so as ensure that the incumbent remains in power (Bateman, 2011). Finally, the outcome of this study will also help to enlighten the general public on factors to consider when electing the incumbent.

Literature Review

Some researchers have employed economic methods and techniques such as statistical analysis, theoretical analysis and the econometric analysis to establish the relationship between economic events and political outcomes. Theoretical wise, economist Samuelson is credited for attempting to theoretically explain how government economic decisions affect political outcomes. Samuelson formulated the theory of optimal public expenditure in a bid to explain how government policies affect social welfare of the citizens (Ricardo, 2001). However, his study lacked relevant empirical data to justify his claim.

Akerman, another economist, attempted to use statistical analysis in finding out how economic events affect political outcomes. He collected and analyzed the US data in order to find out whether the fundamental economic cycles referred to as 3 ½ -year Kitchin cycles were connected with the institutional change (Bateman, 2011). He found out that actually the 3 ½ -year Kitchin cycles represented politico-economy cycles that spanned the four years of presidential elections in the US (Bateman, 2011). The results of his findings showed that indeed changes in factors affecting real interest rates and employment had influence on the presidential election outcome. Other researchers have attempted to employ econometric analysis to find out whether economic events affect political outcomes. For instance, Gary Smith obtained a regression model linking unemployment rate change and percentage of votes garnered by the incumbent by using the results of the US presidential elections for four yearly periods between 1928 and 1980 (Bateman, 2011). He used the percentage of votes garnered by the incumbent as the dependent variable and unemployment rate change as the dependent variable. The outcome of his study showed that a decline in unemployment rate positively influenced the percentage of votes garnered by the incumbent (Bateman, 2011).

Some methods and techniques that have been used in the past are not appropriate for this study. For instance, focusing on the theoretical aspects without taking into consideration the empirical evidence may lead to wrong conclusions being drawn (Creswell & Plano Clark, 2007). Some of the empirical techniques used also relied on obsolete data and therefore the results of the study may not make economic sense today. But nevertheless, the past studies provide great insights into the topic. In this study, the hypothesis to be tested would be whether interest rate change and unemployment rate change affect the outcome of presidential elections. This hypothesis can be translated into an empirical model by gathering data of the specific variables and obtaining a regression linking these variables (Baltagi, 2011).

Data ONE

The data on various variables will be used to obtain the empirical model. The data on economic variables will be obtained from each state in the United States. Each state in US has an archive of data collected over the years. The advantage of state-level data is the fact that it is more representative and hence it is more likely to improve the statistical significance of the results. The regression model is then estimated using this data. The regression results to be used in explaining the outcome of the study include the coefficients of estimates, t-ratios, R2, F-values, standard errors and p-values (Baltagi, 2011). A priori, I expect a negative relationship unemployment rate and the percentage of votes garnered by the incumbent. This relationship implies that an increase in unemployment rate will lead to a decline in votes garnered by the incumbent. How this is just a priori expectation and not necessarily true.

The US election results data will be obtained from both online and print resources. The online resources to be utilized include Federal Election Commission, American Presidency Project: Presidential Election Data, Dave Leip’s Atlas of U.S. Presidential Elections, National Archives: Historical Election Results and Office of the Historian websites. The print sources to be used in the study include various book and journal articles such as the Guide to U.S. Elections (6th ed) handbook published in 2010 by CQ Press, The Election Book: A Statistical Portrait of Voting in America published in 1993 by Bernan Press, A Statistical History of the American Electorate written by Jerrold Rusk and published in 2001 by CQ Press and the Presidential Elections: 1789-2008 book published in 2010 by CQ Press. Both the online and print resources will provide crucial election results data to be used in the study.

Data-TWO

The data to be used in this study will be obtained from the secondary sources. Secondary data is readily available from various sources (Creswell & Plano Clark, 2007). However, secondary data may impose some limitations on the choice of empirical method. First, secondary data is mainly collected for a general purpose and thus inappropriate for a specific study (Özerdem & Bowd, 2010). This means that secondary data may not answer our specific research question. For example, if we intend to conduct a study by relying on secondary data collected over the past forty years and the only relevant data available is for the past thirty years then this means that we our results of the study will be compromised. Second, secondary data is subject to errors or misinterpretation based on data collection methods that were used (Özerdem & Bowd, 2010). The errors in secondary data will have a negative impact on the results of the study

Finally, secondary data collected may be lacking the required quality and sometimes the variables are expressed in an inappropriate form (Creswell & Plano Clark, 2007). Quality data refers to the data that is reliable and relevant. Reliability in this case means having an appropriate definition of the variable. In essence, the variables should be expressed in a form that can be easily utilized by the researcher. Absence of quality secondary data may lead to poor estimation of the regression model. In general, weaknesses of secondary data will have a negative impact on the overall regression results.

To address this issue, I intend to use the most reliable secondary sources of information. In this regard, I will utilize a variety of both online and print resources especially when obtaining the US election results data. Examples of online resources that I intend to use include Federal Election Commission, American Presidency Project: Presidential Election Data, Dave Leip’s Atlas of U.S. Presidential Elections, National Archives: Historical Election Results and Office of the Historian websites. I will also make use of print sources such as the Guide to U.S. Elections (6th ed) handbook. I intend to obtain data for the other variables from each state in the United States. The advantage of state-level data is the fact that they are reliable and hence they are likely to produce statistically significant results.

Empirical Approach

I intend to use the ordinary least squares (OLS) estimation technique to estimate my empirical model. The ordinary least square (OLS) regression shows the relationship between the dependent variable and explanatory variables (Baltagi, 2011). In this case, the assumption made when using the OLS technique is that the dependent variable is a linear function of explanatory variables (Baltagi, 2011). Therefore it would be appropriate to use this method since it would enable us to establish the linear relationship between interest rate change, unemployment rate change and percentage of votes garnered by the incumbent.

However, OLS estimation technique has got some disadvantages. First, presence of outliers, i.e., extremely high or small values for the dependent variable in comparison to the values of explanatory variables may lead to poor estimation of the regression model (Baltagi, 2011). Second, sometimes the explanatory variables are correlated and this may lead to a biasness and inconsistency of estimated coefficients (Baltagi, 2011). The estimated coefficients may have wrong signs which do not make economic sense. Finally, OLS estimation technique assumes that dependent and explanatory variables are always linearly related (Baltagi, 2011). This is not always true in reality since most systems are non-linear. Other possible alternative methods that can be used include the logit and probit models (Baltagi, 2011). These models are preferred over OLS in certain circumstances. For example, the logit model allows the properties of a linear regression model to be exploited. The model specification is as follows:

-The independent variable (Y) is the percentage of votes garnered by the incumbent.

- The explanatory variables are:

a) Interest rates (X2) and

b) Unemployment rate (X3).

Percentage of votes garnered by the incumbent is taken to be the dependent variable since it is influenced by the other two variables.

Since the above variables are expressed in levels, I have converted each of them into a log to obtain the log-transformed model shown below:

InYt = B1+B2 InX2t +B3 InX3t + ut

Note that:

InYt = natural log of Y

InX2t = natural log of X2

InX3t = natural log of X3

A log-transformed model helps to produce more consistent results since it eliminates heteroskedasticity. Heteroskedasticity refers to the inequality in variability of variables across a range of data values (Baltagi, 2011). Converting variables into logs helps to eliminate heteroskedasticity. The coefficients of variables in the log-transformed model can also be interpreted as elasticities. In this case, since we are regressing Y on X2 and X3 , the coefficient on X2 is interpreted as a percent change in Y holding X3 . Similarly, the coefficient on X3 is interpreted as a percent change in Y holding X2 .

Results and Robustness Check

-The preliminary results are as follows:

InYt = 1.54- 0.023InX2t -0.075 InX3t

The interpretation of each coefficient is as follows:

· The coefficient on X2 tells us that one percent increase in interest rates holding X3 constant leads to 2.3% decline in votes garnered by the incumbent.

· The coefficient on X3 tells us that one percent increase in unemployment rate holding X2 constant leads to 7.5% decline in votes garnered by the incumbent.

I had earlier predicted a negative relationship between the dependent variable and the explanatory variables and the preliminary results indicate that indeed the explanatory variables negatively influence the dependent variable. Hence the regression results are significant.

I had also anticipated existence of correlation between the two explanatory variables. Usually, OLS estimation technique assumes that there is absence of multicollinearity between the explanatory variables. To test for the presence and severity of multicollinearity, we compute the variance inflation factor (VIF) of various explanatory variables (Baltagi, 2011). If the variance inflation factor lies between 1 and 10 then there is no multicollinearity. If the variance inflation factor is less than 1 or greater than there is multicollinearity (Baltagi, 2011).

Recommendations on study and Results

Just like the results of studies that have been conducted before, the results of this study can be a potential or a pitfall to the entire research. The primary objective of this study is to establish the role played by sensitivity analysis in future success of a business. However, the results of the study may not be accurate enough to help us make the right judgment about the

study. It is therefore important to present some of the recommendations to the results of this study so that I can draw right conclusions from the study.

            The first recommendation has to do with making a decision on which results should be offered for the study and eligibility of participants. For this study, it will be important to keep in mind that the study is on sensitivity analysis and most of the issues in the study can only be identified during the review process (Doubilet et al., 1984). When the results of the sensitivity analysis are not affected by different decisions made during the research, then the researcher needs to consider those results to significant or have a high degree of certainty. In case of missing information, there will be need to find the missing information from other resources. In case this fails to work out, the results will have to be interpreted with a lot of caution. Furthermore, the reporting of the results should be done with the help of a summary table (Doubilet et al., 1984). This makes it easier to interpret the results and avoid confusion which could be risky for the business.

            Comparing these recommendations to those presented in literature review reveals a very slight difference. From literature on the concept of sensitivity analysis, it is evident that sensitivity analysis can be approached in different ways. The brute force method is suitable for a small model (Caracotsios & Stewart, 1985). Basically using method involves changing initial data and solving the model again to see a change a change in the results.    

References

Baltagi, B. (2011). Econometrics. Berlin: Springer.