A Multivariate Analysis of the Factors that Correlate to the Unemployment Rate. Amit Naik, Tarah Reiter, Amanda Stype

Transcription

1 A Multivariate Analysis of the Factors that Correlate to the Unemployment Rate Amit Naik, Tarah Reiter, Amanda Stype

2 2

3 Abstract We compiled a literature review to provide background information on our research of the correlation between immigration rate and unemployment rate. These articles included findings from other nations to compare with the data from the United States. We conducted a linear regression using SAS taking heteroscedastically consistent errors into account. Our dependent variable was the percentage change in unemployment rate. The independent variables that we included were; political party of the President, the majority political party of the House, change in CPI, GDP growth rate, and change in legal immigration. The time frame of our data is in the period from 1961 to Our results provided no statistically significant conclusion on the relationship between unemployment and immigration. However, we did find significant relationship between the unemployment rate and the political party in power in regards to the President and the House. Also, there was a positive correlation between unemployment and the CPI for the previously stated time frame. 3

4 Introduction People migrate to other countries to escape the horrors of war, oppression, persecution and poverty every day. The wealthy and free countries that protect human rights are often the final destination of these people. These nations are appealing and are usually more than willing to accept refugees trying to find a more peaceful, just, and prosperous nation. War is not the only reason people migrate to new locations. The promise of a more financially secure future for families remaining in the homeland and those family members also emigrating has brought many people into nations such as Canada, the United States, and the high wealth economies in the European Union. However, this can pose a problem for these host countries. In order to cope with this mass influx of immigrants and refugees, laws and statutes must be passed in order to restrict the movement of people, which could result in overpopulation of the prosperous countries. These statutes often result in people to entering into these countries illegally. Restrictive laws against illegal immigrants were put into place in Oklahoma. In 2008, The United States Bureau of Labor Statistics reported that Oklahoma s unemployment rate had fallen, while the unemployment rate in the rest of the nation had risen. This is a direct result of House Bill 1804, the Oklahoma Taxpayer and Citizen Protection act of 2007, which made it a felony to transport or shelter illegal immigrants and forbid states to issue drivers licenses or pay social welfare benefits to illegal aliens or their families. The bill also empowers the state and local police to enforce federal immigration laws. However, the state is estimated to lose $1.8 billion annually in productivity and wages as a result of this law. An estimated 70,000 illegal immigrants left Oklahoma when the legislation was passed. As the above case demonstrates, immigrants, both legal and illegal provide productivity and benefits to the societies, which they move into. However, many argue that these immigrants take jobs away from the original residents of the nation they have moved into. Illegal immigration is especially controversial, because illegal immigrants are circumventing the laws of their host nation, but at the same time many of them are providing useful services to the society, which they are living in illegally. This paper will examine the effect of legal immigration on the unemployment rate in the United States from In order to do this, it will also examine the political party in power in the White House and the House of Representatives. In general, the policies of Republicans and Democrats about immigration have radically differed. Although illegal immigration is part of the topic of this paper, it is impossible to determine exactly how many illegal immigrants are in the country at any given time. Literature Review In his article, Immigration, Income, and Unemployment, An Application of the Bounds Testing Approach to Cointegration (2007), Mete Feridun investigates the nature of the causal relationship between immigration and GDP per capita and unemployment. Feridun explores Sweden s use of autoregressive distributed lag (ARDL) bounds testing procedure and Grangercausality within vector error correction model (VECM) that is based on annual data. The author is testing whether one time series is useful in forecasting another by using the VECM, which is an econometric model, used to capture the evolution and the interdependencies between multiple time series. Results of the ARDL bounds test support the theory that the variables are in a long-run equilibrium level relationship. However, the results of the Grangercausality tests support the existence of a long run, bidirectional causality between immigration and GDP per capita. What this means is that Feridun s results do not support the hypothesis that 4

5 immigration causes unemployment, conversely the findings support that unemployment causes immigration. The working paper by Laszio Konya: Bivariate Causality Between Immigration and Long-Term Unemployment in Australia (2002), is concerned with the possible bivariate Granger causality between immigration and long-term unemployment in Australia. Konya examines the proportion of net permanent and long-term movement to resident population, and long-term unemployment is measured as the proportion of long-term unemployed to the total unemployed. The study is focused on the period of , using quarterly, and both seasonally unadjusted and adjusted data. The study is first focused on determining whether the series are stationary or has non-seasonal and/or seasonal unit roots, thus focusing on the possibility of common features, such as a common trend and seasonality, then testing for causality. The conclusions the author came to was that there is a unidirectional Granger causality, between both the seasonally unadjusted and the adjusted series. During testing of the causality analysis of the seasonally unadjusted and adjusted data, the relationship was found to be negative, which means more immigrants will lead to less long-term unemployment in the future. Konya also notes that it s apparent from our study that seasonal adjustment might significantly affect the outcomes of hypothesis tests for unit root and stationairty. In the article, Unemployment, Immigration, and NAFTA: A Panel of Ten Major U.S. Industries (1997), Catalina Amuedo Dorantes and Wei-Chiao Huang performed a pooled crosssection and time-series analysis of unemployment rates across the ten major industries in the United States from to identify the effect of NAFTA and immigration. These major industries are mining; construction; durable goods; nondurable goods; transportation and public utilities; wholesale and retail trade; finance, insurance and real estate; services; agriculture; and government. The results obtained by the authors specified that the output produced by the industry, unemployment benefit coverage, and interest rates are significant determinants of industry unemployment rates, but union presence does not affect industry unemployment. Also, immigrants from both Canada and Mexico provide necessary skills that are lacking within the United States. The authors used the Chow test, which provided no evidence that NAFTA has changed the structure of unemployment determination in these industries. The Chow test is a statistical and econometric test of whether the coefficients in two linear regressions on different data sets are equal. In the working paper, Immigration Flows and Regional Labor Market Dynamics (1998), Dominique Gross analyzed the capacity of a regional labor market to understand growing flows of immigrant workers with waning stages of expertise during relatively high unemployment. The factors that are analyzed are the characteristics of skill and the impact of the size of the flow of the immigrants. The working paper concludes that in the short run, immigration is positively correlated to unemployment. However, in the long run, immigration is negatively correlated to unemployment. Additionally, if an immigrant has a higher average skill level among other immigrants, then that makes them more effective in their job search in the short run. Finally, the author stated that, increasing the discrepancy between the skill distribution of immigrants and that of the existing workforce is desirable, as both types of labor appear to be complements in the short-run. The article, The Link Between Immigration and Unemployment in Canada by Marr and Siklos (1994), examines the relationship between immigration and unemployment rates in Canada. They assess this topic by using conditional aggregate demand and supply factors using quarterly data for the period This study is a little more original in comparison to the 5

6 some of the others examined in this paper. The authors assessed the empirical link between immigration and unemployment for Canada by applying time series methods that establish current immigration, current immigration and future immigration to be mutually determined by the past. Marr and Siklos conducted studies about transitory versus permanent effects of unemployment on immigration in addition to performing tests based on relationships estimated for every possible subsample. This is opposed to examining only some ad hoc subsample selections. The findings of this research were that increases in the unemployment rate reduced further immigration rates before After 1978, there was a positive association between past immigration and current unemployment. In the article Immigration Unemployment Relationship: The Evidence from Canada (2007), Asadul Islam determines and identifies the relationship between unemployment and immigration in Canada. Asadul used the bidirectional causality test and he found that there was no significant effect of Canadian immigration on unemployment. He concludes, the results from the causality test confirm that in the short run, past unemployment does cause less immigration, but the vice versa is not true. Also, there is a long run positive relationship amongst per-capita GDP, immigration rate and real wages. The author s results indicated that in the short run, more immigration is possibly associated with attractive Canadian immigration policies, and in the long run, as the labor market adjusts, Canadian born workers are likely to benefit from increased migration. As these various articles show, the connection between immigration and unemployment is extremely complex and indefinite relationship. This topic is one that is being researched around the world and there are many different contributing variables. Data and Data Sources This section begins with a brief description of the data we used and the sources we used to find this data. It will then explain the methodology used within this analysis. For this analysis, the response variable was the percent change in unemployment rate. We examined average unemployment rate for the years ranging from as measured by the Bureau of Labor Statistics. The response variables used in this study include; percent change in inflation rate, Gross Domestic Product (GDP) growth rate, percent change in legal immigration, the political party of the president, and the political party that holds the majority in the House of Representatives. The inflation rate used within this paper was the Consumer Price Index (CPI), which excludes changes in oil price and is calculated by examining the change in price for a basket of representative consumer goods. The Bureau of Labor Statistics calculates the CPI. The change in CPI was calculated as a percentage. The growth rate in GDP was measured using the value of the dollar in the year 2000 to remove the influence of inflation on this variable. It is important to remove inflation because inflation is already accounted for in the change in CPI. The Bureau of Economic Analysis maintains data on the GDP of the United States. This rate is expressed as a percentage. The number of legal immigrants into the United States per year was found at the Department of Homeland Security s Office of Immigration Statistics. We then took these numbers and calculated the percent change in legal immigration between years. This study was 6

7 also conducted using the number of legal immigrants per 10,000 people in the United States. The change in immigration as a percentage made our model more accurate, so this indicator was selected. The political party of the president from was found using encyclopedia sources. If the president was a Democrat, then this variable was coded as a 1. If the president was a Republican, the variable was coded as 0. Although inauguration is typically in January or February, we said that the party of the president came into power at the beginning of that year. For example, 2009 would be coded as a 1 because Barack Obama will spend the majority of this year in the White House. Likewise, 2001 was coded as a 0, even though democrat Bill Clinton was still in the White House for part of this year. Due to the lame duck period that occurs during transitions between political parties, this should not adversely affect this variables effectiveness as an indicator. The political party that held the majority of the house was also determined by using an encyclopedia. Once again, Democrat was coded as 1 and Republican was coded as 0. The equation used in this regression was: [1] SAS was utilized to run a linear regression with heteroscedastically consistent errors and variance inflation factors. This regression had two dummy variables (political parties of the president and the house) and three quantitative variables. Results: The estimate provided by SAS was: + ε [2] The resulting Adjusted R-Squared for this model is The parameter estimates and non-heteroscedastic errors, as well as the variance inflation factors can be seen in Table 1. Note that the resulting t-values are not entirely correct, because the errors used are not heteroscedastic. However, the results show that the only statistically significant indicator of unemployment is change in CPI, which is a measure of inflation. The relationship between unemployment and 7

8 inflation is direct. When inflation goes up by 1%, it is estimated that unemployment increases by 2.89%. Table 2 shows the estimates, Heteroscedastic consistent error estimates, and resulting t- values. Because the degrees of freedom for this analysis is more than 30, it is statistically acceptable to use the normal distribution. Thus, the rejection region when α=.05 is when the t- value is less than or greater than Once again, the only parameter which is shown to be statistically significant when α=.05 is the percent change in CPI. Although α=.05 is the most commonly used value for type I error rate, when α=.10, the party that has the majority in the house becomes statistically significant. If α=.20, the party of the president also becomes statistically significant. Depending on the level of Type I error rate, inflation and political party in power in the house and the White House are all statistically significant. However, Immigration is not statistically significant in this model. Analysis: As can be seen from Equation 2, each of the variables examined in this model can be shown to have an effect on the outcome, which is the percent change in unemployment rate between years. It is important to note that the dependent variable is the percent change in unemployment rate and not the change in unemployment rate. If the unemployment rate was 4% and then became 5%, that would be a 20% change in the unemployment rate, even though the unemployment rate itself increases by one percentage point. Likewise, if unemployment went from 5% to 4%, the percent change in unemployment rate would be -20. Therefore, although some of the results in our equation may sound large, they relate to the percent change in unemployment rate, not a change in the rate itself. Likewise, a decrease in the percent change in the unemployment rate does not always mean that the unemployment rate becomes less volatile. If for example, there were was already a -2% change in unemployment rate, and then a 4% decrease in the unemployment rate, the result would be a -6% change in unemployment rate. As the percent change in inflation, which is measured by CPI, increases by one percent, the percent change in unemployment rate is shown to increase by approximately 2.9%. This contradicts the Phillips Curve, which described the inverse relationship between unemployment and inflation in the 1950 s and 1960 s. The Phillips Curve was an economic phenomenon that was used as a policy making tool. In general, politicians and decision makers believed that they could control unemployment by targeting inflation and vice versa. Liberal politicians typically favored a lower unemployment rate and higher inflation, while conservative politicians typically favored a lower inflation rate which led to higher unemployment (Case, Fair, & Oster, 2008). The Phillips Curve is no longer an observable phenomenon when examining the relationship between unemployment and inflation from 1970 to the present. The reasons for this are varied. In the 1950 s and 1960 s, inflation was low and stable. Starting in the 1970 s, as the economy and trade became more global and oil shocks began, both cost and demand curves began shifting at the same time. Inflation was also a large factor in the end of the applicability of 8

9 the Phillips Curve. As the people began to expect inflation and demanding higher wages in anticipation of inflation, the Phillips Curve failed to adequately explain the relationship between unemployment and inflation (Case, Fair, & Oster, 2008). As the percentage of legal immigration increases by one percent compared to the previous year, the percent change in unemployment rate decreases by about half of a percent. This result does not support the argument that immigration is taking jobs away from Americans and leading to higher unemployment. As our research shows, immigrants typically come into the country looking for employment. Immigration laws oftentimes require immigrants to have a job established before immigrating to the United States. This would lead one to believe that as legal immigration increases, the unemployment rate should decrease, as immigrants are hired to supplement the domestic labor force. Legal immigration increasing could be viewed as a sign of an increased need for labor within the United States, which would imply a lower unemployment rate. However, this estimate for beta was not statistically significant. Therefore, we must be careful in making policy conclusions about immigration. It is possible that percent change in unemployment rate and the change in percent of legal immigration could be positively correlated. Also, this indicator does not address the effect of illegal immigrants and laborers on unemployment within the United States. Our analysis shows that when the majority of the House of Representatives associates itself with the Democratic party, the percent change in the unemployment rate decreases by a little more than 6.5%. Likewise, when the president is a Democrat, the percentage change in the unemployment rate decreases by about 5.5%. It is outside the scope of this paper to speculate as to why the unemployment rate decreases when Democrats are in power. However, this result is in line with the policy goals of Democrats during the days of the Philipp s Curve. One result, which could be anticipated, was the relationship between percent change in GDP and the percent change in unemployment rate. As our results show, as GDP increases, the percent change in unemployment rate decreases. The surprising factor is how little the percent change in GDP appears to affect the percentage change in unemployment rate. This change is less than 1%. When GDP increases, output is increasing. In theory, assuming technology is held constant; this requires more workers to generate this output. Therefore, as GDP increases, unemployment should decrease. This beta value was also not statistically significant; therefore, it is possible that the percent change in GDP may have no impact on unemployment rate. Conclusion: Although the purpose of this paper was to determine the effect of both legal and illegal immigration on unemployment rate, the chosen variables did not supply us with this ability. However, we were able to determine that the political party in power in both the House of Representatives and the White House has a statistically significant impact on the percent change in unemployment rate for the years Further research should be conducted as to the different policies enacted by the various presidents within this time frame that may have had an 9

10 effect on the unemployment rate. It is important to remember that politics does not operate in a vacuum so one must also examine world events within this time frame, such as the oil shocks, the concentration of foreign trade, and outsourcing. Another variable that would be useful to include within this research is the unemployment rates of nations where the majority of the legal immigrants into the United States come from. This would include neighboring countries such as Mexico. This information should not be difficult to find because, as can be seen in our literature review, many nations are conducting research on their unemployment rate. It appears from our results that legal immigration is actually inversely correlated with unemployment rate, and thus the more legal immigration occurring, the more the unemployment rate will decrease. The causality of this relationship is difficult to determine. This phenomenon is probably the result of the United States issuing more Green Cards and Work Visas when there are jobs to be filled for which the local labor market cannot provide the manpower or skills. The 2010 Census intends to include illegal immigrants. It is possible that after this census is conducted, a viable estimate for the illegal immigrant population within the United States may become available. If these estimates do become available, this research should be conducted again, expanding to include illegal immigration and foreign unemployment rates. 10

11 Table 1: Output with Non-Heteroscedastic Errors Variable Estimate Error t-value Pr > t VIF Intercept CPIChng LIMChng House Pres GDPCh Table 2: Heteroscedastic-Consistent Error Values and Resulting t-values Variable Estimate Heteroscedastic Consistent Error t-value Intercept CPIChng LIMChng House Pres GDPCh

12 Works Cited Case, Fair, & Oster. (2008). Principles of Macroeconomics, 9th ed. Prentice Hall. Corantes, C. A., & Huang, W.-C. (1997). Unemployment, Immigration, and NAFTA: A Panel of Ten Major U.S. Industries. Journal of Labor Research, Feridun, M. (2007). Immigration, Income and Unemployment: An Application of the Bounds Testing Approach to Cointegration. Journal of Developing Areas. Gross, D. (1998). Immigration Flows and Regional Labor Markets Dynamics. Simon Frasier University; School of International Studies. Islam, A. (2007). Immigration Unemployment Relationship: The Evidence from Canada. Monash University- Department of Economics: Accepted Paper Series, Konya, L. (2002). Bivariate Causality Between Immigration and Long-Term Unemployment in Australia. Victoria University Applied Economics Working Paper. Marr, W. L., & Siklos, P. L. (1994). The Link Between Immigration and Unemployment in Canada. Journal of Policy Modeling,