012613 (Sa) A Quantitative Research Project- The Internet Usage and Low Income Mothers

 Do the Knowledge of Using the Internet and the Annual Income Level Affect the Total Life Skill Assessment of Rural Low Income Mothers?      
By  Li-chin (Crystal) Huang at University of Minnesota -Twin Cities

Abstract

This paper investigated the knowledge of using Internet and income levels relating to rural low income mothers’ self-report life skill assessments. The data for this study are from a multistate state research effort- the Rural Families Speak project (RFS) focusing on the well-being of rural low-income families in the context of welfare reform. A two-way between subjects ANOVA shows a main effect of the knowledge of Using the Internet on the Total Life Skill Assessment Scores but not the Income Level, nor the interaction between these two independent variables. Due to the nature of the data collection, the imitations of research results were discussed and further research direction and policy implication were provided.

Key words: Internet (Information technology), rural low income mothers, two-way between subject ANOVA, purposeful sampling, and welfare reform.

                                                             Introduction

Having grown up from an early era that socio-economic deprivation was the norm of the 50s-60s in Taiwan, I experienced to facilitate my never-got-a-chance to enter a day of  elementary school education mother to use the telephone or elevator when I was a young adult.  She passed away without succeeding in how to use any of them in the post-modern stage of Metropolitan Taipei. And she is the most wise, noble, and intelligent individual that I have ever known on this planet, who raised 10 good children with her bare hands.  I always have this imagination- if my mother got an education and learned the power of information technologies, who would say that she would not be much more capable than that of Hillary Rodham Clinton, Marie Curie, or Augusta Ada Byron King/Babbage!

Though information technology (IT)[1] is no longer considered a luxury in many affluent societies nowadays, the digital divide[2] continues to be a major concern both in domestic and international discourse (NPA Report, 2002; PEW, 2004).  Statistics shows a significant gap in access and knowledge of IT among the rural, urban and suburban areas where socio-economic status, gender, and racial background distinguish such disparity (World Bank, 2002).  The far behind in accessing the opportunity and resources relating to IT implies the increasingly marginalized mothers who, as Dean (2000) remarks that people lacking the access to the information technologies will become the second class citizens of the future in the American society.[3] 

The aforementioned perspectives are from my personal experience and observations from larger societies that lead me to ponder and investigate issues concerning social affordance and justice relating to information technologies, in particular, issues relating to information technologies that can serve as a positive mechanism to shorten the socio-economic gaps among various social strata.

 Luckily enough, during my being a full-time student in my sabbatical leave of 2007-8, I got the chance to join a team project that was important and linked to my academic background. That was the RFS- Rural Families Speak.  The data of this study is from a multistate longitudinal research project focusing on the well-being of rural low-income families in the context of welfare reform.   My team leader is associate professor Dr. Susan Walker of Department of Youth, Family, and Community.  Dr. Walker has been engaged in applying technologies in teaching and learning, thus that provided me the great opportunity to investigate information technologies in the rural low income mothers’ lives tied to my current major – Learning Technologies. Among so many published papers and projects associating with RFS, we are the first group studying IT in rural low income mothers’ lives.

Due to the nature of our inquiry and the original data collection, there is not so much quantitative information available for our project.  So a predominantly qualitative investigation has been conducted, and the proposal was submitted in the early March of 2008.

For this project, I consulted Dr. Walker as well about a possibility to conduct a mini research from a quantitative approach using the RFS data set, and I got the permission as long as the relevant research ethics was kept intact.

Rural women’s access to computers in general, as people predicted, is limited. The impacts on children’s well-being can be significant due to mothers’ limits on employment, education and the exchange of information.  Welfare reform policy intends to support the low income mothers to be independent by providing job and skill training, education and other relevant self-sufficient strategies. Based on this policy implication, I found in the original survey, there was a set of questions[4] surrounding on mothers’ self-report life skill assessments. I was curious about whether the Internet usage played a role in mothers’ daily lives that assisted their daily engagements in a certain way. 

Furthermore, as our class coverage got into richer dimension, I also added up an independent variable – annual income level into this study.  Though low income mothers living in under poverty line, due to various types of resources available to some of them, such as number of children, child supports, food stamps,  personal job and partner’s income, WIC,[5] EIC,[6] and other subsidies, the income presents a range of variation.

 According to PEW’s 2004 general survey, it reported that nearly half of the rural households earn under $30,000, which is a significant threshold for Internet use. In rural areas like other places, the likelihood of going online rises as income rises (PEW, 2004, p.19).
             Based on the aforementioned information, my research question is “Do the knowledge of using the Internet and the annual income level affect the total life skill assessments” ?   

Both of the independent variables -  Knowledge of Using the Internet and the Annual Income Level have 2 levels with Yes or No answered by the participants to the former, and high income and low income manipulated (by recoding) to the latter variable.  The depend variable Total Scores of Life Skill Assessment Scores is ranging from 14 to 25 points which is a quantitative variable.
            The three sets of hypotheses according to my research question are:
              H₀: μinternetY= μinternetN
              Ha: Not H₀

  H₀: μincomeH= μincomeL
              Ha: Not H₀

  H₀: There is an interaction between the Internet usage and Income level
              Ha: There is no interaction between the Internet Usage and Income level

                                                     Description of the Study

RFS project began in 2000. Data were collected from families in rural counties (with population centers of less than 20,000) in 15 states. So this was not a random sampling from the fifty states but was a purposeful sampling[7]  Participants were mothers with at least one child under the age of 12 years living at home and with an income at or below 200% of the federal poverty line. Participants were recruited primarily through social service providers. Participants were interviewed by trained researchers for approximately 2-3 hours, either at their homes or in another mutually agreed-upon location, and answered questions in a semi-structured format. Qualitative and quantitative data were gathered on many aspects of individual and family well-being. To date, three waves of data have been collected.

Because participants in the Rural Families Speak project were not randomly selected, demographics of the RFS sample were compared to a Current Population Survey (CPS) subsample of low-income, rural families with minor children and similar incomes (200% of poverty or below). Although the ethnic representation of the RFS sample was comparable to that of the national CPS subsample, the RFS mothers were younger, less likely to have a high school diploma or equivalent, less likely to be married, and had more children. RFS households had lower incomes and were more likely to receive government assistance. Compared to other low-income, rural families with children, the RFS sample represents families at greater risk for poor outcomes and more likely to be using social services. With a significant proportion of poor children and their families living in rural areas (U.S. Bureau of Census 2002), the RFS project provides rich data about a largely ignored at-risk population.

Sample

Among the 3 waves,   the research team added two questions relating to the knowledge of Internet into the W2 questionnaire. And these were the ONLY questions among many relating to information technology and my research interests. These two questions were “Do you know how to obtain assistance to using the Internet”, and “Do you know how to use the Internet”.  Thus I selected W2 as my targeted samples. Among 413 cases in the wave 2, only 124 participants completed all the interview procedures.[8] I am interested in the knowledge of using the Internet relating to the life skill assessments which might have interaction with the income level to affect the total life skill assessment scores of the low income mothers.
Design
            To make this paper not too simple nor too complex, I considered that the two-way between subjects ANOVA was appropriate to use the given data set that I was working with. Why? Because as mentioned in a previous section, both of the independent variables -  Knowledge of Using the Internet and the Annual Income Level have 2 levels of analyses and the dependent quantitative variable Total Scores of Life Skill Assessment Scores is ranging from 14 to 25 points of measurements.

                                                Analysis of the Data

Informal Analysis

            Before my running the two-way between Subjects ANOVA to examine whether the “Income level” and “Use the Internet” (independent variables) have effects on the life skill assessment scores (dependent variable), and informal analysis is beneficial and important. It gave me the overall outlooks of means, standard deviations (center and spread) and outliers of the samples which could lead to a better understand the data set and a possible preliminary prediction of the outcome.  The following are the descriptive statistics of Use the Internet, Life Skill assessments, and Annual Income information to form an initial understanding of my questions and the data set.

Figure 1. Total scores of life skill assessment

Use the internet	Mean	N	Std. Deviation	Variance
no	17.4545	44	5.17801	26.812
yes	21.3500	80	3.21035	10.306
Total	19.9677	124	4.41530	19.495

                                              

Figure 2. Annual income

N	Valid	124
Mean		22774.2118
Median		21500.0200
Std. Deviation		14929.95942
Skewness		1.194
Std. Error of Skewness		.138
Minimum		.00

(For this study purpose, the annual income was re-coded into high and low two categories:
 1 is low income, which is less than $20,000 and 2 is high income, which is greater than $20,001)

Figure 3. Between-Subjects Factors

		Value Label	N
Use the internet	.00	no	44
	1.00	yes	80
Annual income	1.00	low	53
	2.00	high	71

                                                         

Figure 4. Descriptive Statistics -  Dependent Variable: Total scores of life skill assessments

Use the internet	Annual income	Mean	Std. Deviation	N
no	low	17.2000	5.09489	20
	high	17.6667	5.34600	24
	Total	17.4545	5.17801	44
yes	low	21.0000	2.41091	33
	high	21.5957	3.67518	47
	Total	21.3500	3.21035	80
Total	low	19.5660	4.06443	53
	high	20.2676	4.66586	71
	Total	19.9677	4.41530	124

Figure 5

                 There are several things we can note about our groups simply by examining the above boxplot. First we see the life skill assessment scores of mothers who said Yes to the “Use the Internet” are slightly higher (a little bit skewed to the left) than those mothers who said No to the “Use the Internet” no matter in low income (less than $20,000 annually) or high income group (more than $20,001) group.  Mothers who said No to the “Use of the Internet” did not differ in their life skill assessment scores and had greater variability than those mothers who said “Yes” in Using the Internet. Second, we noticed 3 outliers all happened in the Yes to Use the Internet groups: one (#222) was in the low annual income who said Yes to the “Use the Internet”. Two were (#223 and #148) in the high income who said Yes to the “Use the Internet”. It helps to identify the similarity and differences among the total life skill assessment scores groups.
                 As addressed in the early section that the RFS samples were not generated from any kind of random sampling procedures but provided by relevant welfare agencies through specific criteria selection processes such as income below the 200% poverty line and with children age under 12.  Thus questions regarding potential violating the homogeneity of variance or the normality of residuals were my basic concerns.
                To test the assumption of the homogeneity of variances, the Levene’s test of equality errors of variances showed the following result:

Figure 6.

F	df1	df2	Sig.
7.075	3	120	.000
Dependent variable: Total Life skill Assessment scores. Tests the null hypothesis that the error variance of the dependent variable is equal across groups. Design: Intercept+yr_incm2(Income Level)+kcr26_2 (Use the Internet)+yr_incm2*kcr_26_2

              The null hypothesis being tested by Levene’s is that the variance across all group are equal.  From the descriptive statistics of this case shown, the combination of 4 factors standard deviations, the largest (high/no 5.34 by squared, variance 28.5) is 4 times bigger than the smallest (low/yes 2.41 by squared, variance 5.8), and the  p-value <.05 was too small which means that the null hypothesis was rejected. The variances are not equivalent.

Figure 7.

To test the assumption of normally distributed residuals (from normal population), I saved the unstandardized residuals when running the ANOVA.  And the residuals looked like normal but skewed to the left.

            From the above informal analysis, it confirmed my concern that the variances across groups were not equal, and our data was not generated from random sampling processes.

ANOVA analysis

Tests of Between-Subjects Effects

Figure 8. Dependent Variable: Total scores of life skill assessment

Source		Type III Sum of Squares	df	Mean Square	F	Sig.
Intercept	Hypothesis	41888.869	1	41888.869	5316.143	.009
	Error	7.880	1	7.880(a)
kcr26_2 (Use the Internet)	Hypothesis	417.035	1	417.035	3585.509	.011
	Error	.116	1	.116(b)
yr_incm2 (Income Level)	Hypothesis	7.880	1	7.880	67.746	.077
	Error	.116	1	.116(b)
kcr26_2 * yr_incm2 (Use the Internet*Income Level)	Hypothesis	.116	1	.116	.007	.933
	Error	1957.852	120	16.315(c)

a.   MS(yr_incm2)   b.   MS(kcr26_2 * yr_incm2)    c.   MS(Error)

Figure 9.

Interpret the ANOVA output:

             In the 3 F-values, one corresponding to the first independent variable (Income), one corresponding to the second variable (Use the Internet), and one corresponding to the interaction between Income and Use the Internet. If either of the independent variables is significant, then there is a Main Effect of that variable on the life skill assessment scores. By looking at the table, only Use the Internet (kcr26_2) had a significant effect on the life skill assessment scores, which showed a large F statistics and a very small significance level of p-value. So “Use the Internet” (kcr_26_2) has a main effect on the total life skill assessment scores, but not the Income Level (yr_incm2) nor the interaction between income level (yr_incm2) and Use the Internet (kcr_26_2). See the Figure 9.
          In this study, both the two factors only had two levels (one of them had significance). I don’t need to proceed with post hoc comparison.

                                              Discussion, Critique, and Implication

           My research question is “Do the knowledge of using the Internet and the annual income level affect the total life skill assessments”?   From the ANOVA analyses, the knowledge of “Using the Internet” (kcr_26_2) does show a main effect on the Total Life Skill Assessment Scores but not the Income Level, nor the interaction between  these two independent variables.
           For the policy implication concerns, within this mini research scale, I might jump into a conclusion to suggest that provide inexpensive Internet (in particular, the broadband) accessibility to rural low income mothers would be a possible solution to enhance their life skill potentials, which in turn, might manifest in their skills and employability to reach the policy goal – independent mothers raising healthy children. However, as discussed in the informal analysis, due to the data was not generated from random sampling procedures as well as the violation of the assumption of homogeneity of variance, many other limitations need to be addressed before a discreet conclusion could be made. 

             Among the limitations, first of all, I only tapped the tip of the iceberg – selecting a very small amount of variables (two independent variables and one dependent variable isolated) from the whole RFS data set to investigate an aspect of the multi-faceted complex phenomena. Thus the conclusion might have very limited power of generalization to other similar situations. 
            Second, more specifically speaking, participants who have knowledge of using the Internet might not be the typical rural poor mothers. Educational level, generational dependency, race and ethnicity, age, marital status, how many children (in particular the very small ones), supporting system, knowledge of community resources, isolation conditions and so on could covariate with the current studied dependent variable.

           Third, the test reliability and validity [9] as well as internal consistency[10] of the 25 items in the life skill assessment questions (see in the footnote 3) need to be examined carefully as well. Were these items measuring the real life skills in rural low income context? Even in rural areas, there are deeper, more isolated areas than those of regions which are vicinity to the urban or suburban areas. Then some questions in the 25 items might not measure what exactly needs to be measured to represent the life skills across general rural low income mothers.

           In short, rural poverty, like many other poverty problems, is contextual and ecological. There are many confounding factors affecting the life skill assessment buildup. My simplified version of this research did not render those bigger issues as many other senior researchers attempted to tackle.[11]

              If I can redo this research in more details according to what I learned from this course, I would like to add either education, or age, or race/ethnicity as independent variable(s), each with more than two levels of analysis. That is to recode the current income into 3 levels, so as education, or age, or race/ethnicity. Then the 3-way between subjects ANONA with more than 2 levels of analyses will make this paper busy with potentially using various post hoc methods and analyses to understand what main effects occur under what specific conditions. And the interactions could be better understood.   In addition, the currently covered ANCOVA design can extend this mini research project by adding multiple variables that are believed to be strongly related to the dependent variable; or try the MANCOVA with two or more measured dependent variables such as the Total Community Knowledge Scores and the Total Parenting Ladder Scores added to the current Total Life Skill Assessment Scores if they are correlated with each other.  The other alternative approach is to use the mixed method of quantitative and qualitative approach as many other RFS researchers did and are doing from this data set.
         This two-way between subjects ANOVA project was started with a very simple idea generated from my belief that information technology (in this case, using the Internet) can accelerate the transformation of human condition, in particular, shorten the digital gap.  As this project evolved, I learned many things that I did not pay enough attention before, in particular, the insight gained from the explorative procedures on the configuration and patterns of the raw data.  Now I see the importance of conducting the informal analysis and selecting research methods and deign carefully.

         Human conditions are complex. In both projects -Dr. Walker’s and this EPSY 5262 final work, the necessary humility and caution in interpreting the research outcome is an important lesson which cannot be emphasized enough. Within the existing parameters, research can be done with carefully examining and assessing of the limitations and employing appropriate methods and design, then the vision and revision of the research will move toward more authentic research outcome depicting the genuine human lived and living experiences.

         I love both works.  I appreciate the opportunities from these two projects enhancing my research capacity. And I wish the current RFS project can be continued and extended to include more information technology questions into the questionnaire with open-ended inquiries in a longitudinal format. This is my DREAM - to probe into what have been exactly happening in the current rural low income mothers’ lives – the experience, activities, values, attitudes, barriers and supports  in the information explosive era.

                                                References

Field, A. (2005). Discovering Statistics Using SPSS, 2^nd ed. London: Sage Publications Ltd.

Gall, M, Gall, J, and Borg, (2003) W. Educational Research. 7^th ed.  Boston: Pearson Education,
       Inc.

---

[1] Information Technology (IT), as defined by the Information Technology Association of America (ITAA), is "the study, design, development, implementation, support or management of computer-based information systems, particularly software applications and computer hardware."  In this study, it is simplified as “Internet usage”.

[2] The term "digital divide", in this case, refers to the gap between those with access to the Internet and those without access. In an updated term, it can refer to having the high-speed access (broadband/wireless) or not.

[3]   Refer to http://www.netschools.net/whynetschool/cs hundred.htm 

[4] Do you have a driver's license? Car  insurance; Car  registration; Health Insurance; Checking account ; A good credit record; Local library card; Do you know how to manage your bills? Write a personal check; Make a family budget; Stretch groceries to the end of the month; Register to vote; Apply for a credit card; Prepare a well-balanced meal for the family; Get a telephone service; Work with the landlord to improve housing; Register a consumer complaint; Talk to your child's teacher; Fill out forms to apply for services; Apply for a job; Write a resume; Dress for a job; Fill out your own income tax forms; Join a local club or organization; and Create a personal support system.

[5] WIC is The Special Supplemental Nutrition Program for Women, Infants, and Children.

[6] EIC is the Earned Income Tax Credit (EITC) sometimes called the Earned Income Credit (EIC), is a refundable federal income tax credit for low-income working individuals and families. Congress originally approved the tax credit legislation in 1975 in part to offset the burden of social security taxes and to provide an incentive to work. When the EITC exceeds the amount of taxes owed, it results in a tax refund to those who claim and qualify for the credit.

[7]  Purposeful sampling is a process of selecting cases that are likely to be “information-rich” with respect to the purposes of a qualitative research (Gall, Gall, & Borg, 2003).

[8] In the vast rural areas, conducting interviews took more time and energy. Low income mothers also confronted many personal and situational factors that left the interview incomplete. These factors came from variety of conditions, such as personal illness, children and relative health conditions, relocation, partner’s job changing, failing to follow-up, as well as some newly participating states did not ask the Internet usage relating questions.

[9] Validity refers to in testing, the appropriateness, meaningfulness, and usefulness of specific interference made from test scores (Gall, Gall, & Borg, 2003).

[10] Internal consistency refers to an approach to estimate test reliability that examines the extent to which individuals who respond one way to a test item tend to respond the same way to other items on the test (Gall, Gall, & Borg, 2003).

[11] So far I have read 3 among about 50+ papers employing the structural equation or path analysis to engage with more complex analysis.