The idea of elderly in-migrants as an important factor or stimulus to local economic development (Serow, 2001) has been confirmed by quite a few studies such as Bennett (1993); Carlson, Junk, Fox, Rudzitis, & Cann (1998); Day & Barlett (2000); Hodge (1991); Serow & Haas (1992); and Stallmann, Deller, & Shields (1999). Large-scale elderly in-migrants can bring several benefits to local economy. First, they can increase property and sales taxes, counties' largest source of revenues, without directly increasing their greatest expense such as public education; also, in-migrant retirees as a large portion of elderly do not compete for jobs so that most of counties consider them as net economic assets (Day & Barlett, 2000; Glasgow, 1991; Graff & Wiseman, 1990; Rowles & Watkins, 1993; Schneider & Green, 1992). Second, large-scale elderly in-migrants can increase local sales and capital pool through investments and savings (Campbell, 2005). Third, they can stimulate job creation and service development (Campbell, 2005). Thus, more and more counties are competing for elderly in-migrants as a source of local economic development. The question of what factors attract elderly in-migrants has been put forwarded by county governors who need to make good strategies or policies to pull them in. However, most previous studies on analyzing those factors of elderly in-migrants have been focused from macro levels such as national perspective, southern US, or state level. Little research has been conducted from a micro level of counties which are increasingly competing for elderly in-migrants with each other. The objective of this study is to determine the factors to attract elderly in-migrants from the perspective of counties in Tennessee. The main contribution of this study is to find out the county characteristics in Tennessee that attract elderly in-migrants and then provide policy implications for county governors on how to pull them in. Literatures such as Serow et al (1996 and 2001), Longino (1995), Newbold (1996), Campbell (2005), Gabriel and Rosenthal (2000) provide intuition and background for empirical models used to do the regression analysis in this paper. According to these studies, the factors of elderly in-migrants, in terms of county characteristics, include economic and non economic aspects such as income, employment, taxes, education, safety/crime rates, population (or population density), and elderly population rate. Based on these factors, the following empirical models are set up to do the regression analysis. A linear fixed-effect model is the conceptual model for this paper, instead of random effect model, because only individuals of the sample obtained are focused on and inferences are drawn restricted to these individuals within the sample (Baltagi, 2005). In other words, the linear fixed-effect model is an appropriate specification for this paper because the sample selected in this paper includes all the counties in Tennessee so that the sample is not randomly selected. Also, only those counties in Tennessee are focused on, and inferences are drawn restricted to those counties in Tennessee. Furthermore, Hausman tests are conducted in the next section to confirm that fixed-effect models should be used instead of random effect models (Baltagi, 2005). Two groups of linear fixed-effect empirical models are used to do the regression analysis. The dependent variable for the first group is in migration rate (per 100 persons) of the 60-plus cohort, which is interpreted as a percentage; and the dependent variable for the second group is in migration rate (per 100 persons) of the 67-plus cohort, which is interpreted as a percentage. The independent variables for the two groups include percentage of people with 65-plus over the whole population, percentage share of police expenditure over total expenditure, percentage share of highway expenditure over total expenditure, percentage of white people over the whole population, percentage of population with high school degree over the whole population, medium family income, property tax assessment, employment, population or population density, county dummy, and year dummies. Data is county level and collected through US Census Bureau. The data includes ninety-five counties in Tennessee for five years of 1962, 1972, 1982, 1992, and 2002 (see table 2 for data statistics summary). Also, the data is balanced except that in migration rate (per 100 persons) of the 67-plus cohort for 2002 cannot be obtained. 475 observations are used to be regressed for the first group of models and 380 observations are used to be regressed for the second group since data for 2002 has to be dropped for the second group. Stata software is applied to do the regression for the two groups of fixed-effect models. The results indicate that the elderly in-migration rate is positively correlated to the share of elderly people over the whole population, the rate of people with high school degree, medium family income, and population (or population density). Also, it is no difference in terms of the results either population or population density is used as one of the independent variables. County governors could make appropriate strategies or policies to pull those elderly in according to the results by improving amenities or life quality for elderly in each county. The weakness of this paper is that it is hard to test the endogeneity of independent variables because 1) there is no instrumental variables so that the hausman test result calculated from the difference between the original model and 2SLS cannot be conducted; and 2) it is hard to get all the factors out of the error term as explanatory variables and as a result, it is difficult not to allow correlation between the explanatory variables and unobserved factors in the error term.