Understanding teen mothers: a ZIP code analysis.

Misra, Kaustav ; Goggins, Kylie ; Matte, Amber 等

I. Introduction

"True motherhood is the most beautiful of all arts; the greatest of all professions," (McKay, 1957, p. 116). In general, most who experience motherhood would probably agree with this statement. But for those mothers who experience pregnancy as a teenager, the point could also be argued that along with this beautiful gift comes an abundance of challenges. Undoubtedly motherhood, and parenthood in general, is always a challenge, but teenage pregnancy adds to this tribulation. Perhaps one of the greatest challenges teenage mothers face is the inability to financially support their new families. Because of this, the government strives to provide social and economic programs to assist new, young mothers. Additionally, social costs of teen motherhood are reflected in foster care expenses, smaller tax revenues raised due to smaller tax bases generated by teen parents, higher rates of incarceration for adult sons of teen mothers, and additional costs for the provision of medical care for young children (Hoffman, 2006). In 2004, it was estimated that taxpayers in the U.S. spent $9 billion on children bom to teens; and in 2008 the dollar amount was estimated at $10.9 billion (Hoffman, 2006; and National Campaign, 2011b). While the rates of teen births have fallen in the U.S. (by approximately 33 percent from 1991 to 2008), as of 2008 there were more than 400,000 teen births annually. On average, taxpayers in the U.S. spend $1,647 annually per child born to a teen mother. And in Michigan alone, the cost to taxpayers for teen births was $308 million in 2008 (National Campaign, 2011b; 2011c).

While the rates of teen motherhood are declining, national and state specific levels still represent substantial costs to society at large (Kearney and Levine, 2011, Kearney and Levine, 2012a). Three in ten girls in the U.S. become pregnant at least once by age 20 (National Campaign, 2011a). State comparisons show that adolescent mothers are the most likely to stay in poverty in the long run (Moore, Glei, Driscoll, Zaslow, and Redd, 2002); live in poor communities (Maynard, 1996), give birth to an unhealthy child, and fail to complete a proper education (Klepinger, Lundberg and Plotnick, 1995).

Poverty implies a state of condition in which an individual faces limited resources to do well enough in his or her day-to-day life. Poor teens represent the portion of the teenage population that suffers from this inadequacy. Having taken notice of the substantial social and financial costs of poverty, policy makers, as well as government and non-government officials, have placed increasing importance on the topic--seeking to understand and address its causes, consequences and means of effectively fighting and preventing it. Interestingly, the relationship between poverty and teen mothers has not been explored before from a spatial aspect, which brings us to the contribution of this paper. The primary goal of this paper is to investigate how teen motherhood is associated with poverty at the ZIP code level.

Rates of teenage motherhood have been found to vary across the country as well as within a state (Kearney and Levine, 2012b). For this reason, it is beneficial to be able to identify the spatial patterns that underlie varying rates of teenage mothers. We attempt to study this relationship at the ZIP code level in the state of Michigan. Utilizing this narrow geographic unit of measure, our goal is to perform analysis and form conclusions regarding teen mothers that are more precise than earlier studies and studies that use larger, alternative geographic units, such as state-level or county level data. The underlying theory is that heterogeneity in important relationships between teen motherhood, poverty, and correlated community characteristics might be eclipsed by employing county (or state) level data. By exploring these relationships at the ZIP code level, we hope to reduce this type of noise; we also present estimates at the county level for comparison purposes.

II. Literature

Research regarding teen mothers is not a recent phenomenon; it dates to the early 1990s. The literature suggests a variety of explanatory factors for teen pregnancy and births--poverty is among those most frequently cited as contributory influences (Moore, Glei, Driscoll, Zaslow, and Redd, 2002; Maynard 1996; Klepinger et al., 1995). Throughout the literature, studies that examine the relationship between poverty and teen motherhood often fall into one of two opposing sides of a cause-and-effect, or chicken-versus-egg, debate.

Furstenberg (1991), for example, analyzes whether teen births drive poverty, as opposed to poverty driving teen pregnancy. The study assesses the argument that teens see becoming pregnant at a young age as a positive addition to their lives. Furstenberg ultimately refutes the idea, concluding that teen motherhood is more of an "adaptive response" to poverty. Studies that hypothesize that poverty drives teen births include Rich-Edwards (2002). While the author acknowledges that in some cases teenage childbearing and the receipt of welfare assistance are linked--observing a logical step from teen births to poverty--Rich-Edwards finds the life trajectories of young women in the U.S. are not substantially changed by the fact that they became mothers in their teen years, suggesting that teen mothers are not a cause of poverty, but rather a consequence.

Research suggests that teen mothers who are disadvantaged in their youth, or come from poverty, will stay in poverty after becoming a teen mother. A sixteen-year longitudinal study conducted by SmithBattle (2007) observes a group of nine white and seven African-American teen mothers every four years over a sixteen-year period. The participants differ with regard to marital status, socioeconomic status, parental education and family structure. The author finds that in most cases, teen mothers attain the same social class position as their family had when they were growing up. SmithBattle argues that the trajectories of the teen mothers are dominantly influenced by their "opportunities" as children--as illustrated by the finding that teen mothers who were considered "poor" at the end of the sixteen-year observation period generally came from poor backgrounds. SmithBattle (2007) also finds that teen mothers respond heavily to the opportunities and resources that are readily available to them.

Proudfoot (2008) supports the SmithBattle (2007) argument that impoverished childhoods lead to poverty among teen mothers. Proudfoot describes the life of a specific teenage mother who goes on to attend college. Identifying such an outcome as a rarity among teen mothers supports the hypothesis that teenage mothers generally have poor educational attainment. Proudfoot argues that although the chance of completing higher levels of education may diminish upon becoming pregnant as a teen, if the teen's parents went to post-secondary school, her chances increase.

Throughout the literature, the lack of consensus regarding whether poverty drives teen births or vice versa most likely signals mutual causal effects. There are obvious links between early childbearing and lower social and economic achievement. Nord et al. (1992) find that teenage mothers tend to acquire a lower level of education, earn lower wages, have lower career goals, and enjoy their careers less. Lowenthal (1997) suggests that most teenage mothers do not marry and continue to live in poverty as well.

Furstenberg (1991) also recognizes a pattern between marital status and teen mothers. The author finds that the total number of births to unmarried teenagers increased by more than 50 percent from 1960 to 1975, even though the total number of births to teens declined. The number of births to unmarried teens continued to increase through the late 1980s. Sandfort and Hill (1996) report that by 1993, 72 percent of all teen births were to single mothers.

Unemployment rates and average income levels in a community have also been shown to be related to rates of teen motherhood. The average income within a county is shown to be a predictor for teenage motherhood (Young et al. 2004). Areas with a larger percentage of families with lower incomes are found to have higher numbers of teen mothers. Furthermore, communities with higher levels of unemployment are also found to have higher levels of teen mothers.

Tantivorawong (2009) suggests that there is a relationship between the number of hours worked and teenage motherhood. Studies show that an increase in the number of hours worked tends to increase the probability of teenage motherhood. The author suggests this finding could reflect female teens working more, thus having more financial means to engage in risky behaviors. However, the model does not control for community characteristics. Moreover, Misra (2009) shows that the type of workers is also associated with college or university retention rates; hence, there is also a link between various workers and students' success in a community. Whether the magnitude and ratios of part-time and full-time workers influences teen motherhood rates in a community is something that has not yet been explored and may be worthwhile to investigate.

Brown and Lichter (2004) assess whether residing in a metropolitan area, versus residing in a nonmetropolitan area, affects the economic status of teen mothers. The authors find that mothers who reside in nonmetropolitan regions are financially worse off than those who live in metropolitan areas. A possible explanation serves to support SmithBattle's (2007) claim that the opportunities and tools available to a teen mother affect her outcome - as a metropolitan area generally has more opportunities for employment and education along with public transportation, greater choices for housing and daycare, etc., compared to nonmetropolitan areas. Previous authors also state that community characteristics such as race, educational level, and income can influence poverty in a community as well as increase the number of teen mothers (Maynard, 1996; Nord et al., 1992). Similarly, Young et al. (2004) show that low educational expectations, low locus-of-control, and low family socioeconomic status are all factors that may contribute to teenage motherhood.

Research on the effects of an area's industrial structure on the probability of teen motherhood is rather scarce, perhaps even non-existent. Geronimus and Korenman (1992) find that urban areas tend to have larger proportions of teen mothers. Urban areas tend to have more professional and management oriented jobs, whereas rural areas tend to offer agricultural related jobs. Beyond the type of occupation, job stability may also play a role in predicting teen motherhood rates. A study by Diebold, Neumark and Polsky (1994) finds that job stability within the U.S. fluctuates with job type. The authors find that between white-collar and service jobs, white-collar jobs tend to have more stability, measured by retention rates. During their study from 1983 to 1991, retention rates were observed for good-producing sectors and service-producing sectors in four-year increments. Retention rates among good-producing jobs increased while retention rates among service-producing jobs decreased. Furthermore, retention rates were found to be highest in professional and technical occupations, followed by clerical jobs and then by workers in the service industries. Thus it is reasonable to assume that socio-economic status is related to job security or stability within a region.

Teen motherhood is a very complex social issue that can be viewed from a multiplex of directions. Research continues in an attempt to untangle the relationship of mutual causality between poverty and teen motherhood; and studies continue to seek the most efficient and practical implementation of policies targeted at reducing the social costs imposed by teen births.

Why ZIP Code Level?

Our primary contribution with this work is the assessment of teen motherhood rates at the ZIP code level--so why use the ZIP code? Counties, while stable over time, containing their own political infrastructure, and offering observable borders, can have quite extensive internal diversification. County-level measures for larger counties, with both rural and urban areas, can easily hide underlying heterogeneity. Thus, our contribution is to probe beyond the county, to a smaller geographic unit. We stop short of census tracts and blocks. While these more finely defined geographic areas offer more precise measures, their intangible nature leave them somewhat un-relatable to the individuals who occupy them, as opposed to the familiarity one has regarding his/her own county or ZIP code. In addition, the census tract being so small would arguably suffer more from spatial autocorrelation than a ZIP code. Thus we conduct our analysis at the ZIP code level; we also provide county level analyses for comparison purposes.

The term ZIP code refers to Zone Improvement Plan, a streamlining mail-delivery system designed and employed by the U.S. Postal Service in the early 1960s. The various ZIP codes throughout the U.S. represent postal carrying areas and routes, and their boundaries are determined, and subject to change, by the needs of the postal service. ZIP codes are not bound by traditional geographies such as counties; occasionally, they can even cross state lines. Our employment of a ZIP code level analysis is not unique in and of itself, as researchers often look at them as neighborhood or community proxies, rich with economic data. Bingham and Zhang (2001) profile neighborhoods as distinctive, contiguous communities that contain, "(1) labor market characteristics of the residents, and (2) the social, economic, demographic, and physical characteristics of the neighborhood," (pg.3). Bingham and Zhang equate Ohio ZIP codes (or sets of ZIP codes) to central-city neighborhoods, based on the layout of the state, its urban areas, and its unique and identifiable communities that were established before the institution of ZIP codes--such an equality between neighborhood and ZIP code(s) may or may not be transferable to other states. However, many of the community attributes described here have been examined and exploited at the ZIP code level across a variety of economic fields of study.

With respect to the socioeconomic and demographic data, in a national study of how community characteristics influence environmental outcomes, Arora and Cason (1999), argue that, for relatively broad studies, ZIP codes are more practical and straightforward than alternative community measures. Neidell (2004) employs a ZIP code level analysis in his study of the effects of air pollution of child asthma rates in California--noting favorably the ZIP code's variation in SES measures and well defined geographic specificity.

With respect to the traditional economic data even macroeconomic data--Mian and Sufi (2009) exploit financial data at the ZIP code level in a nationally representative study, examining the relationship between income growth, mortgage credit, and subprime mortgages leading in to the financial crisis of 2008. Bondonio and Greenbaum (2007) credit the ZIP code as the smallest geographic unit for which business level data is made available, employing ZIP codes as their unit of analysis in their study of the effects of tax incentives on economic growth. The results of Rosenthal and Strange (2001), however, who find less evidence of agglomeration at the ZIP code level for U.S. manufacturing industries than at the county or state level, suggest studies should be cautious of (but not necessarily avoid) employing ZIP code analyses of business data.

Outside the discipline of economics, though related to some economic fields of study, public health studies are seeing more ZIP code studies recently. According to Krieger et al. (2002), over 20 percent of articles in the National Library of Medicine (PubMed) dating from 1989 to 2002 employed ZIP code analysis. Generally these studies examine whether socioeconomic variation predicts various health outcomes. The field debates whether ZIP codes or census tracts (or blocks) are preferable for the smaller neighborhood unit. Issues that arise in the debate are the increased heterogeneity within the often larger (in population) ZIP code versus the smaller, more homogenous census tracts and blocks. Additionally, ZIP codes are more subject to spatiotemporal instability than census tracts and blocks. Estimates suggest that three to five percent of ZIP codes change annually (Clapp et al., 1997). Census tracts--and, by default, census blocks--are designed to be more permanent, left unchanged for ten-year periods. From 1997 to 2001, approximately 390 new zip codes were introduced in the U.S., and 120 were terminated (Krieger et al., 2002). Research that yields unique findings for ZIP code analysis versus alternative measures, such as census tract or block, should be cautions of whether the ZIP code studies offer a more complete community measure or are plagued by moving boarders and shifting populations. An additional concern regarding ZIP code heterogeneity is that because they are designed to facilitate and expedite mail delivery, in extreme cases, a ZIP code could represent a rural county or a large business firm. Thus, when comparing outcomes by ZIP codes, researchers should examine the underlying populations.

An additional and unique quality that has contributed to the use of ZIP code driven analysis in health related studies includes the relationship between health care providers and beneficiary addresses (Coburn et al., 2007). Research that focuses on the outcomes and unique struggles faced by residents of rural areas, and the policies those studies impact, sometimes use the number of health care providers in specific ZIP codes to identify the availability of medical resources for specific populations. ZIP codes generally offer a geographic unit that is not so small as to eclipse potential resources a few blocks or miles away, but are also not so large as to assume all the resources in a given area would be equally available to everyone throughout that geographic location.

While the employment of ZIP code level analysis does introduce new concerns, our employment avoids many of those raised by researchers. ZIP code level analysis is often used to measure demographic and socioeconomic factors (Arora and Cason, 1999; Bingham and Zhang, 2001; Corcoran et al., 1992; and Neidell, 2004). We include labor force participation data to exploit the breadth and depth of the ZIP code unit regarding interwoven and complex community relationships (Bondonio and Greenbaum, 2007; Mian and Sufi, 2010; and Rosenthal and Strange, 2001). And our cross-sectional data, rather than longitudinal or time series data, avoids the potential biases from ZIP code border changes that effectively reallocate populations and introduce new (as well as eliminate) some ZIP codes. Nevertheless, our ZIP code assessment is still subject to the arbitrary lines of distinction that define ZIP code border lines and spatial autocorrelation.

III. Data and Methodology

The data analyzed in this study are obtained from the Michigan Department of Community Health (MDCH). We include data from mothers between the ages of 12-18 who gave birth in 2000 in this paper, which are obtained from the Live Birth files for the state of Michigan. The time frame is chosen because the data necessary for this study is not available for other years.

In order to evaluate what factors affect the total number of births to teenage mothers in each ZIP code, a number of explanatory variables are used in the regression model (see appendix for a complete list and set of definitions). Income (log income is used in the analysis), location of residence, marital status, educational level, amount of employment, type of employment and race are all analyzed. These variables are collected from the Census SF3 2000 survey. Many of the variables are transformed into proportions in order to control for variances in total population across ZIP codes. The data reported for teen mothers is categorized by the residing ZIP code of the mother. Each ZIP code in Michigan is listed, ranging from ZIP code 48001 to ZIP code 49971. In total there are 967 ZIP codes observed.

We provide spatial figures to illustrate the distribution of teens across Michigan's ZIP codes (Figure 1), as well as the distribution of Michigan's poor (Figure 2). These figures are suggestive, though not conclusive, of a spatial relationship between teens and the poor population. That is to say, a ZIP code with a higher number of teens appears to have a higher poverty rate.

[FIGURE 1 OMITTED]

Table 1 presents the summary statistics for the sample. A typical ZIP code in the state has a 1.0 percent teen pregnancy rate (on average, the number of births to teen mothers in each ZIP code is approximately seven). Across a typical Michigan ZIP code, 23.3 percent of the population (15 and older) has never been married; 9.9 percent is divorced. With respect to labor force participation, 3.9 percent of the labor force (defined here as the population 16 and older) report being unemployed. Of those 16 and older in any given ZIP code, 51.8 percent work at least 35 hours per week on average. Another 12.8 percent work between 15 and 35 hours per week on average, while only 3.2 percent work between one and 14 hours per week on average. Per capita income is, on average, $20,400; although log income is used in the analysis, per capita income in dollars is provided in the descriptive statistics. The average poverty rate is 9.6 percent. On average, 73.6 percent of the total population (three years and older) in a given ZIP code is not enrolled in school, while 3.9 percent are enrolled in their undergraduate years of college.

The remaining summary statistics presented in Table 1 further depict a typical Michigan ZIP code. We control for additional employment industry variables, abbreviated as "AG" (agriculture related), "TR" (transportation related), "FI" (finance related), "MNGT" (management related), and "ARTS" (please see appendix for a list of the specific industries that fall within these categories).

[FIGURE 2 OMITTED]

Geographic variables included in the explanatory variables include a rural dummy variable. On average 63.4 percent of Michigan ZIP codes are identified as rural. ZIP codes are further identified and categorized as "north," "south," "east," or "west," based on the location of the county the ZIP code is centered in. On average, 20 percent of Michigan ZIP codes are located in the north part of the state, 51 percent are located in the south part of the state, 13.9 percent are located in the eastern portion of the state, and the remaining 14.8 percent of the ZIP codes are located in the west.

We also control for race and ethnicity within each ZIP code. In a typical Michigan ZIP code, the majority of residents are white. On average, 5.2 percent are black, 1.2 percent are Native American, 0.8 percent are Asian, and 2.6 percent are classified as other. For comparison purposes, Table 1 also presents summary statistics at the county level. A typical county looks similar to a typical ZIP code, with some notable exceptions pertaining to region and industry. Such differences are generally attributable to population density.

IV. Results

Table 2 presents the results from the cross-sectional analysis; Model 1 presents the estimates for the ZIP code level analysis. We estimate the model by ordinary least squares; we also employ White's heteroskedasticity-consistent covariance estimation method. The model is based on 942 observations, after some are dropped due to missing or incomplete information. The proportion of teens that are mothers is the dependent variable, and the measure is calculated at the ZIP code level.

Among the variables found to be positively related to rates of teen motherhood at the ZIP code level are: the proportion of the community divorced; the proportion of the community separated; the proportion of the labor force working, on average, 15-35 hours per week; the proportion of the labor force working in finance related industries; and the proportion of the population living in poverty. The only variables found to be negatively related to teen motherhood rates at the ZIP code level are the proportion of the labor force working in professional, management-related fields and per capita income, at the one and ten percent significance levels, respectively.

With respect to the relationship statuses among the population, the positive coefficients on the fractions of the population divorced and separated are consistent with the expectations of the authors. In particular, a one percentage point increase in the fraction of the population that is divorced is associated with, on average, a 0.039 percentage point increase in the teen motherhood rate, ceteris paribus. (By extension, a 10 percentage point increase in the divorced population is associated with a 0.39 percentage point increase in the teen motherhood rate.) The effect of the separated population--statistically different from that of the divorce rate--suggests that a one percentage point increase in the proportion separated is associated with, on average, an increase in the teen motherhood rate of 0.124 percent. (By extension, a 10 percentage point increase in the separated population is associated with a 1.24 percentage point increase in the teen motherhood rate.)

While the magnitude of these effects appears small, the mean value of the dependent variable is 0.01, or one percent; thus coefficients with relatively small magnitudes need not imply trivial relationships. To provide context, we also estimate these effects at the mean. As such, the suggested effect of a 10 percentage point increase in the divorce rate (a 100 percent increase) would be to increase the average teen motherhood rate from 1.0 percent to 1.4 percent (which represents a 40 percent increase). Because the mean value of the proportion of the population (15 and older) that is separated is 0.012, or 1.2 percent, we consider the effects of doubling the percent separated to have comparable proportion-based effects from the two relationship statuses. The suggested effect of a 1.2 percentage point increase in the proportion separated would be to increase the average teen motherhood rate by 20 percent. Thus, the findings suggest that the teen motherhood rate (at the ZIP code level) is substantially more sensitive to changes in the underlying divorce rate of the population than it is to changes in the separated population. Note this outcome holds even though the coefficient for percent separated is approximately three times that of percent divorced.

We find insufficient evidence to suggest a relationship between the proportions of the community that is never married and widowed and the teen motherhood rate. Additionally, none of the relationship status variables included in the model, all of which represent alternatives to married status, is found to be negatively related to the teen motherhood rate.

Labor force participation variables that are found to contribute to teen motherhood include the proportion of the labor force working an average of 15-35 hours per week and the proportion of the labor force working in finance related industries. The magnitude of the effects, as measured by the coefficients, is comparable, though slightly I larger, to that of the divorced population. When estimated at the mean, the suggested effect of I doubling the percentage of the labor force that I works 15-35 hours per-week (an increase from 12.8 percent, on average, to 25.6 percent) would be to increase the average teen motherhood rate 60 percent. Similarly, the suggested effect of I doubling the proportion of the labor force working in finance related industries (an increase from 4.4 percent, on average, to 8.8 percent) would be to increase the average teen motherhood rate 20 percent. Moving the teen motherhood rate in the opposite direction is the proportion of the labor force working in management-related fields. A one percentage point increase in the fraction of the labor force working in management related fields is associated with an average decrease in the teen motherhood rate of 0.069 percent, all else constant. (By extension, a 10 percentage point increase in management employment is associated with a 0.69 percentage point decrease in the teen motherhood rate.) When estimated at the mean, the suggested effect of doubling the percentage of the labor force that works in finance related industries (an increase from 6 percent, on average, to 12 percent) would be to decrease the average teen motherhood rate by 50 percent, ceteris paribus.

The coefficient for the percentage of the population living below the poverty line is 0.047-significant at the one percent level. The implication is that a one percentage point increase in the poverty rate is associated with an average increase in the teen motherhood rate of 0.047 percent, ceteris paribus. (By extension, a 10 percentage point increase in the poverty rate is associated with a 0.47 percentage point increase in the teen motherhood rate.) When estimated at the mean, the suggested effect of a 10 percentage point increase in the poverty rate (nearly a 100 percent increase) would be to increase the average teen motherhood rate from 1.0 percent to 1.5 percent (which represents a 50 percent increase).

The coefficient on the proportion of the labor force that is unemployed is insignificant, a rather surprising finding. This may be due to the correlation between this measure and the poverty rate and per capita income measures at the ZIP code level. With respect to racial minorities, all but percent other population have insignificant coefficients. Additionally, the proportions of the population (three years of age and older) enrolled in college and not enrolled in school are not statistically significant. The model itself has limited explanatory power: the R-squared indicates that only 22 percent of the variation in the ZIP code level teen motherhood rate is explained by the model.

Finally, while rural status has an insignificant coefficient in Model 1, those of south and west are positive and statistically significant. On average, ZIP codes in the south part of the state have a teen motherhood rate of 0.3 percent higher than ZIP codes in the east portion of the state, ceteris paribus. ZIP codes in the west, as opposed to those in the east, are associated with teen motherhood rates that are, on average, 0.4 percent higher.

We also run the model using county level data to compare the results with those of the ZIP code analysis (see Model 2 in Table 2). We find that proportion separated, proportion of people that work 15-35 hours per week, and proportion of other population are significant factors both at the county and ZIP code levels. Differences in coefficients across the ZIP code and county level analyses signal a classic modifiable areal unit problem, MAUP (Openshaw and Taylor, 1979). While such differences are worthy of deeper exploration, such an exercise is beyond the scope of this paper.

Robustness

As a robustness check of the model, we also run an OLS model estimating the poverty rate for the ZIP code; the results are presented in Table 3. The robustness model, Model 3, is identical to the main analysis with respect to independent variables, with one exception: the robustness model replaces proportion of the population below the poverty line with proportion of teen mothers. Model 3's R-squared value indicates that 73 percent of the variation in ZIP code level poverty rates is explained by the model.

For comparison purposes, Table 3 also provides estimates for Model 4, which estimates the poverty rate using county level data. The proportion unemployed; proportion of population that works 35 or more hours per week; per capita income; proportions black, Native American, Asian, and other; and proportion of teens that are mothers all have significant coefficients at the county level as well as at ZIP code level. Therefore, the determinants of poverty remain robust in both of these models. There are, however, a number of additional coefficients that are only significant in the ZIP code analysis--those regarding relational status of the population and college and/or school enrollment measures. Overall, it is fair to conclude that regardless of spatial specifications--ZIP code level versus county level--the number of teen mothers is positively associated with poverty in the state of Michigan.

Table 4 summarizes the findings from Models 1 and 3. The divorced proportion of the population is found to be positively related to both the teen motherhood rate and the poverty rate within a given ZIP code. Per capita income is found to be negatively related to both the teen motherhood rate and the poverty rate. And interestingly, the labor force participation coefficients that are statistically significant across models 1 and 3 change sign per given model specification. The proportion of the labor force working on average 15-35 hours per week is found to be positively related to the teen motherhood rate while negatively related to the poverty rate. And the coefficient on percentage employed in management industries is negative in Model 1 and positive in Model 3.

Additionally, rural status, proportion never married, proportion (of the labor force) unemployed, and proportions black and Asian, all have positive and statistically significant coefficients in Model 3. These explanatory variables have insignificant coefficients in Model 1. Similarly, the proportions (of the population three and older) enrolled in college and not enrolled in school and the proportion (of the labor force) working full time (35+ hours per week, on average) are negatively related to the ZIP code poverty rate. However, these measures all have insignificant coefficients in Model 1. Finally, in Model 3, the coefficient on proportion of teen mothers is positive and significant, as expected. The interpretation is that a one percentage point increase in the proportion of mothers that are teenagers is associated with an average increase in the ZIP code poverty rate of 0.303 percentage points. (By extension, a 10 percentage point increase in the teen motherhood rate is associated with a 3.03 percentage point increase in the ZIP code poverty rate.) When estimated at the mean, the suggested effect of a one percentage point increase in the teen motherhood rate (a 100 percent increase) would essentially have no measurable effect on the average ZIP code level poverty rate.

The results of Models 1 and 3 reinforce one another with respect to the observed positive relationship between teen motherhood and the poverty rate at the ZIP code level. The results, however, add little more evidence to the cause-and-effect debate (Nord et al., 1992 and Lowenthal, 1997). As the primary goal of this paper is to analyze determinants of teen mothers, not to assess factors of poverty, the authors acknowledge that the probable endogeneity issue between these two variables of interest is beyond the scope of this paper.

V. Discussion

With respect to the main findings of the analysis as presented in Table 2, the positive relationships between rates of divorce and separation in the community and the teen motherhood rate are of particular interest. The model controls for the ZIP code's poverty rate; thus the implication is that higher rates of such instances of broken homes is positively associated with the teen motherhood rate--holding constant the poverty rate (among other control variables).

Higher rates of broken homes may be positively associated with teen motherhood due to the emotional and financial tribulations such families are prone to suffer, creating instability within these households. Such instability within a family's structure may have an effect on teenage upbringing including attitude and core values. Higher rates of broken homes may also affect teenagers psychologically and emotionally--skewing their view of marriage and even eroding their sense of family and social values (Corak 2001). As a consequence, such teenagers tend to become teen mothers more often than teenagers who have grown up in stable families (Young et al, 2004). The effects of broken homes are found to influence teenage marriage rates as well. Though teenage marriage rates have been declining over the years, a significant number of teenagers continue to get married. Frequently marriages so early in life end relatively quickly, and these teens become young parents (Bahr and Galligan, 1984).

Additionally, a potential inference of this broken home effect, as seen in Table 2, is rooted in the ZIP code level analysis. There is the possibility that at the ZIP code level, our findings reflect sorting into specific communities (i.e. low-rent apartments with disproportionate occupancy rates among divorced or separated households). There could be an affordability component, or other community characteristic, that attracts financially strapped households such as divorced and separated households as well as households with teen mothers.

With respect to the effects of various labor force participation measures presented in Table 2, the positive coefficient of finance-related industries and the negative coefficient of management related industries also warrant some discussion. Individuals employed in finance related industries (i.e.: finance, insurance, real estate) may be more likely to experience volatility in the characteristics defining their employment, due to the nature of these jobs. For example, a real estate agent, insurance broker, or leasing agent's salary is often highly dependent upon that individual person's performance. Thus, pay differentials among such workers (and over time) may be substantial. Such instability in earnings could arguably contribute to family instabilities. Alternatively, individuals employed in management related industries (i.e.: professional, scientific, management, administrative) may be less likely, on average, to experience this type of volatility within their occupations (Diebold et al., 1994). Diebold et al. argue that professional-type jobs, such as those in the management related industries, tend to have higher job security compared with performance based jobs in the service sector, such as those in the finance related industries. Many jobs within management industries are more skill based than performance based, relying heavily in many cases on higher levels of education and skills. The more stable nature of these positions could arguably lead to a more stable environment for families.

Possible sorting into neighborhoods based on employer location and commute could also influence these industry coefficients. As seen from the summary statistics in Table 1, ZIP codes generally have higher proportions of their populations employed in specific sectors--as opposed to county level averages. Based on the discussion presented above, stronger concentrations of industries may simply reinforce previous findings by researchers. The concentrations of industry in ZIP codes may also imply closer and more complex relationships among residents, compared to residents of the same county. This line of thought, however, does not immediately shed light on the relationship between industry type and teen motherhood rates in a community. Whether the findings of the analysis support industry influence on family stability or can be reduced to an anomaly, further research into these relationship is certainly warranted.

The positive effect of part-time employment, specifically workers working an average of 15-35 hours per week, on the teen motherhood rate is also of interest. This type of employment may be consistent with teenagers and young adults working more hours (having, on average, less parental supervision and more disposable income than their counterparts who work fewer hours per week) and ultimately choosing to engage in risky behavior (Tantivorawong, 2009). Perhaps, though, a more likely interpretation is that the higher rates of such part-time employment are consistent with teenagers and young adults working with time constraints imposed upon them from school and child-care. Though this interpretation introduces endogeneity--as the elevated part-time employment would be a consequence of higher rates of teen mothers--it appears more plausible to the authors. Additional study will need to be undertaken to examine this relationship further.

With respect to the statistically significant and positive coefficient on the poverty rate variable, our findings are consistent with the literature (Lowenthal, 1997; Maynard, 1996; Moore et ah, 2002; Nord et ah, 1992; Rich-Edwards, 2002; and Young et ah, 2004). Recall, when estimated at the mean, the results suggest that, at the ZIP code level, doubling the poverty rate would, on average, be associated with a 50 percent increase in the teen motherhood rate, ceteris paribus. When one considers the overall variation in poverty rates across the ZIP codes in the sample, the standard deviation is 0.002, or 0.2 percent, while the average rate is 0.096, or 9.6 percent. The standard deviation, at approximately 2 percent of the mean value of the variable, implies limited variation in the poverty rates across the ZIP codes. Further, if we estimate at the mean the effects of such a typical change, we find that the suggested effects of a 2 percent change in the poverty level (from 9.6 percent to 9.79 percent) essentially has no measurable effect on the teen motherhood rate at the ZIP code level, ceteris paribus. Other variables with statistically significant coefficients, as presented in Table 2, also have limited variances--approximately 2 percent of the mean value of the variable at most. Thus the observed relationships discussed above hold in theory, but comparisons made across any two typical ZIP codes, with limited variation in our control measures, would likely present relationships that appear much smaller in scale.

The main contribution of this study is the spatial aspect that focuses on ZIP code specific rates and factors of teen motherhood. The ZIP code measure allows us to more finely measure teen motherhood rates in a given area, yet at the same time, it introduces arbitrary lines of distinction and separation in the sample: placing individuals in ZIP code A versus ZIP code B, etc. ZIP codes, unlike counties, do not have their own governing entities, school districts, etc. While ZIP codes could arguably act like a neighborhood (or group of neighborhoods), the larger picture of a typical community with the provision of public goods, etc. is eclipsed at the ZIP code level--rather it holds more at the county level. Thus when ZIP codes are profiled, the gain of the revelation of the underlying group's homogeneity is offset by the loss of the more comprehensive picture of how this group interacts in society. Such issues help to explain the limited power of Model 1, which recall, the R-squared is only 0.22. While the model's explanatory power is limited, the suggested relationships are worth further investigation. Further research will help identify which effects may be overstated here due to the arbitrary lines of distinction and sorting behavior across ZIP codes that are not controlled for other spatial autocorrelation and will investigate the merits and limitations of the findings presented here.

VI. Summary and Conclusions

This paper addresses spatial issues with empirical methodologies to better understand the factors related to teen mother rates, a topic which has been relevant in academic research for decades. In our empirical analysis we find factors associated with the number of teenage mothers in Michigan to include the proportions of the population that are separated and divorced, work status, industry type, the community's poverty rate, and geographic location.

In order for government officials and policymakers to be successful in reducing teen motherhood and its costs on society, there should be a focus of attention and resources in areas where the problem is most prominent. Based on these results, ZIP codes in the south or west regions of the state, with high percentages of separated or divorced households; with high percentages of part-time employment (15-35 hours per week) and employment within finance related industries; and with high percentages of people whose income is below the poverty line may be helpful in identifying ZIP code level communities to be given ample consideration when funds are allocated.

Previous studies (for example: Kearney and Levine, 2011) concerning teen motherhood rates in the U.S. have explored spatial differences, yet have not studied this issue at the ZIP code level. This unit of analysis offers a means of more finely identifying areas in need as well as providing additional information regarding potentially contributory factors of teen motherhood. It is likely that ZIP code level analyses will supplement rather than replace county level studies that address the same issue.

Over the years, teen parenting has inundated the government with increased obligations, especially financially. Research in this area continues to try to identify where this social issue needs more extensive attention. This ZIP code level analysis, improving on the precision of teen-motherhood rate, seeks to contribute to this vast literature about the issue by revealing the value of using smaller geographic units of measure, as well as assessing the contributing factors given the smaller geographic scope. The goal of this research is to continue pushing this conversation forward and to assist in helping to equip policy makers as they face complicated decisions regarding directing assistance where the social costs of teen pregnancy are the greatest.

Appendices

Definitions of Variables

Proportion of teen mothers: Proportion of female teens that are mothers.

Rural: I = if reside in rural area; 0 = if do not reside in rural area.

Proportion never married: Proportion of population 15 years and over who have never been married.

Proportion separated: Proportion of population 15 years and over who are separated.

Proportion widowed: Proportion of population 15 years and over who are widowed.

Proportion divorced: Proportion of population 15 years and over who are divorced.

Proportion enrolled in college: Proportion of population 3 years and over who are enrolled in their undergraduate years of college.

Proportion not enrolled in school: Proportion of population 3 years and over who are not currently enrolled in school.

Proportion unemployed: Proportion of population 16 years and over who are not currently employed.

Proportion that works 35 hours per week or more:

Proportion of population 16 years and over who worked 35 hours or more per week on average in 1999.

Proportion that works 15-35 hours per week:

Proportion of population 16 years and over who worked 15-35 hours per week on average in 1999.

Proportion that works 1-14 hours per week:

Proportion of population 16 years and over who worked 1-14 hours per week on average in 1999.

Proportion employed in "AG" industries: Proportion of population 16 years and over employed in agriculture, forestry, fishing and hunting, or mining.

Proportion employed in "TR" industries: Proportion of population 16 years and over employed in transportation and warehousing, or utilities.

Proportion employed in "FI" industries: Proportion of population 16 years and over employed in finance, insurance, real estate and rental or leasing.

Proportion employed in "MNGT" industries: Proportion of population 16 years and over employed in professional, scientific, management, administrative, or waste management services.

Proportion employed in "ARTS" industries: Proportion of population 16 years and over employed in arts, entertainment, recreation, accommodation or food services.

Per capita income: Average income per person in 1999.

Proportion whose income is below poverty line: Proportion of population whose income is considered to be below the poverty level in 1999.

North: Dummy variable for north ZIP codes, as arbitrarily determined, which includes all of those within Keweenaw, Houghton, Ontonagon, Gogebic, Baraga, Iron, Marquette, Dickinson, Menominee, Delta, Alger, Schoolcraft, Luce, Chippewa, Mackinac, Emmet, Cheboygan, Presque Isle, Leelanau, Charlevoix, Antrim, Otsego, Montmorency and Alpena counties.

South: Dummy variable for south ZIP codes, as arbitrarily determined, which includes all of those within Montcalm, Gratiot, Saginaw, Genesee, Oakland, Macomb, Wayne, Washtenaw, Livingston, Shiawassee, Clinton, Ionia, Allegan, Barry, Eaton, Ingham, Jackson, Calhoun, Kalamazoo, Van Buren, Berrien, Cass, St. Joseph, Branch, Hillsdale, Lenawee and Monroe counties.

East: Dummy variable for east ZIP codes, as arbitrarily determined, which includes all of those within Crawford, Oscoda, Alcona, Roscommon, Ogemaw, Iosco, Gladwin, Arenac, Midland, Bay, Tuscola, Huron, Sanilac, Lapeer and St. Clair counties.

West: Dummy variable for west ZIP codes, as arbitrarily determined, which includes all of those within Kent, Ottawa, Muskegon, Newaygo, Mecosta, Isabella, Clare, Osceola, Lake, Oceana, Mason, Manistee, Wexford, Missaukee, Kalkaska, Grand Traverse and Benzie counties.

Proportion of black population: Proportion of population who is of black ethnicity.

Proportion of Native American population: Proportion of population who is of Native American/ Indian ethnicity.

Proportion of Asian population: Proportion of population who is of Asian ethnicity.

Proportion of other population: Proportion of population who is of non-white or other ethnicity.

References

Arora, S. and T.N. Cason. 1999. "Do Community Characteristics Influence Environmental Outcomes? Evidence from the Toxics Release Inventory." Southern Economic Journal 65(4): 691-716.

Bahr, Stephen J. and Richard J. Galligan. 1984. "Teenage Marriage and Marital Stability." Youth and Society 15(4): 387-100.

Bingham, R.D. and Z. Zhang. 2001. The Economies of Central-city Neighborhoods. Boulder, CO: Wesview Press.

Bondonio, D. and R. T. Greenbaum. 2007. "Do Local Tax Incentives Affect Economic Growth? What Mean Impacts Miss in the Analysis of Enterprise Zone Policies." Regional Science & Urban Economic 31: 121-136.

Brown, Brian J. and Daniel T. Lichter. (2004). "Poverty, Welfare, and the Livelihood Strategies of Nonmetropolitan Single Mothers." Rural Sociology 69(2): 282-301.

Clapp, J. M., M. Rodriguez, and G. Thrall. 1997. "How GIS Can Put Urban Economic Analysis on the Map." Journal of Housing Economics (6): 368-386.

Coburn, A. F" A. C. MacKinney, T. D. McBride, K. J. Mueller, R. T. Slifkin, and M. K. Wakefield. 2007. "Choosing Rural Definitions: Implications for Health Policy." Columbia, MO: Rural Policy Research Institute Health Panel. Retrieved June 3, 2013. (http://www.rupri.org/Forms/ RuralDefinitionsBrief.pdf)

Corak, Miles. (2001). "Death and Divorce: The Long-Term Consequences of Parental Loss on Adolescents." Journal of Labor Economics 19(3): 682-715.

Corcoran, M., R. Gordon, D. Laren, and G. Solon. 1992. "The Association Between Men's Economic Status and Their Family and Community Origins." The Journal of Human Resources 27(4): 575-601.

Diebold, Francis X., David Neumark, and Daniel Polsky. 1994. "Job Stability in the United States." Cambridge, MA: National Bureau of Economic Research. Accessed April 13, 2012 (http://www .nber.org/papcrs/w4859.pdf?ncw_window-1).

Furstenberg, Frank F. Jr. 1991. "As the Pendulum Swings: Teenage Childbearing and Social Concern." Family Relations 40(2): 127-138.

Geronimus, Arline T. and Sanders Korenman. 1992. "The Socioeconomic Consequences of Teen Childbearing Reconsidered." The Quarterly Journal of Economics 107(4): 1187-1214.

Hoffman, S. (2006). By the Numbers: The Public Costs of Teen Childbearing. Washington, DC: National Campaign to Prevent Teen Pregnancy.

Kearney, M.S. and P. B. Levine. 2011. "Income Inequality and Early Non-Marital Childbearing: An Economic Exploration of the 'Culture of Despair.'" Working Paper No. wl7157, National Bureau of Economic Research, Cambridge, MA.

Kearney, M.S. and P. B. Levine. 2012a. "Explaining Recent Trends in the U.S. Teen Birth Rate." Working Paper No. 17964, National Bureau of Economic Research, Cambridge, MA.

Kearney, M.S. and P. B. Levine. 2012b."Why Is the Teen Birth Rate in the United States So High and Why Does It Matter?" Journal of Economic Perspectives 26: 141-63.

Klepinger, Daniel H., Shelly Lundberg, and Robert D. Plotnick. 1995. "Adolescent Fertility and the Educational Attainment of Young Women." Family Planning Perspectives 27(1): 23-27.

Krieger, N" P. Waterman, J. T. Chen, M. Soobader, S. V. Subramanian, R. Carson. 2002. "Zip Code Caveat: Bias Due to Spatiotemporal Mismatches between ZIP Codes and US Census-Defined Geographi Areas-The Public Health Disparities Geocoding Project." American Journal of Public Health: Research and Practice, 92(7): 1101-1102.

Lowenthal, Barbara and Richard Lowenthal. 1997. "Teenage Parenting: Challenges, Interventions, and Programs." Childhood Education 74(1): 29-32.

Maynard, Rebecca A. 1996." Kids Having Kids: A Robin Hood Foundation Special Report on the Costs of Adolescent Childbearing." New York: The Robin Hood Foundation: 1-2.

McKay, David O. 1957. Pathways to Happiness Salt Lake City: Bookcraft.

Mian, A. and A. Sufi. 2009. "The consequences of mortgage credit expansion: Evidence from the U.S. mortgage default crisis." The Quarterly Journal of Economics 124(4): 1449-1496.

--. 2010. "The Great Recession: Lessons from Microeconomic Data." American Economic Review 100: 51-56.

Misra, Kaustav. 2009. "The Impact of Part-time Jobs Availability on College and University Retention Rates: A State Level Analysis." International Journal of Economic Perspectives 3(4): 249-256.

Moore, K.A., Glei, D.A., Driscoll, A.K., Zaslow, M.J., Redd, Z. (2002). "Poverty and Welfare Patterns: Implications for Children." The Journal of Social Policy 31(2): 207-227.

National Campaign to Prevent Teen and Unplanned Pregnancy, Fast Facts 2011a. Retrieved September 17, 2012 (http://www.thenationalcampaign .org/resources/pdf/FastFacts_3 in 10.pdf).

__. Counting It Up: The Public Costs of Teen Childbearing: Key Data. 2011b. Retrieved September 17, 2012 (http://www.thenationalcampaign .org/costs/pdf/counting-it-up/key-data.pdf).

--. Counting It Up: The Public Costs of Teen Childbearing in Michigan in 2008. 2011c. Retrieved September 17, 2012 (http://www.the nationalcampaign.org/costs/pdf/counting-it-up/factsheet-michigan.pdf).

Neidell, M. J. 2004. "Air Pollution, Health, and Socio-economic Status: The Effect of Outdoor Air Quality on Childhood Asthma." Journal of Health Economics 23: 1209-1236.

Nord, Christine W., Kristen A. Moore, Donna R. Morrison, Brett Brown, and David E. Myers. 1992. "Consequences of Teen-Age Parenting." Journal of School Health 62(7): 310-318.

Openshaw, S. and P. J. Taylor. 1979. "A Million or so Correlation Coefficients: Three Experiments on the Modifiable Areal Unit Problem." Pp. 127-144 in Statistical Applications in Spatial Sciences, edited by N. Wrigley. London: Pion Ltd.

Proudfoot, Shannon. 2008. "Teen Moms Combat Negative Images With Education; Early Pregnancies Do Not Equal Poverty." Edmonton Journal: A6.

Rich-Edwards, Janet. 2002. "Teen Pregnancy is Not a Public Health Crisis in the United States. It is Time We Made it One." International Journal of Epidemiology 31(3): 555-556.

Rosenthal, S. S. and W. C. Strange. 2001. "The Determinants of Agglomeration." Journal of Urban Economics, 50: 191-229.

Sandfort, Jodi R., and Martha S. Hill. May 1996. "Assisting Young, Unmarried Mothers to Become Self-Sufficient." Journal of Marriage and Family 58(2): 311-326.

SmithBattle, Lee. 2007. "Legacies of Advantage and Disadvantage: The Case of Teen Mothers." Public Health Nursing 24(5): 409-420.

Tantivorawong, Athip. 2009. "An Econometric Analysis on Factors that Affect Teenage Pregnancy Based on 1979 National Longitudinal Survey of Youth." Senior Honors Thesis, Tepper School of Business, Carnegie Mellon University. Accessed January 6, 2012 from http://tepper.cmu.edu/undergraduate-economics/ academics/curricula/tepper-senior-honors-program in-economics/list-of-epper-senior-honors-theses in-economics/index.aspx.

Young, Tamera, Jean Turner, George Denny, Michael Young, 2004. "Examining External and Internal Poverty as Antecedents of Teen Pregnancy." American Journal of Health Behavior 28(4):4: 361-373.

This research is supported by a grant from the Allen Foundation 2011, under contract #12142200, for which the authors are indebted.

by Kaustav Misra, Corresponding author: Department of Economics, College of Business and Management, Saginaw Valley State University, University Center, MI 48710. E-mail: [email protected]. Phone: (989)964-2651; Fax: (989)964-7497.

Kylie Goggins, Department of Economics, College of Business and Management, Saginaw Valley State University, University Center, MI 48710. E-mail: [email protected]. Phone: (989)964-4340; Fax: (989)964-7497.

Amber Matte, Undergraduate Student, College of Business and Management, Saginaw Valley State University, University Center, MI 48710. E-mail: [email protected]. Phone: (989)964-8617; Fax: (989)964-7497.

Averetta E. Lewis, 216 Health Sciences, Department of Nursing, College of Health and Human Services, Saginaw Valley State University, University Center, MI 48710. Phone: (989)964-7146; Fax: (989)964-4925.

TABLE 1.
Summary Statistics: Determinants of Teen Mothers

Variables Mean Std. Dev.

 ZIP Code Level

Dependent Variable:

 Proportion of Teen Mothers 0.010 0.000
 (Total number of Teen
 mothers/Total Female Teens)

Explanatory Variables:

 Rural 0.634 0.016
 Proportion Never Married 0.233 0.003
 Proportion Divorced 0.099 0.001
 Proportion Widowed 0.068 0.001
 Proportion Separated 0.012 0.000
 Proportion Enrolled in College 0.039 0.002
 Proportion Not Enrolled in School 0.736 0.002
 Proportion Unemployed 0.039 0.001
 Proportion that Works 35+ Hours/Week 0.518 0.003
 Proportion that Works 15-35 Hours/Week 0.128 0.001
 Proportion that Works 1-14 Hours/Week 0.032 0.000
 Proportion Employed in "AG" Industries 0.029 0.001
 Proportion Employed in "TR" Industries 0.043 0.001
 Proportion Employed in "FI" Industries 0.044 0.001
 Proportion Employed in "MNGT" Industries 0.060 0.001
 Proportion Employed in "ARTS" Industries 0.081 0.002
 Per Capita Income 20,403.93 228.08
 North 0.2 0.013
 South 0.51 0.016
 West 0.148 0.011
 East 0.139 0.011
 Proportion of Black Population 0.052 0.005
 Proportion of Native American Population 0.012 0.001
 Proportion of Asian Population 0.008 0.001
 Proportion of Other Population 0.026 0.001
 Proportion Whose Income is 0.096 0.002
 Below Poverty Line
 N 942

Variables Mean Std. Dev.

 County Level

Dependent Variable:

 Proportion of Teen Mothers 0.009 0.004
 (Total number of Teen
 mothers/Total Female Teens)

Explanatory Variables:

 Rural 0.698 0.462
 Proportion Never Married 0.232 0.051
 Proportion Divorced 0.100 0.012
 Proportion Widowed 0.071 0.015
 Proportion Separated 0.012 0.003
 Proportion Enrolled in College 0.052 0.046
 Proportion Not Enrolled in School 0.887 0.050
 Proportion Unemployed 0.040 0.011
 Proportion that Works 35+ Hours/Week 0.501 0.057
 Proportion that Works 15-35 Hours/Week 0.133 0.022
 Proportion that Works 1-14 Hours/Week 0.033 0.005
 Proportion Employed in "AG" Industries 0.016 0.010
 Proportion Employed in "TR" Industries 0.015 0.010
 Proportion Employed in "FI" Industries 0.024 0.008
 Proportion Employed in "MNGT" Industries 0.029 0.014
 Proportion Employed in "ARTS" Industries 0.049 0.015
 Per Capita Income 18,857.66 3,065.77
 North 0.506 0.503
 South 0.240 0.430
 West 0.132 0.341
 East 0.120 0.327
 Proportion of Black Population 0.035 0.062
 Proportion of Native American Population 0.014 0.025
 Proportion of Asian Population 0.006 0.009
 Proportion of Other Population 0.026 0.014
 Proportion Whose Income is 0.102 0.031
 Below Poverty Line
 N 83

TABLE 2.
Results for Models 1 and 2: Determinants of Teen Mothers

Variables Coefficient Std. Error

Dependent Variable:

 Proportion of Teen Mothers
 (Zip Code Level)

Explanatory Variables:

 Rural -0.001 0.001
 Proportion Never Married 0.003 0.013
 Proportion Divorced 0.039 ** 0.019
 Proportion Widowed -0.015 0.020
 Proportion Separated 0.124 * 0.064
 Proportion Not Enrolled in School 0.015 0.017
 Proportion Enrolled in College -0.023 0.019
 Proportion Unemployed -0.010 0.025
 Proportion that Works 35+ HoursAVeek -0.002 0.011
 Proportion that Works 15-35 HoursAVeek 0.052 ** 0.022
 Proportion that Works 1-14 HoursAVeek -0.053 0.042
 Proportion Employed in "AG" Industries -0.015 0.016
 Proportion Employed in "FI" Industries 0.050 ** 0.020
 Proportion Employed in "MNGT" Industries -0.069 *** 0.019
 Proportion Employed in "ARTS" Industries 0.009 0.014
 Log Per Capita Income -0.006 * 0.003
 North -0.001 0.002
 South 0.003 ** 0.002
 West 0.004 ** 0.002
 Proportion of Black Population 0.006 0.005
 Proportion of Native American Population 0.003 0.020
 Proportion of Asian Population 0.038 0.031
 Proportion of Other Population 0.039 ** 0.017
 Proportion Whose Income is 0.047 *** 0.013
 Below Poverty Line
 Constant/Interccpt 0.041 0.031
 R-Squared 0.22
 Adjusted R-Squared 0.20
 N 942

Variables Coefficient Std. Error

Dependent Variable:

 Proportion of Teen Mothers
 (County Level)

Explanatory Variables:

 Rural -0.001 0.001
 Proportion Never Married -0.098 *** 0.025
 Proportion Divorced 0.035 0.044
 Proportion Widowed 0.065 0.043
 Proportion Separated 0.474 ** 0.232
 Proportion Not Enrolled in School -0.119 ** 0.056
 Proportion Enrolled in College -0.060 0.050
 Proportion Unemployed -0.132 *** 0.045
 Proportion that Works 35+ HoursAVeek 0.021 0.016
 Proportion that Works 15-35 HoursAVeek 0.067 * 0.038
 Proportion that Works 1-14 HoursAVeek -0.175 * 0.096
 Proportion Employed in "AG" Industries -0.033 0.045
 Proportion Employed in "FI" Industries -0.130 * 0.077
 Proportion Employed in "MNGT" Industries 0.024 0.092
 Proportion Employed in "ARTS" Industries -0.006 0.043
 Log Per Capita Income -0.006 0.009
 North 0.000 0.002
 South 0.000 0.001
 West 0.001 0.002
 Proportion of Black Population 0.026 * 0.014
 Proportion of Native American Population 0.015 0.023
 Proportion of Asian Population -0.036 0.089
 Proportion of Other Population 0.133 *** 0.038
 Proportion Whose Income is 0.041 0.030
 Below Poverty Line
 Constant/Interccpt 0.173 0.112
 R-Squared 0.77
 Adjusted R-Squared 0.68
 N 83

Please note: *; **; ***; represent 10%; 5%; and 1% levels
of significance, respectively.

Reference categories include: Proportion Married; Proportion
in "TR" Industries; East; Proportion White.

TABLE 3.
Results for Models 3 and 4: Determinants of the Poverty Rate

Variables Coefficient Std.
 Error

Dependent Variable: Proportion Whose
 Income is Below
 Poverty Line
 (Zip Code Level)
Explanatory Variables:
 Rural 0.006 * 0.003
 Proportion Never Married 0.198 *** 0.033
 Proportion Divorced 0.339 *** 0.047
 Proportion Widowed 0.028 0.051
 Proportion Separated 0.171 0.163
 Proportion Enrolled in College -0.172 *** 0.047
 Proportion Not Enrolled in School -0.146 *** 0.043
 Proportion Unemployed 0.396 *** 0.061
 Proportion that Works 35+ I lours/Week -0.276 *** 0.028
 Proportion that Works 15-35 Hours/Week -0.225 *** 0.056
 Proportion that Works 1-14 Hours/Week -0.136 0.106
 Proportion Employed in "AG" Industries 0.060 0.042
 Proportion Employed in "FI" Industries -0.053 0.050
 Proportion Employed in "MNGT" Industries 0.115 ** 0.048
 Proportion Employed in "ARTS" Industries 0.019 0.035
 Per Capita Income -0.062 *** 0.008
 North -0.006 0.004
 South -0.003 0.004
 West -0.001 0.004
 Proportion of Black Population 0.060 * 0.013
 Proportion of Native American Population -0.015 *** 0.052
 Proportion of Asian Population 0.224 *** 0.078
 Proportion of Other Population 0.210 *** 0.044
 Proportion of Teen Mothers 0.303 *** 0.083
 Constant 0.880 *** 0.074
 R-Squared 0.73
 Adjusted R-Squared 0.72
 N 942

Variables Coefficient Std.
 Error

Dependent Variable: Proportion Whose
 Income is Below
 Poverty Line
 (County Level)
Explanatory Variables:
 Rural 0.002 0.004
 Proportion Never Married -0.056 0.093
 Proportion Divorced 0.119 0.139
 Proportion Widowed -0.202 0.137
 Proportion Separated -0.453 0.711
 Proportion Enrolled in College 0.205 0.127
 Proportion Not Enrolled in School 0.047 0.159
 Proportion Unemployed 0.760 *** 0.131
 Proportion that Works 35+ I lours/Week -0.239 *** 0.040
 Proportion that Works 15-35 Hours/Week 0.093 0.107
 Proportion that Works 1-14 Hours/Week -0.290 0.303
 Proportion Employed in "AG" Industries 0.124 0.151
 Proportion Employed in "FI" Industries -0.384 * 0.205
 Proportion Employed in "MNGT" Industries 0.475 0.285
 Proportion Employed in "ARTS" Industries 0.021 0.137
 Per Capita Income -0.116 *** 0.027
 North -0.001 0.004
 South 0.000 0.004
 West -0.005 0.005
 Proportion of Black Population 0.137 *** 0.037
 Proportion of Native American Population -0.198 *** 0.070
 Proportion of Asian Population 0.564 ** 0.219
 Proportion of Other Population 0.211 * 0.123
 Proportion of Teen Mothers 1 091 *** 0.376
 Constant 1.269 *** 0.323
 R-Squared 0.69
 Adjusted R-Squared 0.68
 N 83

Please note: *; **; ***; represent 10%; 5%; and 1% levels of
significance, respectively.

Reference categories include: Proportion Married; Proportion in
"TR" Industries; East; Proportion White.

TABLE 4.
Comparisons of ZIP Code Analysis Results, Models 1 and 3

Model 1 3
 Y = Teen Y =
 Mother Rate Poverty
 Rate

Independent Variables
 Proportion Whose Income is +++ N/A
 Below Poverty Line
 Proportion of Teen Mothers N/A +++

 Rural +
 Proportion Never Married +++
 Proportion Divorced ++ +++
 Proportion Widowed
 Proportion Separated +
 Proportion Enrolled in College ---
 Proportion Not Enrolled in School ---
 Proportion Unemployed +++
 Proportion that Works 35+ Hours/Week ---
 Proportion that Works 15-35 Hours/Week ++ ---
 Proportion that Works 1-14 Hours/Week
 Proportion Employed in "AG" Industries
 Proportion Employed in "FI" Industries ++
 Proportion Employed in "MNGT" Industries --- ++
 Proportion Employed in "ARTS" Industries
 Per Capita Income - ---
 North
 South ++
 West ++
 Proportion of Black Population +
 Proportion of Native American Population ---
 Proportion of Asian Population +++
 Proportion of Other Population ++ +++
R-Squared 0.22 0.73

Plus and minus signs (1, 2 or 3) denote sign and significance at the
10, 5 and 1 percent level, respectively.