CC BY
12
Section 1. Demography
https://doi.org/10.29013/EJEMS-21-4-3-22
Jingyi Zhao,
under the direction of Prof. Philip Liang University of Chicago Department of Economics The Bear Creek School, United States E-mail: jingyi0624@gmail.com
THE EFFECTS OF INTEGRATING EDUCATIONAL AND HUMAN CAPITAL INVESTMENT TO PROMOTE ECONOMIC DEVELOPMENT: MODELING THE RATE OF RETURN AND ANALYZING STUDENT BENEFITS
Abstract. This paper examines the effects of investing in an educational program that focuses on cognitive and non-cognitive development and the benefits social-economically disadvantaged students obtain. We arrive at conclusions through data collected from the program addressed in the paper as well as results of quantitative summaries of outside literature. This paper finds that investment in such developmental programs for the financially disadvantage yields significant benefits in direct cognitive and non-cognitive abilities, leading to greater success in adulthood. This paper offers uniqueness in its conclusions and data supporting significant benefits revealed during both short term and long term programs. We utilize customized randomization methodology for the selection and grouping of candidates, in order to guarantee to the best of our ability accurate representation. Results of this early childhood educational program suggest an internal return of 35.85%.
Keywords: Economics, Education, Mathematics, Effect sizes, Learning, Investment, Incentives.
1. Introduction poor indicators of economic success and poverty al-
There is continuous debate about how to im- leviation. This paper aims at conducting and exam-
prove educational outcomes for American children, ining an educational program for socially-econom-
yet much of the data collected targets strategies (class ically disadvantaged young children with emphasis
size, motivators, location) rather than the fundamen- on both foundations of cognitive and non-cognitive
tal skills that students gain. While exploring strat- achievement. This program incorporates deciding
egies is important to developing educational pro- factors of adulthood success, overall abilities, and
grams, there is a lack of focus on the bigger picture later life decisions.
of a student from a young age to adulthood. Because In order to build the necessary foundation for a of this trend, education is naturally shaped by small better educational system, policymakers and pro-indicators of ability at a specific point in a student's grams ought to examine which skills should be delife. Unfortunately, using this framework for govern- veloped and how to incorporate both knowledge and mental suggestions as well as benefit predictors are social awareness into a student's education. This paper
models the effects of investing in different aspects of such an educational program and benefits for students.
This early childhood education program for social-economically disadvantaged students is conducted in the Greater Seattle area of Washington state. Beginning at ages 5-8, children were given both cognitive and non-cognitive skills classes 4 days a week in addition to specialized days for improvement tracking and background data collection. We use permutation testing and worst case approximations to validate our estimations for the program.
We calculate Hotelling's multivariate statistics for two samples for our randomization procedures. Our program has a significant internal rate of return at 35.85%, as well as significant treatment effects in cognitive development, non-cognitive development, adulthood economic success, and wage growth.
2. Background and Experimental Design
The experiment was performed with two phases between May 31 to August 5, 2021, centered in the Greater Seattle Area (4.018.598 people) ofWashing-ton state. 50 students were taken into consideration for the original sample from housing and family shelters that accurately represent all areas in Greater Seattle (with over 20 miles of distance between each home). Children were 5-8 years old, and data collection began with background information from their parents. Participants received three small-group (5 students per group) cognitive classes per week for the duration of the experiment in the format of online video calling, with an additional testing (Testing and result recording included: CogAT (Cognitive Abilities Test), Wechsler Intelligence Scale for Children (WISC-V), Stanford Binet Intellig reading comprehension testing, mathematical skills testing, social sciences/critical thinking testing, and emotional development testing). day every two weeks for progress tracking. Cognitive classes were 1.5 hours each, with a 1-hour follow-up session attached. Every week, financial or objective rewards (Financial incentives delivered by cash (for each family): $20.00 for a week of attendance in lessons,
$10.00 for each testing day and demonstrated effort for learning the material. Different rewards are offered for parents and younger children: New Playing Sets, puzzles and games, homeware and furniture, gift cards to eateries and stores) [4] were given to the treated group for the main purpose of testing the effectiveness of economic investment in the rate of return of student benefits. Specific education content was chosen carefully (Among the most widespread and versatile educational tools, tutoring - supplemental one-on-one or small group instruction - has been promoted as an effective method for helping students learn) (Nickow, Oreopoulos, Quan [13]) in order to reach the largest improvement within the given time frame. Participants and their families received social and background follow ups every two weeks, and 100% of the families were surveyed at every checkpoint. Background information collected included age, gender, economic status, parental income, education, status of housing, emotional habits, and parental investment in education. Children received a daily education plan with accordance to their age and grade level as well as the cognitive learning previously described. Students were guaranteed a face-to-face talking session with the researcher once every week.
3. Eligibility Criteria
The pandemic limited face-to-face interviewing with potential families and children, but the digital alternative of interviews allows the same effectiveness in choosing candidates (Digital learning also motivates educators and students to use latest technologies to communicate and deliver and share information with each other) (Tejasvee et al. [15]). In order to qualify for participation in the experiment, children had to (i) be 5-8 years old on the commencement of the first data-collection date; (ii) Be economically disadvantaged in consideration of U.S and Washington State poverty numbers (U.S Federal Poverty Guidelines used to determine financial eligibility for certain federal programs. For a more accurate representation of results within the specific area where candidates are
chosen, this eligibility factor is impacted heavily by housing costs where candidates live); (iii) Be socially disadvantaged without a stable shelter (Households are considered to be cost burdened if they spend more than 30% of their income on housing and severely cost burdened if they spend more than 50% of their income on housing. Cost-burdened households struggle with necessities and utilities (Kushel et al. [12]) In 2014: 83% of households earning less than $15,000 a year were cost burdened. This includes consideration of the number of people per bedroom within the household. Regular bedrooms with 3 or more people is considered burdened, as apartment and small house bedrooms are not designed to comfortably accommodate more than 2 people. A second method of measuring the burden of housing costs is calculating the absolute amount of money left after paying for rent; we consider this factor with the caveat that candidates within this specific experiment may not be in a situation of rental)or have parents with a level of education equal to or lower than community college. (iv) Lack educational resources compared to the average student of the same age (In this program specifically, educational resources are defined by: (i) Previous Schooling with respect to the student's age and respective grade level; (ii) Material resources such as books and online learning opportunities; (iii) Parent input and effort to assist student). Chosen families and children must be open to accepting money or small gifts, and they must be willing to accept interviews and background data collection. Relative to the average child in the United States, these children were at a disadvantage. Although other factors might have contributed to these children's disadvantage, the primary factors contributing to these children's disadvantage are poverty, lack of educational opportunities, lack of education, lack of stable housing, and difficulty in social interactions. The group of candidates are an accurate representation of those in poverty or require shelter in Washington State.
Children and families are committed to the full duration of the study. During classes and meetings,
they are required to refrain from using their electronic devices for other purposes.
This requirement eliminates a potential distraction. 4. Randomization Protocol with Mathematical Models
4.1 Describing the Protocol It is essential to understand the randomization protocol used to conduct any research. Before conducting randomization procedures, background information was tabulated into a spreadsheet program for organization. At the end of the first step, all 50 students were listed in the same manner on one platform, ready to be randomized.
Tentative treatment and control groups were created by simple random assignment. Then, these tentative groups were checked for balance on socioeconomic status (SES), gender, race, age, previous education, and parental income. If the groups were too imbalanced, the randomization was redone. The first two pairs of tentative treatment and control groups were rejected and the third pair was accepted.
The second step involves the usage of a volatile function ^AND for absolute randomization. The first helper column holds random values generated with the RAND function. Subsequently, the second helper column holds numerical values used to sort data, generated with a formula: = RANK (C5, rand) \+COUNTIF ($C$5: C5, C5) -1 This formula (By default, RANK will 1 will be associated with the highest value, 2 with the second highest value, and so on. Duplicated numbers are resolved by using the COUNTIF function. We do this by adding the result of this step to the value generated by RANK) backs up the displayed data. Additional information regarding the implementation is recorded in the additional appendix.
The third step is the redraw tentative treatment and control groups until tentative treatment and control groups are balanced. Group were balanced based off of the mean of an index of socioeconomic status (SES), the ratio of boys and girls, the ratio of race, the ratio of age, previous education, and
parental income. Exactly three tentative treatment and control groups were drawn, each with the same randomization method as step two.
The fourth step is moving separated siblings into one group. The only separated siblings were two pairs of siblings. One of these pairs was randomly assigned to the treatment group and the other was randomly assigned to the control group. This step completes the randomization.
5.2 Mathematically Modeling Randomization Procedures
The assignment procedure produced treatment and control groups that are more similar than 79% of treatment and control groups created by simple randomization. We arrive at this conclusion through a simulation. We transformed the data into numerical form using indicators for whether students had attended pre-K, Kindergarten, first grade, and second grade, binary variables for whether the student was Asian, White, Black/African, Hispanic, or Pacific Islander, and a binary variable for whether the student was male. We next generated 1000 random partitions of the data into treatment and control. For each of these partitions, we calculated Hotelling's multivariate two-sample i-squared statistic: x2c.
i
t2(A,B) = (Z. -Z„
V
At Îa + t1 îl A B
(za - zb
The Hotelling's multivariate two-sample i-squared statistic t2c maps a partition (A, B) of S
(such that |a| =
1 2
and B =
S
to № > 0 and is
given by
where Z. = Ia| 1P .Z , where Z is a vector containA | | l&A l 7 ,
ing information about the age, education, race, gender, and family income of participant i, ZB — 111" pBZ i,
Z A =-
A -1'
^^^¡^a(Z, -ZA)(Z, -ZA)' and Zb -
= -L- Y Z —zr)(Zt —zr)'.
B — 1Lu leBy l B'K l B'
Pi "
6. Present Value of Benefits and Costs Determination with Internal Rate of Return
Functions and surveys strongly suggest that enhancing education increases earnings earnings (Studies that estimates were retrieved from: Burkhead et al. [5], William J. et al. [16], Brown [10]). One obvious benefit is that better academic performance prepares students for higher education. Higher education, in turn, results in better job placements and higher earnings in adulthood. With this in mind, the effect ofinvestment in educational resources is most directly modeled with student performance on cognitive tests and evaluations. In order to present the conversion ofevaluation scores to potential earnings, functions and models will be utilized.
6.1 The Human Capital Earnings Function
6.1.1 HCEF Framework and Model
The correlation between education and wage determination will be established within the framework of Mincer (l993)'s human capital earnings function (HCEF).
2
Representative Symbol Usage
S Years of Completed Education
X Years an Individual Worked Since Completing Schooling
e Statistical Residual
The log of individual earnings (y) in a given time period can be modeled as an addition function of both a linear term for education and a quadratic term for experience:
log(y) - a + bS + cX + dX2 + e (l)
This structural pattern of earnings by age and education had been recorded at least since the ear-
ly 1950 s (Miller, 1955. P. 64-67) (shows the age distribution of earnings per year for three different education groups and on (i) the concavity of these profiles). The HCEF therefore is successful in utilizing both inductive and deductive reasoning to model accurate results (Mincer's equation acts as an approximation to a generic function,
log(y) - F(S, A) + e Since both S and A in the dataset are recorded as discrete, the function F() can be estimated nonpara metrically by including a complete set of dummy variables for all (S, A) pairs or by using smoothing methods non-parametrically, Zheng (1996): formal testing to compare the fit of expanded various versions of (1) to kernel density estimates) in smaller data-sets. Alternatively, higher-order terms have been added to incorporate age/experience and examine the improvement in fit relative to Mincer's original specification, providing an important improvement in fit:
ple size was increased because of cross-tabulation) is best represented in logarithmic form. (i) The distribution of logarithmic earnings relates closely to a normal distribution. (ii) The logarithmic transformation shows the success of the semi-log HCEF. (iii) It is fairly convenient for further interpretation.
Representative Symbol Usage
A Annual Earnings
w Hourly Earnings
h Hours per Week
d Weeks
Representative Symbol Usage
g Third/Fourth-Order Polynomial
log(y) = a = bS + g(X) + e (2)
6.1.2 Brief Discussion of the Education Coefficient The methodology of modeling earnings (A change in the data collection method of the U.S Census Bureau complicates the calculations, therefore, the sam-
A - wxhx6 (3)
Regression of logarithmic annual earnings on education and other controls causes the estimated education coefficient (Educational coefficients of statistical models are usually the educational return. The educational return is generally lower in models that control for potential experience using time not spent in school rather than experience) to be the sum of the education coefficients for parallel models of:
w ,h,9
Figure 1. Relationship between Mean log Hourly Wages and Completed Education
Table 1. - Estimated Education Coefficients from HCEF for Men and Women in 2019
w h d Annual Hours (log) Annual Earnings (log)
(a) Male
Education 0.100 0.018 0.025 0.042 0.142
Coefficient (0.001) (0.001) (0.001) (0.001) (0.001)
R-Squared 0.328 0.182 0.136 0.222 0.403
(b) Female
Education 0.108 0.020 0.032 0.054 0.163
Coefficient (0.001) (0.001) (0.001) (0.001) (0.001)
R-Squared 0.252 0.073 0.076 0.112 0.251
6.2 Present Value Benefits and Costs
Because this paper emphasizes the investment perspective, the precision with the timeline of modeling costs and benefits is important. If the research conditions are implemented in current education systems, the cost of hiring additional teachers and obtaining learning spaces are significant. All of this comes with the fact that the benefits are not realized until years later, after current students who benefit from the program join the labor market. In order to show the benefits and costs, we compare the treatment and control groups.
All costs of this program are incurred in a single year. We assume that tutors cost $17.76 per hour. We assume that we spend 4 hours of tutor labor training each tutor. Tutors also spend 14 weeks teaching classes. During each week, they spend 5.5 hours. We assume each tutor teaches 5 students. We also assume that a social worker spends 1 hour per week for 14 weeks supporting a group of 25 students. We assume that the cost of social worker labor is $42.21 an hour. These assumptions imply that each student requires $311.35 of labor.
The rapid and concentrated data collection time fits completely within the year 2021, which means we can somewhat accurately predict additional costs. Figure 3 illustrates the age-earnings profile for workers in 2021. The figure displays average income per year for workers at each age between 18 and 65. Refer to Figure 3, earnings rise before the 40 s, reach the maximum in the early 50 s, and steadily decline in later
stages oflife. Earnings for students who just j oined the labor force will remian low until the 20 s or 30 s. Let Et represent the average real earnings every year after age 18. Results from the research program indicate that a one standard deviation increase in mathematics or reading performance causes a 14% increase in earnings. We use P to represent the increase in earnings caused by a one standard deviation increase in mathematics or reading test scores and performance.
To compute the IRR, we use age-specific treatment effects on math and reading skill level improvements. Let 5 M denote the improvement in mathematics in standard deviations and let 5R denote the improvement in reading in standard deviations. It is
initial years, because a dollar earned in future years is less valuable than a dollar earned at the current time.
Representative Symbol Usage
n Number of Weeks
nh, j Hours per week Required from Workers of Type j
Two types of workers are required for this program: tutors (j = t) and social workers (j = s). The present value (PV) of costs of the program over a temporary duration of 14 weeks can be calculated by three parts. The primary expense is the payments to the tutors for teaching students:
Tutor Wages = nw x nh t x 17.76, (4)
where the tutor wage is reasonably assumed to be 17.76 dollars per hour. In this program, the tutor wage can be calculated with respect to every 5 students, resulting in 1367.52 dollars for 14 weeks. A modification can be made to incorporate the cost of training instructors for the program:
Tutor Cost = nw x nh tx 17.76 + (4 x 17.76) (5)
Therefore, the total tutor cost is 1438.56 dollars. On a per-student basis, the cost will be 281.71 dollars.
The second part considers the necessary work time of social workers with the children when instructors are not present:
Social Worker Cost = n x n, x 42.21 (6)
w h, s v '
We assume that 1 hour of social worker time per week is required for each group of25 students. Therefore, the total labor cost is 311.35 dollars per student.
The third part accounts for financial or other forms of rewards given to students and parents. Money or item rewards are a cost to the organization
funding the program but a benefit to their receiver which offsets the time and effort costs of learning. Specific financial incentives are discussed in the respective section of the paper.
In the future, I plan to implement a more intensive 4-year program. The present value (PV) of the costs for that program given the interest rate r is:
PV of Costs =Y—. (7)
¿1(1 + r) ( )
Ifa 5-year-old joins the program in the current year,
PV of Benefits - f E '^(5m +5r). (8)
¿3 (1+r y
6.3 IRR: Internal Rate of Return
In order to decide whether investments in education programs like the one undertaken in this specific research are worth making, we use the internal rate of return as a direct evaluation. However, we take into account that the IRR is only as accurate as the assumptions that drive it, so we use permutation testing to illustrate the credibility of the IRR.
Representative Symbol Usage
e\ estimate for permutation
<jp associate analitic standard error for permutation
yv e original estimate
/V <T A analitic standard error
R p Number of block permutationwithin small groups
In order to test if treatment and control groups have a common outcome distribution with a small sample size, control and treatment labels must be exchangeable. In the context of this program, the experimental labels (wage, gender, race, SES, education) are indeed exchangeable. Since we cannot
assume equal distributions, we will utilize studen-tized test statistics for our permutation testing. Using a simplified testing model that reaches asymptotic validity is suggested by Chung [17] and Heckman [5], we calculate for our p-value:
pP,S - (1 + RP
1II
p-1
e* e
>
<Ja
(9)
Representative Symbol Usage
CF t h t net pre-tax cash inflow or outflows during a single period t for individual i
r IRR that can be earned in alternative investments
t time period cash flow is received
T last period of the working life of the treated individuals
n number of individual cash flows
As background information on the IRR, trustees predict a 1.29 percent annual growth in real wages over the next 30 years (PSID, [14]). We calculate the IRR by solving,
F CFit
¿—¡t=1 i >f
0 =
(10)
t=0 (1 + r )t
Using the earnings data from 2018, data on wage growth between 2018 and 2021, and our projection of the growth rate of real wages, we see a predicted internal rate of return at 35.85%, a significant value.
7. Estimator of Treatment Effects
Representative Symbol Usage
D treatment status of participant i
Z . his or her vector of the four preprogram covariates"
Y participant's result of interest ib
Yd participant's counter-factual result ic
R binary indicator of whether the outcome Y is on record^
a. Cognitive Performance, index of socioeconomic status, gender, educational background
b. Within a relevant sub-sample P containing NP = |p participants.
c. His or her treatment status D is fixed at d g {0,10}
d. R. = 0 if Y. is missing and R. = 1 if Y. is not missing. Additionally, cluster-robust asymptotic standard errors can be used to studentize the estimator given by equation (8), allowing for correlation between error terms (participant siblings).
We estimate the effect of the program treatment with respect to this sub-sample P given by:
* =^ZY-if) (11)
N P ieP
Then, using observed data from the treatment:
Y, - DYl + (1 - D, Y (12)
Proceeding, we will use two different estimation methods: (i) the unconditional differencein-means (UDIM) estimator, (ii)the augmented inverse probability weighting (AIPW) estimator.
Step one estimates the parameter of treatment effect (6) with the UDIM estimator Iudim :
V RD.Y. V R. (1 - D, )Y ,
¿—l ieP i i i ¿—I ieP iV ,y i
n
u dim — '
(13)
V RiDi V Rt (1 - Dt)
tieP i i ¿—lieP i
In step two, we assume that D. (treatment status) is unconditionally independent of the counterfactual outcomes. The randomization process of this specific program justifies unconditional independence, which suggests no further need for a conditional ordinary least squares (COLS) estimator.
Step three takes into consideration that the UDIM estimator assumes non-response as a random factor. In order to accurately model growth and return in adulthood after completion of treatment during childhood, the outcome should not depend on conditions involving non-response and pre-program covariates (R, an indicator of whether Y. is missing, may be dependent on D. and the preprogram covariates Z).
Step four uses the augmented inverse probability wei ghting (AIPW) estimator, allowing for our considerations by using a weaker assumption that
Y 1 RkDZ;:
n aipa — / (jl 't
NP ieP
(14)
Here,
= ?f + (Y,d _ Y ) (15)
Representative Symbol Usage
Y, gender-specific ordinary least squares-based estimator e
e. With conditional expectation: E[Yt\Zt, D, = d, R, = 1] for d g {0,1}
8. Potential Modification: Additional Investment in Developing Non-Cognitive Skills
Representative Symbol Usage
i task
t age
Information set based on the come-evaluating agent
Rlt I-i) Anticipated reward per unit effort (activity j, time t)
A Other determinants of effort
u Vector of parameters characterizing prefences
Because there is considerable evidence of early disadvantage causing difficulty in cognitive and financial growth, this program also serves to target the development of non-cognitive skills: personality, socio-emotional, and character. Performing this modification on children is exponentially more effective due to the greater malleability of personality and character skills.
Investigating treatment effects of investment in non-cognitive skills offers more specific results and discussion. Academics and future schooling relies heavily on cognitive skills, however, future economic contribution and success is equally affected by cognitive and noncognitive abilities. In order to account for both cognitive and non-cognitive skills, this program uses a variety ofachievement tests (Refer to the Additional Appendix for detailed explanations of attributes in personality psychology used to design treatment methods of this program).
Vectors of skills at age t can be denoted by d, This program allows a model of the age-specific
outcome Y
i> t'
Y ,t ,t (0t ,e} ,t, X, ,t), j el..., Jt and t e1,...,T (16) Here, X, t is the purchased inputs that affect outcomes in vector form. e, t is effort and it is defined by the supply function,
The above equation encourages analysts to treat the actions of agents as primarily responses to incentives agents face (Mullainathan and Shafir [12]). I also allows non-cognitive factors to influence actions. Enough empirical evidence (Borghans et al. [2; 3]) demonstrates the importance of stable personality and other capacities in performance outcomes.
By controlling for different incentives and situations, we are able to accurately identify capacities. However, there is still difficulty in associating a factor with exact measurements. (define factor) Therefore, we can only identify factors (See Anderson and Rubin (1956) and Williams (2012)). relative to each other (Almlund et al. [2]). Equation (15) allows one to account for how different influence variables affect Y, showing the effects of investing in non-cognitive development in addition to cognitive improvement. Outcomes affected include but are not limited to: economic success, earnings, education, and various behaviors.
9. Financial Incentives and Effects on Student Education
The motivations considered while designing the methods of this program includes incorporating the impacts of financial incentives and rewards presented to the parents. This section focuses on how incentives and investments factor into student education and future abilities. Recent literature demonstrates how incentives promote learning among students, focusing on the technology of skill formation (Cunha [17]).
Representative Symbol Usage
ot self productivity and cross effects
f(t) twice continuously differentiate"
t stage of the life cycle
dimension)
e, t =St (6t, Aj t, Rl (Ij u))
(17)
a. increases in all the arguments, concave in It The affects can be modeled by,
dt +1 - f(t )(dt, It ,6P ,t ) (18)
This allows accurate depictions of ability-forming stages and the effectiveness of investments during critical stages in a student's life. The earliest version of this model was developed by Ben-Porath (Browning [10]). The first term of Equation 17 also demonstrates the correlation between non-cognitive and cognitive abilities (one promotes the other). In order to track long term correlations and effects, we model the connection of investments and later life stages of students,
8 -0(t+1)_ > o, At Later Life Stages (19)
e-et -8-i :
As age grows,
82 -a
(t+i)
111
(20)
8-et-8-i't
This model is consistent with other literature in the area, strongly suggesting that investments during younger ages is considerably more effective than at later life stages (Heckman and Kautz, 2014; Knud-sen et al, [17]). Here, It T, therefore, 0t+1 T . We will account for self-productivity when calculating,
ot+i t^et+s t, 5 > 1, 82 -a
(t + 5+1)
8-1 -8-1'
> 0, 5 > 1
(21)
Investing during period t+s will complement any investment made earlier in period t, proving
how investments in early life contributes to later life investments. Here it is shown that investment in economically disadvantaged young children provides just social outcomes and is economically efficient.
10. Financial Incentives and Effects on Parents
In addition to the effects financial incentives have on children, parents are also noted to play an important role. This section of the paper will discuss both the effects parental skills have on students and also how economic incentives affect parent decisions.
Deviating slightly from the models that analyze student effects, parental input can also be placed into the effectiveness model,
82 -a
y(t+i)
8-aP ,t-8-i :
> o
(22)
Outside literature provides evidence that more engaged parents strongly increases the success of investments made (Lareau [2]). Other early childhood programs also notes important results measured by the quality of home environments. Other effects include having more motivation about parenting and fulfilling their role in helping students discover their abilities and character (Cole et al., 2012).
Below, we will discuss the results recorded from this program.
Table 1.- Demographic Sample: (S = 50)
Characteristic Detected Result(S%) Mean Value(Standard Deviation)
1 2 3
Parent Age N/A 32.5(8.6)
Mother 39(78.0%) N/A
Father 11(22.0%) N/A
Hispanic 6(12.0%) N/A
Pacific Islander 4(8.00%) N/A
African American 5(17.24%) N/A
Other 24(48.00%) N/A
Less than a high school diploma/GED 15(30.00%) N/A
High school diploma/GED 19(38.00%) N/A
Some post-secondary education 16(32.00%) N/A
Married 11(22.00%) N/A
1 2 3
Unmarried, but living with partner 29(58.00%) N/A
Single 10(20.00%) N/A
$10,000-$19,999/year 19(38.00%) N/A
$20,000-$39,999/year 12(24.00%) N/A
$40,000 or more/year 19(38.00%) N/A
Full-time 15(30.00%) N/A
Part-time 7(14.00%) N/A
School 4(8.00%) N/A
Not working 24(48.00%) N/A
No medical care 9(18.00%) N/A
No dental care 14(28.00%) N/A
During this program, parents were constantly engaged and received communication in order to track their mindset changes.
Figure 2.- Parent Response to Financial Incentives After 14 Weeks of the Program
The methodology for measuring parental response is adopted from the PARI (Parental Attitude Research Instrument). The model was estimated by maximum
Table 2.- Overview of Effects on
likelihood factors and categorical deciders. A parent who comes out with a larger value believes in the importance of education more (Moon, 2014).
Attitudes and Actions of Parents
Variable Age Conditional Effect Size Asymptotic p-values) Permutation Single p-val
Home Environment 5 0.333 0.003 0.04
Overall Parent Attitude 5 0.288 0.012 0.05
Home Safety 5 0.346 0.004 0.04
Home Environment 6 0.299 0.010 0.03
Overall Parent Attitude 6 0.003 0.012 0.06
Home Safety 6 0.305 0.011 0.05
Home Environment 7 0.276 0.013 0.03
Overall Parent Attitude 7 0.371 0.003 0.05
Home Safety 7 0.289 0.013 0.04
Home Environment 8 0.264 0.014 0.02
Overall Parent Attitude 8 0.333 0.001 0.06
Home Safety 8 0.278 0.015 0.03
Close administering of the children and parents ed for the control group. (ii) Difference in Means:
allowed us to collect data of parental interference Difference between the averages of the treatment
with their children's education after the program. A and control group. (iii) P-value: average values for
variety of data was collected surrounding different the two genders being equal. variables. (i) The control mean: average value record-
Table 3.- Parental Interference with Student Education as a Result of the Program
Variable Age Control Mean Difference in Means P-value
Mutual Cognitive Activities 5 33.575 32.478 0.942
Mutual Non-Cognitive Activities 5 379.144 68.385 0.045
No Interaction Between Parent and Child 5 -453.922 -76.341 0.063
Mutual Cognitive Activities 6 35.208 39.410 0.603
Mutual Non-Cognitive Activities 6 350.019 63.590 0.011
No Interaction Between Parent and Child 6 -658.820 51.834 0.065
Mutual Cognitive Activities 7 42.035 84.457 0.011
Mutual Non-Cognitive Activities 7 339.616 150.324 0.414
No Interaction Between Parent and Child 7 -857.624 160.408 0.397
Mutual Cognitive Activities 8 70.687 25.677 0.232
Mutual Non-Cognitive Activities 8 480.1535 66.485 0.075
No Interaction Between Parent and Child 8 -702.74 59.786 0.039
11. Results and Discussion 11.1 Cognitive
Table 4.- Program Effects on Cognitive Ability of Students
Variable Age Treated SD Permutation P-value
SD of Improvement in Math 5-6 0.360 0.273
SD of Improvement in Reading 5-6 0.268 0.214
SD of Improvement in Intelligence Tests 5-6 0.232 0.1348
SD of Improvement in Math 7-8 0.230 0.292
SD of Improvement in Reading 7-8 0.111 0.342
SD of Improvement in Intelligence Tests 7-8 0.124 0.329
Based on math achievement scores, Murnane, Wil-let and Levy (1995) report that students who score one SD higher earn on average 9.3% higher when they reach the working age. Based on reading achievement scores, Currie and Thomas (1999) report that students who score in the upper quartile of reading exams earn 20% more than those in the bottom quartile, which is associated with 8.0% increase in earnings per one SD of increase in testing performance (Additionally, Neal and Johnson (1996) estimated similar effects of student scores using the National Longitudinal Survey of Youth and Armed Forces Qualification Tests. Students were reported to receive 20% higher earnings per one SD increase). Reasons for the percentage gap could be (i) students who took the AFQT exam were older, (ii) Results reported by Currie and Thomas discover certain mean regression in exam results. Therefore, it is more reasonable to associate one SD increase with an 8.0% increase in earnings. We take into consideration the latest predictions on real earning growth in section 5.c.
Table 5.- Program Effects on I
There are certain caveats to our estimated results. First, our calculations do not account for fringe benefits, which could cause estimated benefits to be understated by around 33.3%. Second, the effect of testing results on earnings have the potential to change in the future, which might require a different implementation ofthe equations. Third, there is ambiguity in the growth of earnings in the future, although an argument can be made for the reliability of wage growth predictions.
From table (Table 4), we see significant improvement in the cognitive abilities of treated students in the program. It is important to note the short time frame of the program conducted. With 14 weeks of the program, children ages 5-6 have achieved 0.36 and 0.268 SD of increase in reading and math, respectively. Children ages 7-8 have achieved 0.230 and 0.111 SD of increase in reading and math, respectively. It can be argued that the scope of these cognitive improvements alone is sufficient to prove the effectiveness of the program.
11.2 Non-cognitive
n-Cognitive Ability of Students
Variable Age Treated SD Permutation P-value
1 2 3 4
SD of Improvement in Social Science Test 5 0.230 0.273
SD of Improvement in Critical Thinking Test 5 0.361 0.094
SD of Improvement in Social Awareness Test 5 0.199 0.02=633
SD of Improvement in Social Science Test 6 0.460 0.205
SD of Improvement in Critical Thinking Test 6 0.240 0.0774
SD of Improvement in Social Awareness Test 6 0.132 0.0350
1 2 3 4
SD of Improvement in Social Science Test 7 0.382 0.413
SD of Improvement in Critical Thinking Test 7 0.380 0.0513
SD of Improvement in Social Awareness Test 7 0.122 0.0848
SD of Improvement in Social Science Test 8 0.340 0.344
SD of Improvement in Critical Thinking Test 8 0.202 0.0561
SD of Improvement in Social Awareness Test 8 0.281 0.198
The importance of cognitive abilities is held at high regard, while there is significantly less focus on non-cognitive improvement. In order to better predict success in life, other factors should be considered as well (character skills, character traits, social awareness, and so on. It has also been suggested by economists that in order to improve the educational system, a reevaluation of which skills matter in life and when to form them is necessary. Cognitive abilities and non-cognitive abilities are reflected the best when improved together in terms of socioeconomic status (Heckman, 2010). Evidence arises from re-
cent studies on the GED's economic efficacy. It was reported that students who did not graduate from high school but received a GED were not as successful as those who did graduate from high school. Interestingly, this was more caused by both cognition and character than just one factor. Economic success for both individuals and society is motivated by a combination of the factors described above, which the program studied in this paper considers.
The chart below completed by Conti and Heckman (2010) demonstrates considerations for these factors.
M-Males, F=Females
Figure 3. Disparities by Education (Post-compulsory Education)
We see that factors in early life accounts for 50% The program conducted in this paper show sig-or more of the disparities (wages, health, obesity, de- nificant results on non-cognitive improvement as pression, smoking) in adulthood. shown by table (need to fill). Achievement tests
show that cognitive skills are boosted by development of the non-cognitive (30-40 percent of the variation in test results across students is due to character skills and not academic ability (Borghans et al. [2; 3], which is demonstrated by the program tested in this paper.
The effects of non-cognitive improvement are both statistically and economically significant.
For females, enhancements in academic motivation causes treatment effect to rise by 30 percent.
Our estimated value suggest a significant difference in earnings and wage growth between those treated by the program and those who are untreated. As recorded by the table, the treated children are predicted to earn significantly more throughout all stages of adulthood compared to the untreated, and the difference peaks in late adulthood. Treated children are estimated to earn $98,776 more in late adulthood than the untreated children, with a wage growth rate at 8.66%. The permutation p-value for the estimated earnings and rate of growth, respectively, are 0.1451 and 0.0119.
11.4 Effects on Education with Financial Investment
All policies are financed by a flat income tax rate designed such that the government budget is balanced every period and all effects are evaluated in the new long run steady state. We assume that self-productivity increases with the developmental stages. Literature on parental investments recognize the correlation between parental skills and student skills and established that they stabilize from age 1 on (Cunha [17]).
Through our results, we conclude that the impact of parental investments is larger during the first developmental stage of children, yet the effects last for a relatively short amount of time.
This estimate is statistically significant (p = 0.057). For CAT scores at a teenage age, little can be said about academic motivation (p = 0.161 and 0.528).
Outcomes in the labor market illustrate that 18 percent of the treatment effects on earnings per month at the middle age (p = 0.089) and also about 21 percent of the treatment effect on the chances of employment (p = 0.085) can be explained by early improvements in outside behavior.
11.3 Economic Success in Adulthood
Further data was collected shortly after financial investment was received by the parents to track how they used the rewards.
Table 6.- Demographic Sample: (S = 50)
Characteristic Detected Result(S%)
Books or Educational Material 23.3%
Utilities 20.8%
Health/Medicine 13.1%
Food 13.0%
Clothing for Children 9.8%
Savings 7.9%
Items for Home 6.7%
Other 5.4%
12. Conclusion
Utilizing experimental data from an educational program for young children, we examine treatment effects while modeling the rate of returns and benefits. By examining effects of cognitive and non-cognitive education while also testing potential modifications to the program, we discuss and make conclusions on treatment effects, the present value of costs and benefits, financial investments and incentives, and economic success in adulthood. The young age of candidates
Variable Untreated Mean Treated Mean AIPW Estimate P-value
Estimated Earnings in Early Adulthood 78.326 114.345 11.967 0.3777
Estimated Earnings in Mid Adulthood 81.646 135.450 74.874 0.03432
Estimated Earnings in Late Adulthood 219.950 318.726 81.531 0.1451
Rate of Growth of Earnings -0.0612 0.0866 0.8300 0.0719
is significant in the success of the program. One reason why is that behavioral problems are more of a concern with older children. We present the benefits that arise from an organized and concise educational program for young children and show how they will potentially increase success in life. We show that the development of both cognitive and non-cognitive skills is both achievable and optimal for socially and economically disadvantaged children and families. Therefore, we argue that public policies ought to direct attention at the implementation of programs that reduce poverty and increase the ability of children.
Specifically, federal and local policies aimed at improving the educational system should both focus on improvement of cognitive ability as well as personality and social awareness. Programs that target cognitive development can be directly evaluated via their performance, which results in clear benefits for children at a young age and later in adulthood. Current policy-making structures lack proper methods and resources to support these areas of development for the disadvantaged. We show how investing in early-childhood development prepares children for successful adult lives. Both policy makers and those who are involved in the education process should (i) look to considerable investment in early-childhood educational programs for children ages five to eight (Both short-term and long-term programs provide considerable benefits. The short-term program tested in this paper can be best implemented as an education resource during times where regular school does not occur) (ii) Incorporate both cognitive and noncognitive curricula consistently, (iii) Incorporate the results of non-cognitive development in data collection and analysis of educational programs, (iv) be aware that outside motivators for parents and students are beneficial, yet the way they are used are difficult to control and may not benefit the families the way researchers predict.
13. Additional Appendix
13.1 Figure: Human Capital Earnings Function
Recent evidence on the shape of the F function and the performance of a specification displays age profile and earnings realistically. Differences between fitted and actual data suggest that age-earnings profiles for US men and women are rather smooth (well-approximated by a simple variant of the human capital earnings function).
Figure 4. Age profiles of hourly wages for men (a) and women (b).
13.2 Earning Profiles
The Average Earning and Age Function below is interpreted from data collected by PSID.
Figure 5. Average Earnings as a Function of Age in 2021
Year Qtrl Qtr2 Qtr3 Qtr4
2011 755 753 753 764
2012 769 771 758 775
2013 773 776 771 786
2014 796 780 790 799
2015 80S 801 303 825
201G 330 824 327 849
2017 865 859 859 857
201S 331 876 387 900
2019 905 908 919 936
2020 957 1002 994 984
2021 939 990
Figure 6. Median Usual Weekly Earnings of Full-time Wage and Salary Workers by Age
Figure 7. Median Usual Weekly Earnings of Full-time Wage and Salary Workers by Sex
13.3 Additional Material on Non-Cognitive Treatment
Detailed Explanation and Definition ofAttributes in Personality Psychology
Table 7. - Big Five Domains and Facts
Big Five Personality Factor American Psychology Association Dictionary' Description Facets (and correlated skill adjective) Related Skills Analogous Childhood Temperament Skills
Conscientiousness The tendency to be organized, responsible, and hardworking Competence (efficient), Order (organized). Dutifulness (not careless). Achievement striving (ambitious), Self-discipline (not lazy), and Deliberation (not impulsive) Grit, Perseverance, Delay of gratification, Impulse control. Achievement striving. Ambition, and Work ethic Attention/(lack of) dis-tractibility, Effortful control. Impulse control/ delay of gratification. Persistence, Activity*
Openness to Experience The tendency' to be open to new aesthetic, cultural, or intellectual experiences Fantasy (imaginative), Aesthetic (artistic), Feelings (excitable), Actions (wide interests), Ideas (curious), and Values (unconventional) Sensory sensitivity, Pleasure in low-intensity activities, Curiosity
Extraversion An orientation of one's interests and energies toward the outer world of people and things rather than the inner world of subjective experience; characterized by positive affect and sociability Warmth (friendly), Gre-gariousness (sociable), As-sertiveness (self-confident), Activity (energetic). Excitement seeking (adventurous), and Positive emotions (enthusiastic) Surgency, Social dominance, Social vitality, Sensation seeking, Shyness*, Activity*, Positive emotionality and Sociability/affiliation
Agreeable-ness The tendency' to act in a cooperative, unselfish manner Trust (forgiving), Straightforwardness (not demanding), Altruism (warm), Compliance (not stubborn). Modesty (not show-off), and Tender-mindedness (sympathetic) Empathy, Perspective taking. Cooperation, and Competitiveness Irritability*, Aggressiveness, and Willful- ness
Neu-roticism/ Emotional Stability J Emotional stability is «Predictability and consistency in emotional reactions, with absence of rapid mood changes.» Neuroticism is «a chronic level of emotional instability and proneness to psychological distress» Anxiety (worrying), Hostility (irritable), Depression (not contented), Self-consciousness (shy). Impulsiveness (moody), Vulnerability to stress (not self-confident) Internal versus External, Locus of control, Core self evaluation, Self-esteem, Self-efficacy, Optimism, and Axis I psychopathologies (mental disorders) including depression and anxiety disorders Fearfulness /behavioral inhibition, Shyness*, Irritability'*, Frustration, (Lack of) soothability. Sadness
Notes: 'These temperament attributes may be related to two Big Five factors. Facets specified by the NEO-PI-R personality inventory (Costa and McCrae, 1992). Adjectives in parentheses from the Adjective Check List (Gough and Heilbrun, 1983).
Source: Table adapted from John and Srivastava (1999).
References:
1. Bettencourt Amie F. et al. "Evaluating Implementation Fidelity of a School-Based Parenting Program for Low-Income Families." The Journal of School Nursing,- Vol. 35.- No. 5. 2018.- P. 325-336. URL: https://doi.org/10.1177/1059840518786995
2. Borghans Lex et al. "Identification Problems in Personality Psychology." 2011. URL: https://doi. org/10.3386/w16917
3. Borghans Lex, et al. "The Importance of Intrinsic and Extrinsic Motivation for Measuring IQ." Economics of Education Review,- Vol. 34. 2013.- P. 17-28. URL: https://doi.org/10.1016/j.econe-durev.2013.01.008
4. Gross Deborah, Amie F. Bettencourt., Breitenstein S. "Financial Incentives for Promoting Participation in a School-Based Parenting Program in Low-Income Communities." Prevention Science.- Vol. 20.-No. 4. 2019.- P. 585-597. URL: https://doi.org/10.1007/s11121-019-0977-y
5. Hartley HarryJ. "Book Review:Jesse Burkhead. Thomas G. Fox, John W. Holland, Input and out Put in Large-City High Schools (Syracuse: Syracuse University Press, 1967).- 110 p." Urban Education,- Vol. 3.- No. 3. 1968.- P. 181-182. URL: https://doi.org/10.1177/004208596800300308. "Contents." Journal ofStatistical Planning and Inference,- Vol. 176. 2016.- p. iii. URL: https://doi.org/10.1016/s0378-3758(16)30033-7
6. Tinbergen J. "Input-Output Analysis in Education." Economics of Education, 1987.- P. 336-338. URL: https://doi.org/10.1016/b978-0-08-033379-3.50070-x
7. "Analyzing Social Experiments as Implemented: A Reexamination of the Evidence from the Highscope Perry Preschool Program." Quantitative Economics,- Vol. 1.- No. 1. 2010.- P. 1-46. URL: https://doi. org/10.3982/qe8
8. Cohn R. A., Lewellen W. G., Lease R. C. and Schlarbaum G. G. "Individual Investor Risk Aversion and Investment Portfolio Composition". Journal of Finance,- Vol. 30.- No. 2. 1975.- P. 605-620. URL: https://doi.org/10.3982/qe8
9. Jencks Christopher and Marsha Brown. "Effects ofHigh Schools on Their Students." Harvard Educational Review,- Vol. 45.- No. 3. 1975.- P. 273-324. URL: https://doi.org/10.17763/haer.453.d06n480616v9567g
10. "Housing Instability and Food Insecurity as Barriers to Health Care among Low-Income Americans." Journal of General Internal Medicine,- Vol. 21.- No. 1. 2006. URL: https://doi.org/10.1111/ j.15251497.2005.0021010013.x
11. Lemieux Thomas. "The 'Mincer Equation' Thirty Years after Schooling, Experience, and Earnings." Jacob Mincer a Pioneer of Modern Labor Economics,- P. 127-145. URL: https://doi.org/10.1007/0-387-29175-x 11
12. Mullainathan Sendhil and Eldar Shafir. "Freeing up Intelligence." Scientific American Mind,- Vol. 25.-No. 1. 2013.- P. 58-63. URL: https://doi.org/10.1038/scientificamericanmind0114-58
13. Nickow Andre et al. "The Impressive Effects of Tutoring on Prek-12 Learning: A Systematic Review and Meta-Analysis of the Experimental Evidence." 2020. URL: https://doi.org/10.3386/w27476. "Panel Study of Income Dynamics Competition (PSID)."
14. NSF. URL: https://www.nsf.gov/pubs/2020/nsf20573/nsf20573.htm.
15. Tejasvee Sanjay et al. "Digital Learning: A Proficient Digital Learning Technology beyond to Classroom and Traditional Learning." Advances in Information Communication Technology and Computing, 2020.- P. 303-312. URL: https://doi.org/10.1007/978-981-15-5421-6 31
16. Fowler William J. and Herbert J. Walberg. "School Size, Characteristics, and Outcomes." Educational Evaluation and Policy Analysis,- Vol. 13.- No. 2. 1991.- P. 189-202. URL: https://doi. org/10.3102/01623737013002189
17. Chung E. Y. Romano J. P. Asymptotically valid and exact permutation tests based on two-sample U-statistics. Journal of Statistical Planning and Inference,- 168. 2016.- P. 97-105. URL: https://doi. org/10.1016/j.jspi.2015.07.004
