Effects of race, gender, and other characteristics of legally blind consumers on homemaker closure.

The Rehabilitation Service Administration (RSA) considers competitive employment in integrated settings the optimal employment outcome of persons with disabilities (RSA, 2004). However, noncompetitive employment outcomes such as homemaker are allowed under the Rehabilitation Act. Concerned with substantially higher rates of homemaker closures among individuals with visual impairments compared with individuals with other disabilities, RSA is currently assessing state vocational rehabilitation (VR) agency policies and practices relating to homemaker outcomes (RSA). This study provides information relevant to this activity by examining demographic and other factors associated with the homemaker closure of VR clients who are legally blind.

Recognizing early-on the commitment of RSA to support individuals with disabilities in the achievement of high-quality competitive employment outcomes and the need to reduce homemaker outcomes, the National Council of State Agencies for the Blind (NCSAB) convened a "homemaker panel" at its Spring 2002 meeting. This panel conducted a preliminary review of state VR agency policies for serving homemaker cases. Advantages and disadvantages in retaining the homemaker closure as a successful employment outcome for blind or visually impaired clients and concerns regarding the perceived excessive use of homemaker closures in some VR programs were discussed.

The homemaker closure has been recognized as a successful employment outcome for clients of the state-federal VR program since its inception in 1920 (Rubin & Roessler, 2001, p. 27). However, a higher percentage of blind clients, compared with other disability groups, historically have been closed as homemakers-a trend that continues today. For example, in fiscal year (FY) 2001, of those with successful employment outcomes (Status 26 closures), approximately 43% of all legally blind VR clients, in comparison with 2% of clients with other disabilities, were homemaker closures. A substantial majority (74%) of these legally blind clients were females. Further, White legally blind females with successful employment outcomes were more likely than African American females (57% vs. 39%) to be closed as homemakers (Cavenaugh, 2003). As VR agencies serving blind clients implement policies to either eliminate or substantially reduce the number of homemaker closures, it is important for policy makers and administrators to be familiar with the demographic and disability characteristics of VR clients with homemaker outcomes and to be aware that consumers with this profile may have fewer opportunities to participate in the VR program.

Historical Review of the Homemaker Closure

Homemaker refers to men and women whose vocational activity concerns managing a home for themselves and/or for others (RSA-911 Case Service Report, 2000). This literature review briefly traces the rate of homemaker closure among blind and visually impaired VR clients during the last several decades. Studies examining demographic (e.g., gender, age, race, work history) and other disability characteristics of blind and visually impaired clients closed as homemakers are then reviewed. Care was taken to note, when available, if findings were specific to (a) only clients who were blind (includes legally blind), (b) only clients who were visually impaired (excludes legally blind), or (c) both clients who were blind and visually impaired.

Homemakers closure status and blindness. In 1979 only about 10% of all homemaker closures were legally blind (RSA, 1982, as cited in Kirchner & Peterson, 1982). An additional 12% of the total homemaker closures in the same year were visually impaired (Kirchner & Peterson), leaving a substantial portion of clients closed as homemakers who have other types of disabilities. The data reported in the RSA document may lead one to underestimate the frequency with which VR clients with blindness or visual impairment are closed as homemakers. In fact, among legally blind and visually impaired clients successfully closed in FY 1980, approximately 40% were homemakers (Kirchner & Peterson, 1982). In comparison, only 14% of successfully closed clients with other disabilities were homemakers. Johnson and Hafer (1985) reported even higher homemaker rates for FY 1981, at least for legally blind clients. They cited RSA data indicating that approximately 56% (5,346) of legally blind clients successfully closed were homemakers. No rates for visually impaired clients were reported.

By the late 1980's and early 1990s, the homemaker closure rate across all disability groups had decreased to between 9% and 10% of all successful closures (U.S. Dept. of Education, 1992). During this same time period, the homemaker closure rates for blind and visually impaired clients remained stable, ranging from approximately 50% to 55% for legally blind clients and 30% to 33% for visually impaired clients (Cavenaugh, 2003). Rates continued to decrease throughout the 1990's for all disability groups with only 5% of all successful closures closed as homemakers during FY 1999. The homemaker closure rate for legally blind and visually impaired clients also declined, with 45% of legally blind clients and 22% of visually impaired clients closed as homemakers in FY 1999.

Other predictors of homemaker closure status. Findings from a 1973 U. S. Department of Labor report showed that between 1920 and 1970, 43% of all rehabilitation closures (all disabilities) were female and that 32% of these females were homemaker closures (as cited in Johnson & ttafer, 1985). Thurer (as cited in Johnson & Hafer, 1985) also reported higher rates of homemaker closures for females than males, indicating that in 1976, 1 of 3 female closures, compared with 1 of 15 male closures, were homemakers or unpaid family workers. Goldner and Liebman (1985) cited a RSA document indicating that in FY 1979, 82% of all homemaker closures were female. Similar to VR clients with other disabilities, blind and visually impaired clients closed as homemakers were more likely to be female than male. For example, approximately 76% of blind and visually impaired homemakers in FY 1980 were female (Kirchner & Peterson, 1982), 77% in FY 1999 were female, and 75% were females in FY 2001 (Cavenaugh, 2003).

Researchers have reported other demographic and disability predictors of homemaker closure. Giesen and McBroom (1986) analyzed data from blind and visually impaired clients closed in five states during FYs 1978 through 1980. They found that the typical homemaker closure was about 56 years of age with late onset of blindness, had multiple disabilities, was married, and had less education than competitive closures. Using FY 1980 data, Kirchner and Peterson (1982) also found that in comparison with competitive closures, blind and visually impaired homemakers were older, were married or widowed, received public support benefits, and had less education. In an analysis of FY 1982 data, Hill (1989) reported that age and gender were the most significant predictors of work status at closure for blind and visually impaired clients. Older clients were more likely to be closed as homemaker than in any other work status, and being female increased the likelihood of being closed as homemaker.

In a descriptive analysis of RSA FY 1999 data on blind and visually impaired clients, Cavenaugh (2003) found that homemakers were generally older, White, not married, had less education, and received public support benefits. Giesen and Cavenaugh (2003), examining RSA FY 1998 data on legally blind clients, reported few overall race differences but striking differences by race-gender combination, notably that a lower percentage of African American females, in comparison with White females, were closed as homemakers.

Summary of Review and Research Questions

As illustrated in the review of literature, the homemaker closure has been most prevalent among female VR clients who are blind or visually impaired. Previous studies also have found that demographic and disability characteristics including age, marital status, education, presence of secondary disability, and self-support/income are predictors of homemaker closure for blind and visually impaired clients. In addition, African American females may be less likely than White females to be closed as homemakers. The current study seeks to further explore the relationship of race and homemaker outcomes for blind VR clients. The following research questions were investigated: Are there differences in VR employment outcomes (competitive or homemaker) for legally blind clients who are African American or White? Further, are there differences in employment outcomes when controlling for education, presence of secondary disability, age, source of support, gender, and marital status?

Method

Data Source

The National Case Service Report (RSA-911) data, obtained from RSA, U.S. Department of Education, for the Fiscal Year 2001 (FY 2001) was used for analysis. The RSA-911 database is population data on all cases closed nationwide in the state-federal VR system in a given fiscal year. Each RSA-911 client record included demographic, socioeconomic, and disability information at referral; information on all types of services received; and outcome information (e.g., type of closure, earnings at closure) for all cases closed during the fiscal year. All RSA-911 FY 2001 complete case records for clients (N = 10,736) who were legally blind, had an Individualized Plan for Employment (IPE), and who were subsequently closed successfully employed (Status 26) were analyzed.

Analysis Variables

Variables employed in the analyses were homemaker vs. competitive closure status (the criterion) and predictor variables, race/ethnicity, secondary disability, self-support at application, gender, marital status, education, and age at application. The dichotomous criterion variable, homemaker closure status, was based on type of successful (Status 26) closure. Homemaker closure status was coded "1," and competitive employment, self-employment, and employment in the business enterprise program were coded "0." Because the focus was to investigate differences in the rates of homemaker and competitive closures, all other successful closure statuses (extended employment and unpaid family worker) were excluded by coding them as missing data on this variable. Race/ethnicity categories were White, African American, Native American, Asian American and Pacific Islander, and Hispanic of any race. Four indicator variables were created using the SPSS logistic regression procedure with "White" designated as the reference category. Four indicators also were created for marital status with "currently married" as the reference category. Secondary disability was dichotomous with "1" indicating presence of a secondary (nonvisual) disability and "0" otherwise. Support at application was coded to reflect findings from previous research (e.g., Cavenaugh, Giesen, & Pierce, 2000) that client self-support at application, in contrast with other types of support, is predictive of competitive employment outcome. Thus, self-support at application was coded "1" and "0" otherwise. Age at application was in years. Gender was coded "1" for female and "0" for male. Education was in number of years of formal education completed.

Procedure

We provide a descriptive profile on the analysis variables crosstabulated by homemaker closure status, race, and gender. We then investigated the exclusive role of race/ethnicity on homemaker closure status. Particular attention was given to the contrast between White and African American clients. This was examined in Model 1--Race/Ethnicity Only Model. The remaining independent variables have been identified as correlates of successful or competitive closure. These variables were added in a block hierarchically to the set of predictors to provide statistical control. Thus, Model 2--Race/Ethnicity Controlled Model--examined the role of race/ethnicity on homemaker closure status while controlling for age, gender, education, secondary disability, self-support at application, and marital status.

Data Analysis

Logistic regression using SPSS Version 11.5 was employed to analyze both models. Logistic regression (LR) was selected for the analyses because the criterion variable (homemaker closure status) was dichotomous, and LR is preferred in this context (Hair, Anderson, Tatham, & Black, 1998). Further, the assumptions of LR were met. Also, LR allows for inclusion of control variables that are expected to influence outcome, and LR provides odds ratios that may be used to compare outcome groups. For example, an odds ratio indicates how a White female and an African American female client may differ in their odds of closure as a homemaker. In this study, odds ratio values greater than 1 indicated that the odds likelihood for a person from a specific group (e.g., African American) was greater than that for a person in the reference group (e.g., White) in obtaining the designated outcome (homemaker closure status). Odds ratio values less than 1 for the same variable indicated the odds likelihood for an African American client was less than that for a White client.

Results

A descriptive profile of homemaker and nonhomemaker closures is given in Table 1. Breakdowns by race and gender also were given.

Race/Ethnicity Only Model

The first research question (Race/Ethnicity Only Model--Model 1) investigated differences in likelihood of homemaker closure for legally blind VR clients who were either White or African American. The likelihood ratio test for race/ethnicity on homemaker closure status (Model 1) was statistically significant, [chi square] (4, N = 10,736) = 93.66, p < .001, indicating that race/ethnicity categories alone were statistically reliable predictors of homemaker closure. The Hosmer and Lemeshow goodness-of-fit test failed because of insufficient available categories associated with only one predictor in the model. However, on indices related to effect size, the percent correct classification did not increase from a base rate of 55.3% when race/ethnicity entered the model, and the Nagelkerke pseudo-[R.sup.2] (see Menard, 2002, pp. 20-25) was .012, both indicating a very small effect. Additional results for the Race/Ethnicity Only Model are shown in Table 2. Race/ethnicity, when considered alone, was found to be a significant but weak predictor of homemaker closure. The regression coefficient for African American versus White/Caucasian (Table 2) was statistically significant, p < .001; the odds ratio for this predictor was .617. Compared to White/Caucasian legally blind clients, African American legally blind clients had 38% lower odds (.617 times as likely) of being closed as homemakers than of being closed in competitive employment--when considering race alone.

Race/Ethnicity Controlled Model

The Race/Ethnicity Controlled Model (Model 2) was hierarchical, entering the control variables, education, secondary disability, age, gender, self-support at application, and marital status, one at a time, and then as a block subsequent to race/ethnicity. The likelihood ratio test for Model 2 was significant, [chi square] (13, N = 10,736) = 5,896, p < .001, indicating that the set of predictors in the model were statistically significant predictors of homemaker closure vs. competitive closure status. The Nagelkerke pseudo-[R.sup.2] = .566, indicating a large effect size. The Hosmer and Lemeshow test was not significant, [chi square] (8, N = 10,736) = 6.99, p = .54, indicating good fit between the model and the data. The overall percent correct classification was 81.5% with sensitivity of 83.9% and specificity of 76.2%. Additional results for Model 2 are shown in Table 3. Race/ethnicity was significant with each of the control variables entered individually except for age at application. With me control variables included in the model as a block, race/ethnicity no longer was a significant predictor of homemaker vs. competitive closure status, p =.273. Results for the odds ratios for the significant control variables (Table 3) indicated that a person with more education or who was self-supporting at application was less likely to be closed as a homemaker. A person with a secondary disability, who was older, female, or widowed (rather than married) was more likely to be closed as a homemaker. Coefficients, odds ratios, and significance levels are provided in Table 3. Interaction effects were investigated and found to be minimal. (1) It was concluded that the above block model without interaction terms was a good, parsimonious fit to the data.

All predictors were examined for intercorrelations and possible multicollinearity. All independent variables, with appropriate indicator codings, were entered into an ordinary least squares (OLS) multiple regression routine in order to obtain tolerance values and intercorrelations to assess multicollinearity. Zero-order correlations between predictors were generally quite small (< .20) except for those between age at application and gender (.23), and age at application and widowed (.55). Obtained tolerance values for all variables clearly indicated no problems related to multicollinearity.

The OLS multiple regression also was examined for importance of individual predictors based on the squared part correlations. These followed the same order as the Wald [chi squae] values from the logistic regression shown in Table 3. Thus, the predictors of homemaker closure status, in order of importance, were age at application, self-support at application, gender, education level, presence of secondary disability, and marital status.

Discussion

Previous research clearly documents the relationship of gender and the presence of legal blindness with homemaker employment outcomes for VR clients. This relationship was again corroborated in this study. Also, the current study added information regarding the impact of race and other demographic variables on homemaker closure among legally blind VR clients. The limited, narrowly-focused examination of race differences alone on homemaker closure appeared to indicate that White legally blind clients are more likely to be closed as homemaker than African American legally blind clients, whether they be male or female. This comparison, however, is misleading because there may be other differences between White and African American legally blind clients that relate to employment outcome.

For White and African American legally blind clients who have the same rate of secondary disability and who are of the same marital status, age, gender, sell-support status, and education, the comparison between White and African American clients does not yield different expectations about their likelihood of being closed as a homemaker. Further, we conclude that higher levels of education and presence of self-support at the time of application are associated with lower likelihoods of homemaker closure for clients who are legally blind. Legally blind clients who are female, older, have a secondary disability, or who are widowed (rather than married) are more likely to be closed as a homemaker.

These findings underscore the importance of examining race differences in socioeconomic and demographic context. What initially may appear to be a difference in outcome due to race/ethnicity may actually reflect differences between race/ethnicity groups on other variables that affect outcome. Our finding that the race/ethnicity effect disappeared only when age at application was included in the model, suggests that the initial race/ethnicity effect is mediated only by age. We note that White female clients tend to be older than African American female clients (Ms = 59.0 vs. 48.3 years, respectively). It is the age difference--not the race difference--that makes for greater likelihood of homemaker closure for White clients. This finding is consistent with Hill (1989) who, using a similar statistical procedure, reported that older clients who were equated on other personal characteristics were significantly more likely to be closed as homemakers.

Additionally, caution needs to be exercised in the interpretation of odds ratios for variables with differing metrics. In this study, race/ethnicity, self-support, secondary disability, and marital status classifications all consist of dichotomous comparisons. Consequently, the odds ratios for these variables can be directly compared to gauge relative importance. However, the odds ratios for age at application and years of education must be interpreted for a one-unit (one year) change. For example, the odds ratio for age at application is 1.072--apparently not a very large odds ratio. However, this should be interpreted as a 7.2% increase in , odds of homemaker closure for each additional year of client age. Such coefficients may be more meaningfully interpreted in increments of the predictor that are of a size that is "clinically significant" (Menard, 2002). For example, a 10-year age difference would be associated with a 72% greater odds of homemaker closure.

Conclusions

Our findings show that among legally blind clients, when personal characteristics (age, self-support, education, secondary disability, and marital status) are controlled, the odds of being closed as homemaker are no different for White and African American males and females. Further, whereas age was the best predictor of homemaker closure, there likely are factors associated with age-unavailable in this investigation--that actually account for the manifest age effect. This observation leads to questions such as why proportionally so many homemaker closures are older White females. Future research should investigate rates of homemaker closures for older clients by race and gender, and investigate other factors associated with age, such as if utilization of blindness VR services differs across age levels by race/ethnicity and gender and if so, possible reasons for these differences. For example, legally blind African American females who would be viable homemaker candidates may avoid VR services for reasons such as lack of minority professionals, reliance on their own extended family/support systems, or unfamiliarity with VR services.

Future studies of outcomes in the state-federal VR system also should closely examine whether effects apparently due to variables of primary interest (e.g., gender, race) are mediated indirectly by correlated factors. In order to accomplish this, future research must avoid specification errors by including potentially relevant variables, employ statistical methods and models that enable statistical control over concomitant variables while focusing on variables of primary interest, and employ approaches that allow assessment of mediated effects. For example, examination of gender differences on homemaker closure rates may need to control for education, race/ethnicity, age, and so on. Procedures such as multiple regression (both OLS and logistic), analysis of covariance, and path analysis methods may be useful in this context.

This study is limited in that there are likely other variables that influence employment outcomes (e.g, counselor effectiveness, gender biases, agency policies on serving homemakers). Further, in an ex post facto study, caution needs to be exercised in making causal inferences. Despite these limitations, ex post facto studies have played an important role in understanding process and outcomes in the state-federal VR system (Bolton & Parker, 1988). This study is important in that it is the first to focus on the role of race/ethnicity and the homemaker closure for legally blind clients. Although homemaker closures are continuing to decrease, it remains the most frequent employment outcome for legally blind females--55% are homemakers (Cavenaugh, 2003). VR administrators must be familiar with the demographic and disability-related profile of homemaker closures to assess how polices that restrict, or eliminate, homemaker closures will impact services to legally blind clients.

Table 1
Descriptive Statistics for Analysis Variables
by Outcome, Race/Ethnicity, and Gender

                                        Homemaker           Competitive
                                                                (a)

Measure                                  African             African
                                White   American    White   American

         (a) Females

Age at Application         M    69.7    57.2        43.3    41.0
                           SE   .32     .75         .35     .57
                           n    2829    431         1920    529

Self-supporting
at Application (b)         M    .09     .07         .37     .33
                           SE   .01     .01         .01     .02
                           n    2846    435         1938    532

Highest Grade              M    11.9    11.2        13.1    12.5
                           SE   .05     .18         .07     .14
                           n    2835    434         1918    527

Secondary Disability (b)   M    .54     .64         .34     .33
                           SE   .01     .02         .01     .02
                           n    2847    435         1938    532

% Married (c)                   34.4    20.7        42.6    23.1
% Widowed                       46.2    29.9        6.7     5.6
% Never Married                 7.6     23.4        30.6    44.0

      (b) Males

Age at Application         M    63.8    54.5        42.1    40.7
                           SE   .61     1.02        .32     .55
                           n    931     179         2289    583

Self-supporting
at Application (b)         M    .12     .09         .36     .36
                           SE   .01     .02         .01     .02
                           n    936     179         2313    588

Highest Grade              M    12.0    11.2        13.1    12.5
                           SE   .12     .27         .06     .12
                           n    934     179         2280    575

Secondary Disability (b)   M    .55     .60         .34     .27
                           SE   .02     .04         .01     .02
                           n    936     179         2314    588

% Married (c)                   47.8    39.7        45.9    35.2
% Widowed                       22.2    11.2        2.0     3.2
% Never Married                 17.2    25.7        39.3    43.7

Note. Other race/ethnicity categories were not shown.

(a) Competitive included competitive, self
employed, and business enterprise closures.

(b) Mean represents proportion with the variable attribute.

(c) Other categories of marital status were not shown.

Table 2
Logistic Regression Results for Race/Ethnicity Only Model

                                               Wald
Variable                     b       SE    [chi square]

Race/Ethnicity                             91.662
  African American         -.482    .055   75.856
  Native American          -.507    .286   3.148
  Asian/Pacific Islander   -.523    .182   8.267
  Hispanic of any race     -.280    .070   15.985
Constant-                  -.104    .023   21.159

Variable                   df       p      Exp (b)

Race/Ethnicity             4        .000
  African American         1        .000   .617
  Native American          1        .076   .602
  Asian/Pacific Islander   1        .004   .592
  Hispanic of any race     1        .000   .756
Constant-                  1        .000   .902

Note. The reference category was "White" for Race/Ethnicity.

Table 3
Logistic Regression Results for Race/Ethnicity Controlled Model

                                           Wald
                            b        SE    [chi square]

Race/Ethnicity                             5.147
  African American         -.060    .073   .683
  Native American          -.689    .363   3.596
  Asian/Pacific Islander   .210     .235   .801
  Hispanic of any race     -.034    .095   .129
Education                  -.108    .008   172.350
Secondary Disability       .670     .054   156.045
Age at Application         .070     .002   1321.954
Self-support at Referral   -1.827   .071   660.580
Gender                     .844     .055   235.255
Marital Status                             74.741
  Widowed                  .773     .094   68.234
  Divorced                 -.003    .077   .002
  Separated                -.160    .144   1.241
  Never Married            .047     .074   .406
Constant                   -3.107   .157   391.813

                           df        p     Exp (b)

Race/Ethnicity             4        .273
  African American         1        .409   .941
  Native American          1        .058   .502
  Asian/Pacific Islander   1        .371   1.234
  Hispanic of any race     1        .720   .967
Education                  1        .000   .897
Secondary Disability       1        .000   1.954
Age at Application         1        .000   1.072
Self-support at Referral   1        .000   .161
Gender                     1        .000   2.326
Marital Status             4        .000
  Widowed                  1        .000   2.166
  Divorced                 1        .966   .997
  Separated                1        .265   .852
  Never Married            1        .524   1.048
Constant                   1        .000   .045

Note. The reference category was "White" for
Race/Ethnicity and "Married" for Marital Status.

Acknowledgement

This research was supported, in part, by Grant H133B000903 from the U.S. Department of Education, National Institute on Disability and Rehabilitation Research, by subcontract from Howard University.

References

Cavenaugh, B. S. (2003). [Analysis of Rehabilitation Services Administration 911 National Data, Fiscal Years 1992-2001]. Unpublished raw data.

Cavenaugh, B. S., Giesen, J. M., & Pierce, S. J. (2000). Rehabilitation of visually impaired persons in separate and general agencies. Journal of Visual Impairment & Blindness, 94(2), 133-145.

Bolton, B., & Parker, R. M. (1998). Research in rehabilitation counseling. In R. M. Parker & E. M. Szymanski (Eds.), Rehabilitation counseling: Basics and Beyond (pp. 437-470). Austin, TX: Pro-ed.

Giesen, J. M., & Cavenaugh, B. S., (March, 2003). Race and gender in blindness VR: Historical trends and neglected interactions. Presented at the 2003 Annual Meeting of the Southeastern Psychological Association, New Orleans, LA.

Giesen, J. M., & McBroom, L. W. (1986). The blind homemaker closure: A multivariate analysis. Mississippi State, MS: Rehabilitation Research and Training Center on Blindness & Low Vision.

Goldner, N., & Liebman, J. (1985). Homemaker as a vocation: A dilemma for rehabilitation agencies. Journal of Rehabilitation, 51(4), 41-46.

Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. G. (1998). Multivariate data analysis, 5th edition. Upper Saddle River, NJ: Prentice Hall, Inc.

Hill, M. A. (1989). Work status outcomes of vocational rehabilitation clients who are blind or visually impaired, Rehabilitation Counseling Bulletin, 32, 219-230.

Johnson, S., & Hafer, M. (1985). Employment status of severely visually impaired men and women. Journal of Visual Impairment & Blindness, 79(6), 241-244.

Kirchner, C., & Peterson, R. (1982). Statistical brief #21. Journal of Visual Impairment and Blindness, 76(10), 426-428.

Menard, S. (2002). Applied logistic regression analysis (2nd ed.). Thousand Oaks, CA: Sage Publications, Inc.

Rehabilitation Act Amendments, Pub. L. No. 102-569 (codified as amended at 29 U.S.C. 701 et seq (1992)). Rehabilitation Services Administration. (2000). Reporting manual for the case service report (RSA-911). Washington DC: U.S. Department of Education.

Rubin, S. E., & Roessler, R. T. (2001). Foundations of the vocational rehabilitation process. Austin, TX: PRO-ED, Inc.

U.S. Department of Education, (1992). Annual Report to the President and to the Congress FY 1992 on Federal Activities Related to the Rehabilitation Act of 1973, as Amended. Washington, D.C.

(1) Based on previous findings (Giesen & Cavenaugh, 2003), we were concerned about interactions involving race and gender. To investigate this possibility, we entered as a block into the previous analysis first-order interaction terms involving race, gender, age, and self-support. Although the change in the model chi-square from 5896 to 5929 was statistically significant ([chi square] (10, N = 10,736) = 33.55, p < .001), our use of a very large sample size (population) makes the test of statistical significance hypersensitive and less useful as an indicator of importance. Further, the overall percent correct classification remained at 81.5% --exactly the same as that for the model with no interaction terms included. Given these results, we concluded that the interactions were negligible.

Paula R. Warren

Mississippi State University

Brenda S. Cavenaugh

Mississippi State University

J. Martin Giesen

Mississippi State University

Paula R. Warren, RRTC on Blindness and Low Vision. P.O. Drawer 6189, Mississippi State, MS 39762. Email [e-mail]