ࡱ> }|@ ˆjbjbqq %l    * 8$&" ( 4PRRRRRR, ~<<<~    <P 6<P  $74 4{ {  Latino Dropouts and High Stakes Testing: A Report Funded by the Carlos H. Cantu Hispanic Education and Opportunity and Endowment Alisa Hicklin The Project for Equity, Representation, and Governance Department of Political Science Texas A&M University 4348 TAMUS College Station, TX 77840 ahicklin@politics.tamu.edu Latino Dropouts and High Stakes Testing Current political rhetoric concerning the U.S. educational system largely centers around one idea: accountability. Over the last twenty years, states have worked to design accountability systems that will set standards, assess performance, and levy sanctions on underperforming school districts, mostly as a response to A Nation at Risk, a federal action calling for the improvement of American public schools. The complex, ever-changing combinations of standardized testing, monetary awards, administrative sanctions, and funding schemes have fueled a frenzy of scholarly studies on the effects on these policies. A large part of this literature is devoted to how these accountability systems disproportionately affect minority groups. Specifically, some studies argue that these systems, and their emphasis on high-stakes testing, have become especially detrimental to the Latino student population. Most of the scholarly literature focusing on the link between Latino dropouts and high-stakes testing consists of either case studies (limiting the number of students in the analysis) or quantitative studies that have found the effect of testing on the dropout rate to be much less than is expected. In the case of these quantitative studies, many of the scholars have argued that the unexpectedly weak statistical relationship can be attributed to the indirect effects of testing on other issues, such as grade retention, over-enforcement of disciplinary standards, limited English proficiency exemptions, and academic tracking all of which have been found to lead to increased dropout rates. Previous Work on Testing and Dropouts The most common objection to high-stakes testing lies in the argument that testing negatively affects minority students, more so than their Anglo counterparts (Fletcher 2002, Clarke, Haney, and Maddus 2000, Velez and Saenz 2001, McNeil 2000, Lomax, West, Harmon, Viator, and Madaus 1995, Orfield and Wald 2001). These studies find that high stakes testing is often found in states where minority education is already lacking, and the addition of such tests widens the achievement gap. Overall, research has found that the presence of high stakes tests increases students probability of dropping out (Reardon and Galindo 2002, Clarke et al. 2000). Fletcher (2002) shows that African-American and Hispanic students fail graduation tests approximately twice as often as Anglo students. Reardon and Galindo (2002) follow this study, confirming that minority students do suffer from the added hurdle of high stakes testing (as does Jacob 2000). Clarke et al. (2000) evaluate testing and find that its most significant effect on minority dropout rates is actually filtering through the effect of grade retention. These scholars show that high stakes testing causes increased grade retention and, in turn, increased grade retention leads to higher dropout rates. Velez and Saenz (2001) evaluate the dropout trends of Latino students and present research suggesting that testing-based policies can put schools in a position to be more inclined to over-enforce disciplinary measures, so as to push out at-risk students in an attempt to increase overall tests scores. This over-enforcement is often seen when students are suspended for minor disturbances (chewing gum, talking in class, etc.), and especially when these suspensions (not coincidentally) begin the day before the first testing day and end just after the test is completed (also see Amrein 2002). Orfield and Wald (2002) conclude that high stakes tests attached to high school graduation lead to increased drop-out rates, particularly for poor and minority students. This perceived bias against minority groups has led to public protests by leading minority political groups (NAACP, MALDEF, ACLU), arguing that high stakes testing leads to increased dropouts and a new discrimination (Guillen 2001). The discrimination issue seems to be most prevalent with regards to Latino students, due in part to language barriers. When state-mandated tests require mastery of English language, Latino students may be discouraged before even taking the tests. With so much emphasis placed on the tests, Latino students often feel that they have no chance of success in the school system (Valenzuela 2000). Although these studies offered multiple conclusions about the effect of testing on dropout rates, the many of these scholars, along with many school administrators and journalists, point to the state accountability systems as the driving force behind the problems with high-stakes testing. The primary argument is that the states emphasis on high-stakes testing, (versus giving weight to multiple assessment measures), forces school districts to focus only on raising test scores. To raise the districts test scores, administrators may choose to narrow the curriculum (teach to the test), artificially inflate the scores (cheating) (Bohte and Meier 2000), or try to reduce the number of students taking the test, leaving only those students who have the best chance of passing. This final strategy is the focal point of this study. The central argument is that school district administrators, specifically superintendents, are very aware of the possibility of state sanctions that could threaten their budgets, their power, and their job security. If superintendents believe that their current student body is incapable of making adequate progress toward successful passing rates on the test, they might find ways to manipulate the districts scores. One possible way to artificially inflate these scores is to identify the students who are most likely to fail and find ways to keep those students from taking the test. This strategy is often seen with the graduation exams, usually administered in grade 10. This argument is linked back to the Latino student population through the classification of limited English proficiency. In short, administrators identify their Latino students as high risk due to limited knowledge of English. To keep these students from taking (and presumably failing) the test, administrators could take many different actions. Students can be retained in 9th grade, can be suspended for the week of test on over-inflated disciplinary charges, can be exempted (in some states) for limited English, can be tracked into special education program (exempting them from the test), or can simply be encouraged to drop out of school. Superintendents who are interested in inflating their districts pass rates, essentially, would have two choices: increase the number of students who are capable of passing the test or remove a significant number of students who are likely to fail. A superintendent who is incapable or unwilling to increase pass rates by better educating their students may choose to encourage (or not discourage) at risk students to drop out of school. Although the state technically factors in the dropout rate when grading districts on their performance, evidence shows that the dropout rates used in the states analysis are seriously flawed (Texas 2000). This paper tests the underlying assumption of the above argument, that higher Latino dropout rates will increase the Latino student pass rate on the high stakes exam. Because most students drop out between 8th and 12th grade, this paper will examine the effects of increased dropout rates on the pass rate for the 10th grade high stakes exam. The central hypothesis for this analysis will be the assumption that administrators make if they choose to push out their at-risk Latino students Increased Latino drop out rates will lead to higher Latino pass rates for the district. Theoretically, the logic is somewhat sound. If lower achieving students are more likely to drop out of school, than encouraging them to drop out will remove a significant group of lower achieving students from the test pool. If removing lower achieving students from the test pool will lead to better district pass rates, then encouraging some at-risk students to drop out should increasing the district pass rate. Additionally, it is possible that superintendents would be pushing out Latino students not to raise the Latino pass rate, but to raise the aggregate district pass rate. A second model is run, therefore, using the district pass rate for all students as the dependent variable. If superintendent cheating is effective, we would expect a positive relationship between the dropout rate and test scores. Data and Methods To test this assumption, this analysis uses data collected from the Texas Education Agency measuring student achievement on the primary high stakes exam from 1999-2001. For each district, a production function of student achievement is used to predict the districts Latino pass rate for the 10th grade exam. Because the 10th grade exam is a high-stakes test (must be passed to graduate), this is exam is highly salient and is administered at a time when many students may be contemplating dropping out. The production function predicts the 10th grade pass rates using the districts 7th grade Latino TAAS pass rate for the same cohort of students, along with multiple controls, and the districts Latino dropout rate. So, for students taking the 10th grade exam in 2000, the districts 7th grade pass rate from 1997 will be used as an independent variable. Data Units of Analysis The units of analysis for this study are Texas school districts. Each school district is an independent unit, led by a superintendent and a school board with principals at each campus. As with most public organizations, these districts are assessed using multiple performance indicators. In Texas, the primary performance indicator for a school district is the districts pass rate on the Texas Assessment of Academic Skills, hereafter referred to as the TAAS test. School districts are given different labels depending on their performance on these tests, along with multiple other factors. Dependent Variable The districts 10th grade Latino TAAS pass rate is used as the dependent variable. Because this is the most salient measure of a districts performance, it is often one of the top priorities for district administrators (Meier and OToole 2000). The pass rates are taken from the Texas Education Agencys (TEA) website, through the Academic Excellence Indicator System (AEIS). The data is pooled for all districts for four cohorts of students (those taking the 10th grade exam from 1999 to 2002). Many school districts have an a very small number of Latino students, and thus do not have data on Latino pass rates, reducing the total number of data points to 1227. These 10th grade district pass rates vary from 0 to 100 with a mean of 72.62. Independent Variables 7th Grade Latino Pass Rate. Because each group of students is unique in their composition and abilities, some measure of past performance is taken into consideration. In this analysis, the districts Latino pass rate on the 7th grade exam is used in the function for 10th grade pass rates, with both rates drawn from the same cohort of students. Therefore, the 7th grade pass rate from 1996 is used to predict the 10th pass rate in 1999. These 7th grade pass rates, taken from TEA, were not available for all districts, serving as a primary limitation in trying to include all Texas districts. These 7th grade pass rates range from 0 to 100 with a mean of 69.05. Including this variable should also provide some insight into the TAAS testing instrument. Theoretically, if the TAAS test is a valid, consistent measure of achievement, then we would expect that there is a consistent predictable level of achievement, or in this case, the pass rates should be relatively consistent. Additionally, if the same cohort of students takes a valid, consistent assessment of achievement in one year, and then takes the same type of test four years later, those students should pass this test at a predictable rate. Statistically speaking, including the 7th grade pass rate should yield a large, statistically significant coefficient with a large r-squared. Latino Dropout Rate. The main variable of interest in the districts Latino dropout rate for that year. If a school districts administrators believe that reducing the number of at-risk students would be profitable for the districts pass rate, a strong, positive relationship is expected. The dropout rate is taken from the Texas Education Agency but is not the primarily rate that is most often reported. Instead, this analysis uses the data in the system that track cohorts of students and reports what percentage of students are considered dropouts, continuing students, graduating students, and GED students, with the percentages adding to 100%. This is considered to be a better measure of dropouts, as it is a cohort measure. Latino Retention Rate. To control for the theory argued by Clarke et al (2000), that administrators will retain schools in the 9th grade to improve TAAS scores, the percentage of students categorized by the TEA as "continuing students" is included. These data represent the percentage of students who have completed four years of high school but do not have the credits to graduate. If the theory is correct, the coefficient for this variable should be positive and significant. Quality of the District. Unarguably, some districts do better than others for a multitude of reasons. In an effort to capture the school districts overall performance ability, this analysis considers multiple resource variables and some constraint variables. Five variables are used to capture district quality. The percentage of the districts funds coming from the state and the average teacher salary are used to control for a districts wealth. The percentage of low income students, a common indicator of educational task difficulty (Meier and OToole 2000), is used to control for how challenging it is to educate students. So as to consider the benefits drawn from more experienced educators, average teacher experience (in years) is also included. Factors Specific to the Latino Population. Because language barriers are often cited as a primary cause of Latino academic problems, districts with a large group of students considered to be of limited English proficiency (LEP) have the added responsibility of teaching basic language skills, prerequisite to the skills tested on the TAAS. Some districts combat this problem with bilingual education problems. To account for both forces, a variable measuring the percent of students considered to be LEP will be including along with the percentage of the districts teaching hours that are dedicated to bilingual education. TAAS success. Many studies argue that some districts have learned to teach to the test better than other districts (Olsen 2000), meaning that some districts have found ways to systematically teach skills needed to score better on the test, without actually teaching the concepts that are supposed to be tested. Controlling for a districts ability to beat the system is difficult. However, using the pass rate for Anglo students in the same cohort should control for these factors. This variable is expected to be positively related to the Latino pass rate, given that the same educational structure is in place. Methods These data are examined using a least squares analysis. So as to remove any variation that can be attributed only to yearly fluctuations, the model is run as a one-way fixed effects model. The data represent Texas school districts for years 1999-2001. Because pass rates form the basis of the analysis, districts with very small Latino population are not included. The sample includes only those districts with at least 10% Latino students to ensure that the data provide meaningful information. This is especially important because we are using pass rates. If the Latino student population makes up a very small portion of the student body, the pass rates are not very reliable. The 10% mark allows for more meaningful interpretations of the data analysis. Findings Many of the findings in this analysis are somewhat surprising. Specifically, the relationship between 7th grade TAAS scores and 10th grade TAAS scores is most unexpected. Given that both variables in that relationship measure performance on the same type of test, administered by the same governmental group, and taken by the same cohort of students, 7th grade scores should be an effective predictor of 10th grade success. Instead, there is a consistent lack of both statistical and substantive significance for the both the 7th grade TAAS Latino pass rate and the 7th district pass rate. We also find striking differences between the determinants of Latino success versus the determinants of district success. Effects of Increased Latino Dropouts on Latino Performance In table one, two of most consistent predictors of student achievement, state aid and percentage of low income students (Meier and OToole 2000), seem to have little effect on student performance in this study. Quality of teaching (average years of experience), along with resources specific to the Latino population (percentage of teaching dedicated to bilingual education), were not statistically significance either. The variables that did reach statistical significance in this analysis are average teacher salary, the percentage of students who are Latino, the percentage of student who are considered to be of limited English proficiency (LEP), the 10th grade Anglo pass rate, and the Latino dropout rate. Average teacher salary, a good measure of district resources, is both significant and in the predicted, positive direction. The variable used to control for the districts ability to teach to the test, the Anglo pass rate, is both positive, significant, and substantively influential. Most importantly, both the dropout rate and the retention rate are negative and significant. This clearly shows that if administrators are targeting certain Latinos to weed them out, they are choosing the wrong students. Increased dropouts and increased retention rates are both negatively related to Latino student performance. For Latino student success, administrative cheating does not work. [Table One About Here] The two measures of student composition are significant but in opposite directions. This is understandable, given what each variable measures. The percentage of Latino students in a district is positively related to the pass rate, lending to the explanation that those districts with higher Latino populations have the most background in equipping Latino students with the skills needed to pass the TAAS exam. Conversely, the percentage of students considered to be LEP has a strong, negative relationship with the TAAS pass rate, which is very expected given the history of literature arguing that language barriers are the primary problem for Latino students. Effects of Increased Latino Dropouts on District Performance In Table 2, a similar model is run to predict the determinants of TAAS passage for the entire district. As was found in the previous model, neither the 7th grade all pass rate, nor the 7th grade Latino pass rate has any significant effect on 10th grade scores. However, unlike table 1, many of the other variables do achieve statistically significance. The percentage of funds from the state, the percent of students considered to be low income, and average teacher experience all achieve statistically significance with coefficients in the predicted direction. Also, the percentage of students considered LEP is no longer statistically or substantively significant. [Table 2 About Here] The independent variables of interest, the Latino dropout rate and the Latino retention, remain statistically significant, with substantively significant coefficients in the negative direction. Once again, increased Latino dropouts lead to poorer performance, even when trying to control for district quality (through the 10th grade Anglo pass rate). Increased Latino retention leads to poorer overall district performance. For the districts in this analysis, targeting a group of students to remove them from the test taking pool does not have a positive impact on district performance. It not only hurts Latino student performance (Table 1), but it also negatively impacts the entire district (Table 2). Interpretations Three findings in this analysis warrant further explanation: the 7th grade Latino scores, the relatively low r-squared, and the Latino dropout rate. First, why is not the district 7th grade Latino pass rate for the same cohort of students a significant predictor of 10th grade district success? Three possible explanations exist the districts in the analysis had enough growth in the Latino population to significantly alter the cohort, the data are just wrong, or the TAAS tests are internally flawed. There are some missing data, but there is no real pattern in which districts or years are missing data points. The mean of the 7th grade scores is approximately 3 points lower than the 10th grade mean, which does not seem unreasonable. The growth possibility also remains unlikely. In the end, it is most likely that the lack of a relationship is due to the TAAS itself. It is possible that the test is simply a poor measure of achievement, a theory well-supported in the education literature, or that the 7th and 10th grade exams test substantively different skills, which is not supposed to be the case. One other possibility is that because the 10th grade exam is high-stakes -- with failure preventing high school graduation -- the tests are viewed differently. The 10th grade exam may be taken more seriously than the 7th grade exam, by both the students and teachers, and therefore losing the consistency between years. Another interesting finding is found in the differences between tables 1 and 2. Not only does the explanatory power of the model increase, but the variables measuring traditional district resources and constraints become significant in table 2. Apparently, the determinants of Latino student success are not the "usual suspects" used to predict Anglo student success. Instead, there must be other factors that play a significant role in achievement, specific to Latino students. The phenomenon is found in the low R-squared of .23 in table 1. When considering what factors would determine Latino student achievement, the model in this paper includes many pertinent issues: district quality, past achievement, resources, needs, Latino-specific variables, and TAAS-specific variables. These are the variables that have been causally linked to student performance in the literature, yet they, as a group, only explain one-fourth of the variance. Although the inevitable issues of model misspecification and some measurement error exist, there would have to be a major, critical flaw in the research design to explain these results. Again, this raises questions about the TAAS tests validity. It is possible that these variables do not explain much of the variance in the test pass rates, because the tests lack a systematic element. Even knowing this, though, it is important to recognize the importance of using the TAAS as the primary indicator of achievement. Although this may be a poor assessment of academic skills, it is still defined as the most important and most salient indicator of district performance in Texas. The pass rates are publicized in the schools, the community, on the nightly news, and in the papers. Students are rewarded with TAAS parties or days off of school for good performance. Most importantly, the state awards (both symbolically and financially) are undeniably grounded in district TAAS performance. If this is how the state and the district administrators measure performance, the factors that influence these pass rates are important for education, public policymaking, and political representation. This brings us back to the issues of district administrators and their so-called cheating through increasing Latino dropout rates. The data clearly leads to the conclusion that Latino dropout rates are statistically and substantively significant in the negative direction. The higher the dropout rate, the worse the districts Latinos do on the TAAS. Of course, this negates the original hypothesis, the rationale behind district cheating. Again, multiple explanations for this relationship exist. A strong possibility is that the multiple variables included to control for the quality of the district were not sufficient enough, and that the statistical results are only showing that higher dropout rates are indicative of a poor district, which would be most likely to have poor pass rates on the TAAS. The other possibility, although highly unlikely, is that administrators do target certain students, but are unable to identify the students who would be mostly likely to fail. Once again, this could be because the TAAS does not truly measure academic skills. Regardless, if the administration is pushing out students (or even if they are openly not encouraging Latino students to stay in school), they could be pushing out the wrong students. Additionally, this type of administrative strategy is likely to be seen by the students as a hostile environment, leading to the further determinant of the Latino student pass rate on the TAAS. Regardless of the specific explanation, one thing is clear: administrative cheating, either through dropouts or retention, if it exists, does not work. According to this analysis, districts who wish to raise Latino student performance on the TAAS should look to decrease the number of students who are labeled LEP, possibly by better bilingual education, and districts should consider redirecting funds to teacher salaries in an effort to recruit better teachers. Although these findings have been echoed in much of the previous education literature, this is the first study to empirically test the effects of district cheating. As mentioned multiple times, statistical results consistently lead to questions about the TAAS tests validity and reliability. If the TAAS is truly an inferior measure of ability and quality, the practical implications are these: the quick fix of pushing out some students does not help in the short term, nor does it improve anything in the long run. Despite the shortcomings of the TAAS, these results fall directly in line with the overarching conclusions of previous research: improving Latino education is a necessary task that requires long term planning. As the Latino population grows, our society will become increasingly dependent on the socioeconomic development of our Latino communities. Education is the critical first step in this process and deserves immediate attention from policymakers, district administrators, and academicians. Table 1: Determinants of Latino TAAS Success VariableCoefficientP-Value7th Grade Pass Rate, Latino0.0350.223District Latino Dropout Rate-0.2160.000District Latino Continuing (Retention) Rate-0.3450.000Percentage of Funds from State0.0001.000Average Teacher Salary0.00060.034Percentage of Students Considered Low Income0.0050.893Average Teacher Experience-0.3530.221Percentage of Students Considered LEP-0.3310.002Percent of Teacher Hours Dedicated to Bilingual0.2480.14410th Grade Pass Rate, Anglo0.3720.000Dependent Variable: 10th District Latino TAAS Pass Rate R-squared 0.2333 N 753 Root MSE 12.16 Table 2: Determinants of District TAAS Success VariableCoefficientP-Value7th Grade Pass Rate0.0190.6417th Grade Pass Rate, Latino0.0060.847District Latino Dropout Rate-0.2570.000District Latino Continuing (Retention) Rate-0.5910.000Percentage of Funds from State0.0350.050Average Teacher Salary0.0010.000Percent of Students Consider Low Income-0.2070.000Average Teacher Experience0.8570.000Percent of Students Consider LEP0.0950.136Percent of Teacher Hours Dedicated to Bilingual-0.1670.098Dependent Variable: 10th District TAAS Pass Rate R-squared 0.5796 N 904 Root MSE 10.078 References Amrein, Audrey L. and David C. Berliner. 2002. High-Stakes Testing, Uncertainty, and Student Learning. Education Policy Analysis Archives. 10(18). Bohte, John and Kenneth J. Meier. 2000. "Goal Displacement: Assessing the Motivation for Organizational Cheating." Public Administration Review. 60(2). Chudowsky, Naomi, Nancy Kober, Keith S. Gayler, and Madlene Hamilton. 2002. State High School Exit Exams: A Baseline Report. Center on Education Policy. Washington, D.C. Clarke, Marguerite, Walter Haney, and George Madaus. 2000. High Stakes Testing and High School Completion. National Board on Educational Testing and Public Policy. 1(3). Curtis, Emory. 1999. Graduation Exit Testing Fails. Exodus News. October 1999. Dropout Study: A Report to the 77th Texas Legislature. 2000. Legislative Budget Board. December 2000. Fletcher, Michael A. 2002. Exit Tests Hurt At-Risk Students. Washington Post. August 14, 2002. Guillen, Liz. 2001. Californias Latino Students: Continuing Challenges That Cannot Be Ignored. Testimony Presented to the Senate Education Committee. MALDEF. January 2001. Heubert, Jay P. and Robert M. Hauser. 1999. High Stakes: Testing for Tracking, Promotion, and Graduation. Washington, D.C.: National Academy Press. Heubert, Jay. 2002. High-Stakes Testing: Opportunities and Risks for Student of Color, English-Language Learners, and Students with Disabilities. In The Continuing Challenge: Moving the Youth Agenda Forward. Ed. M. Pines. Baltimore: Johns Hopkins University Press. Jacob, Brian A. 2000. Getting Tough?: The Impact of High School Graduation Exams. Harris Graduate School of Public Policy Studies. University of Chicago. Lomas, Richard G., Mary Maxwell West, Maryellen C. Harmon, Katherine A. Viator, and George F. Madaus. 1995. The Impact of Mandated Standardized Testing on Minority Students. Journal of Negro Education. 64(2): 171-185. Mathews, Jay. 2001. State Tests Are Becoming a Graduation Hurdle. Washington Post. January 30, 2001. McNeil, Linda. 2000. Contradictions of School Reform: Educational Costs of Standardized Testing. New York: Routledge. Olson, Lynn. 2000. High-Stakes Tests Jeopardizing Hispanics, Panel Warns. Education Week. 19(42):7. Orfield, Gary and Johanna Wald. 2001. High Stakes Testing. In Motion Magazine. April 2001. Reardon, Sean and Claudia Galindo. 2002. Do High-Stakes Tests Affect Students Decisions to Drop Out of School? Evidence from NELS. Presented at the Annual Meeting of the American Educational Research Association. New Orleans, LA. Romo, Harriett and Toni Falbo. 1996. Keeping Latino Youth in School. Hogg Foundation for Mental Health. University of Texas. Sanni, Christine. 2000. Studies Find High Stakes Tests Threaten Disadvantaged Students in Texas. HGSE News. Harvard Graduate School of Education. January 2000. Schrag, Peter. 2000. Too Good to Be True. The American Prospect. 11(4). Students Challenge Discriminatory Impact of MCAs Testing. 2001. American Civil Liberties Union: Massachusetts. www.aclu-ass.org/youth/studentrights/studchallenmcas.html. August 2001. Texas Education Agency. 2000. Dropout Study: A Report to the 77th Texas Legislature. December 2000. http://www.tea.state.tx.us/research/dropout/rider71study/ Velez, William and Rogelio Saenz. 2001. Toward a Comprehensive Model of the School Leaving Process Among Latinos. School Psychology Quarterly. 16: 445-467. Viadero, Debra. 2000. Testing System in Texas Yet to Get Final Grade. Education Week. 19(38): 20-21. Warren, John Robert and Melanie R. Edwards. 2001. The Impact of High Stakes Graduation Tests on High School Dropout. Presented at the Fourth Annual Undergraduate Research Symposium. University of Washington.  The Texas Education Agency changed the high stakes testing instrument as of 2002-2003. The current test is the Texas Assessment of Knowledge and Skills. All of the data in this analysis is taken from years in which the TAAS was administered. PAGE  PAGE 22 9V_ @A""9$;$B$D$$$V%%((**9*;*++@+B+++ ,,W,i,>.?.....00x1z11111112222C3E3y3{333446466%787::::;<>?]AiAFFCGEG^G`G=H>*H*>* j0JUH*6565CJ 5Z     9:;<=>V$a$     9:;<=>V )D_ @c i#P'(((R,W,i,..111t4$7:;>\ACCCFFFIIIMaObOyORRPRTTUWWb )D_ $a$$da$dd`$a$@c i#P'(((R,W,i,..111t4$7:;>\ACCCd`d$a$CFFFIIIMaObOyORRPRTTUWWWWWZ]l_aek$da$dd`=H?HsHuHHHIIIIIIxLzLRPRRR S SESGSJVLVWW#X%XXXXXXX`ZbZZZ[[[[p\r\\\!]#]M_h_ff+g0gjili-l5l6lqqqqqqq r!rZr[rrrrrrrssBsCssssssssGtdtetktltmt5H*5CJCJ5656H*\WWWWZ]l_aekvqqqqqqqqqqqqqqqqqrr r!rMrTrZr[rzrrrrrrrrrrrrssss5sywyyyyyyBzTzz{I{{{Q|||#}S} ~;~~~~ހ=bށNcσHcuЄ\ j0JU CJOJQJ6565H*5CJH*Xtttttttttt uuuuxxxxxx$$&P#$/Ifa$ $&P#$/Ifj$$Ifl,F\!''T' 6P"6    4 l4a uuuu:u@uFuGu^udujukuuuuuuuuuuuuu'v.v4v5vgvhvivjv{vvvvvvv=www*xxx>yyzSzezz{{n{{|a||}k}} ~N~~~ 6VCwɁm‚-yЃ u˄1Æņdž _u:u@uFuGu^udujukuuuuuuxxxj$$Ifl,F\!''T' 6P"6    4 l4a$$&P#$/Ifa$ $&P#$/If uuuuuuuu'v.v4v5vgvhvxxvv $&P#$/Ifj$$Ifl,F\!''T' 6P"6    4 l4a$$&P#$/Ifa$ hvivjv{vvvvvvv=www*xxx>yyzSzezz{ $d`a$ $d^a$$da$$^a$$`a$$a$$a${{n{{|a||}k}} ~N~~~ 6VC $d^a$$^a$$`a$$a$$da$ $d`a$CwɁm‚-yЃ u˄1&`#$$^a$$a$d^d $d^a$$da$ $d`a$†ÆņƆdžˆ0JmH0J j0JU džȆɆʆˆd&`#$džȆʆˆ # 01h/ =!"#$% i(@( NormalCJmH <@< Heading 1$$d@&a$56@6 Heading 2$d@&5<A@< Default Paragraph Font2@2  Footnote TextCJ8&@8 Footnote ReferenceH*(>@( Title$a$5(U@!( Hyperlink>*B*, @2, Footer  !&)@A& Page Number>(ˀ      ˀ z z z z z z z z z z z z z z z z z z z z z z`#)*y08>0ELQX_gwknpuzˀ(5~     .0 P =HmtˆDJT\Ckqqr5ssdtkttuuhv{CˆEGHILMNOPQRSUWXYZ[]W udžˆFKV^ !! _Hlt22364808|̀|̀    $ ( 28AHMQ`epp=qBqqqqqr rrrrretltu"uuukwpwwwww+y4yyyyyzz {{w{|{|%|u~|~ƀɀ̀ahicklinA:\Dropout Revision.docahicklinlC:\Documents and Settings\ahicklin\Application Data\Microsoft\Word\AutoRecovery save of Dropout Revision.asdahicklinlC:\Documents and Settings\ahicklin\Application Data\Microsoft\Word\AutoRecovery save of Dropout Revision.asdahicklinA:\Dropout Revision.docahicklin6H:\meier stuff\Latino Dropouts_files\Dropout Paper.docahickliniC:\Documents and Settings\ahicklin\Application Data\Microsoft\Word\AutoRecovery save of Dropout Paper.asdahickliniC:\Documents and Settings\ahicklin\Application Data\Microsoft\Word\AutoRecovery save of Dropout Paper.asdahicklin6H:\meier stuff\Latino Dropouts_files\Dropout Paper.docahicklin6H:\meier stuff\Latino Dropouts_files\Dropout Paper.doc Nick Theobaldhome:web:reports:report021.doc@"Skkkkkkkkkkkkkll l!lMlTlZl[lzllllllllllllmmmm5m?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_abcdefghijkmnopqrsuvwxyz{~Root Entry FǬw1Table`{WordDocument%SummaryInformation(lDocumentSummaryInformation8tCompObjXObjectPoolǬwǬw FMicrosoft Word DocumentNB6WWord.Document.8