Reducing Child Poverty: Comparing In Kind and Cash Transfer Programs Across the OECD
Kurtis Mikal ‘Llewellyn’ Poole
2023-04-03
Abstract
Social spending is a complicated topic that involves good policy and good execution. The target for many social spending policies are families with children, with an aim to reduce childhood poverty. American Research argues the validity of these programs’ existence, rather than the comparative advantages of the different policies. This paper analyzes social benefits for families across 38 different countries, to compare the correlative effects of In Kind and In Cash benefits spending. My findings show that, for the purposes of reducing child poverty, In Kind spending had no statistically significant effect on poverty and only In Cash spending had significant and substantial effects on childhood poverty.
Section 1: Introduction
Social benefits programs are are implemented for the benefit of impoverished children, to pull them out of poverty. America has had a troubled relationship with social benefits programs for the last 100 years, made much worse by the disproven myth of “Social Disincentives,” which say that an adult with social benefits will become reliant on them. This has restricted the number and funding of social programs, particularly those focused on families. Social benefits programs in total involve easing the financial burden on parents to raise their children out of poverty, and in total have been proven consistently to be effective. These social benefits programs can be split up in In Kind and In Cash benefits. In Kind benefits are for services or funds that are limited: Food Stamps can only be used for certain foods, Medicare is only for medical check up, these kinds of benefits. In Cash benefits are funds that are unrestricted and granted either through unrestricted tax credits or through child allowance programs.
Though much research in America has been done arguing the for the efficacy of Total social benefits, I noticed a gap in research for comparing In Kind and In Cash benefits. I wanted to compare the United States to other similar countries, so I started with selecting an international organization comprised of similar countries developmentally. This led me to using the 38 countries in the Organization for Economic Cooperation and Development (OECD). This organization documents and fact checks tons of statistics on its member countries, which I used to acquire both Child Poverty statistics as well as Social Family Benefits Spending statistics.
The main models I used to prove a meaningful correlation were fixed effect panel regressions, measuring how much Child Poverty changes based on increases in social benefits spending. I controlled for GDP and fixed the regressions based on country, to accurately measure the effects of an individual country’s spending and poverty rates. These models reveal that In Kind social benefits have no real effects on Child Poverty, while In Cash benefits are significantly correlated to reducing Child Poverty. I interpret this to mean that, for purposes of reducing poverty, countries should focus more on In Cash social benefits programs and channel their funding into those programs.
Section 2: Literature Review
The article that started my research was a 2022 paper called “Children and the US Social Safety Net: Balancing Disincentives for Adults and Benefits for Children” which, as the name implies, discusses social benefit spending and its effects on the target demographic of children (Aizer). It also discusses the topic of the potential social incentives of social benefit spending, and how the positives far outweigh the negatives. Many other American papers have been written on this topic, most of which focus on the merit of social spending in any capacity, with most recent academic papers supporting social spending as a way of reducing childhood poverty.
These papers have also proven that not only does it reduce child poverty, but after about twenty years these programs actually begin to pay for themselves: typically making two dollars for every one dollar invested in to these programs. One paper found that Medicaid has a 7% annual return on initial invest, paying for itself and making profit after 15 years (Goodman-Bacon, Andrew.) A gap I noticed in all of these papers, however, was a discussion on which was more effective: In Cash or In Kind, which directed my research.
Section 3: Data
Summary Statistics
##
## =============================================
## Statistic N Mean St. Dev. Min Max
## ---------------------------------------------
## Poverty 390 0.133 0.057 0.027 0.294
## Total Benefits 390 2.103 0.887 0.328 3.921
## Cash Benefits 390 1.217 0.592 0.057 2.919
## Kind Benefits 390 0.886 0.529 0.018 2.471
## ---------------------------------------------All of my data has come directly from the OECD website. The main variables used are Child Poverty, GDP, and Family Benefits Public Spending split up into In Cash and In Kind spending.Family Benefits Public Spending is measured as a percentage of GDP for each country, and was split up into In Cash and In Kind by the OECD (OECD (2023a)). Family Benefits Public Spending is also referring exclusively to to financial support targeted towards families and children, and other public spending such as housing and health are not included in this spending (OECD (2023b)). Child Poverty is measured as a percentage of the population aged 0-17 “whose income falls below the poverty line taken as half the median household income of the total population.” GDP is measured as US Dollars per Capita (OECD (2023c)). This data is collected from 38 different countries in the OECD over 15 years, from 2006 to 2020.
Public spending increased over COVID to help with families displaced by the pandemic, but since the long term effects of this spending has yet to be fully reported I decided to not include years after 2020. Some short term effects have been felt, however, that I will be discussing in the conclusion later.
Section 4: Results
Exploratory Analysis
These results are not surprising, but they’re important to demonstrate nonetheless. As Total Benefits spending increases, Childhood Poverty decreases, which validates the claim that its main purpose is to reduce Childhood Poverty. Giving or saving families money makes intuitive sense as to why it reduces poverty, so these results are not surprising.
In Kind Benefits also results in a negative correlation, though there is a significant amount of spread in the lower benefits spending range that can only partially be explained by country grouping not being represented in the line. This hints at a weaker relationship between In Kind and Child Poverty, but without looking at regressions it’s hard to say purely off visuals.
Cash Benefits have a much more even spread across the line, and have a much stronger visual negative correlation. This hints at Cash Benefits having a more measurable effect on Child Poverty, though again nothing can be said with certainty until the regression models later.
Correlation Analysis
## Poverty Total Benefits Cash Benefits Kind Benefits
## Poverty 1.000 -0.610 -0.500 -0.463
## Total Benefits -0.610 1.000 0.817 0.762
## Cash Benefits -0.500 0.817 1.000 0.249
## Kind Benefits -0.463 0.762 0.249 1.000The important correlations are between benefits spending and poverty. This correlation matrix confirms the moderate negative correlation between spending and child poverty. Notably, Cash Benefits have a stronger correlation (at -0.5) than In Kind benefits (at -0.463), reaffirming at what we could visually interpret from the previous graphs.
Ordinary Least Squares Panel Regression Analysis
##
## Ordinary Least Squares
## ==========================================================================================================
## Dependent variable:
## -----------------------------------------------------------------------------------------
## Child Poverty
## (1) (2) (3) (4) (5) (6) (7) (8) (9)
## ----------------------------------------------------------------------------------------------------------
## Total Benefits -0.039*** -0.033*** -0.032***
## (0.003) (0.003) (0.003)
##
## In Cash Benefits -0.048*** -0.041*** -0.038***
## (0.004) (0.004) (0.004)
##
## In Kind Benefits -0.050*** -0.038*** -0.040***
## (0.005) (0.005) (0.005)
##
## GDP -0.039*** -0.048*** -0.052*** -0.060*** -0.048*** -0.060***
## (0.006) (0.007) (0.006) (0.007) (0.007) (0.007)
##
## Constant 0.215*** 0.609*** 0.669*** 0.191*** 0.724*** 0.780*** 0.177*** 0.668*** 0.763***
## (0.006) (0.063) (0.068) (0.006) (0.065) (0.071) (0.005) (0.070) (0.072)
##
## ----------------------------------------------------------------------------------------------------------
## Observations 390 390 390 390 390 390 390 390 390
## R2 0.372 0.431 0.463 0.250 0.362 0.384 0.214 0.305 0.372
## ==========================================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01
## Columns 3, 9, and 9 control for YearFor the first nine models I ran an Ordinary Least Squares models with no fixed effects for countries over time, and came across some intuitive results. Goodness of fit, measured with R Squared (R2), starts off mildly low and increases with each explanatory variable we add. Which means as we include the GDP and the Years our goodness of fit increases across all models, our model has less scatter. Every result is statistically highly significant, and all have moderate negative correlations with child poverty.
For these models, it seems that In Cash and In Kind benefits have very similar effects on childhood poverty: every 1% increase of public benefits spending reduces child poverty by about 4%. This is very high, as the mean poverty rate across our data set is 13%. That being said GDP still has a stronger effect on poverty, with every 1% increase in national GDP decreasing poverty by about 6%.
While all of the results (particularly in the third, sixth, and ninth columns) align with our previous intuitions and observations,these results are incomplete, as there is still a significant bias in that these models pool all the data points together.
Fixed Effect Panel Regression Analysis
##
## Fixed Effect
## ========================================================================================================
## Dependent variable:
## ---------------------------------------------------------------------------------------
## Child Poverty
## (1) (2) (3) (4) (5) (6) (7) (8) (9)
## --------------------------------------------------------------------------------------------------------
## Total Benefits -0.010*** -0.012*** -0.017***
## (0.003) (0.003) (0.003)
##
## In Cash Benefits -0.014*** -0.020*** -0.023***
## (0.004) (0.003) (0.003)
##
## In Kind Benefits -0.005 0.007 -0.010
## (0.008) (0.007) (0.008)
##
## GDP -0.036*** -0.062*** -0.042*** -0.061*** -0.035*** -0.046***
## (0.006) (0.012) (0.006) (0.012) (0.006) (0.012)
##
## --------------------------------------------------------------------------------------------------------
## Observations 390 390 390 390 390 390 390 390 390
## R2 0.033 0.136 0.273 0.041 0.174 0.291 0.001 0.091 0.201
## ========================================================================================================
## Note: *p<0.1; **p<0.05; ***p<0.01
## Columns 3, 9, and 9 control for YearFor the next nine models Fixed Effect regressions were run, grouping the countries’ data together, and once we account for this we get a much more accurate model. The first big difference between these two results is the results for In Kind benefits. In our Fixed Effect models the In Kind benefits lose all statistical significance and their negative correlation is much weaker, even at one point having a positive correlation which doesn’t make intuitive sense. And while Cash Benefits do lose some goodness of fit and negative correlation after being translated, they are both still significant, reinforcing the results of our previous models and graphs. In Cash benefits reduce poverty by 2.3% for every 1% increase in GDP spent, and while that is lower than in our first nine models it is still very impactful. GDP still has a strong and significant effect on childhood poverty, and it remains at about 6% as seen from our previous models.
[Write your findings from the above table here. Start with R-squared. For each coefficient value explain direction, magnitude, and statistical significance of the relationship between y and x variables. Make sure to discuss the economic significance of the results.]
Section 4: Conclusion
At a conceptual level, these results are not shocking. Family Benefits Spending is implemented directly for reducing child poverty, so it makes perfect sense that they would have a significant and negative correlation. The important part of this paper, being the comparison of In Kind and In Cash Family Benefits Spending, is shocking. My initial hypothesis of In Kind Benefits being more impactful was proven wrong, and it varies from being slightly less effective to being not even significantly effective. This indicates that Cash Benefits are the more impactful half of Family Benefits Spending.
This is important for the United States as our child poverty levels have been, on average, over a full standard deviation above the OECD’s mean child poverty rates. The United States also has the second lowest Total Family Benefits Spending in the OECD, only higher than Turkey (which has a nearly equal average child poverty rate), and that reflects on our higher child poverty rates. There are many reasons why this is higher, but one of these that I find important is the USA’s emphasis on In Kind Benefits over Cash Benefits in fear of a negative social incentive. A negative social incentive which, while it has been argued in other papers, I’d argue isn’t even relevant to the conversation. It shouldn’t matter if a parent might become mildly complacent if it means that their children get to sleep under a roof and eat real food. No child should know hunger, and yet the United States’ McCarthyistic fear of anything that could be perceived as socialism has left us in a severely worse position than our neighbors with regards to caring about some of our most vulnerable citizens.
Partially outside the scope of this paper is an increase in Total Social Spending done by the United States from 2019 to 2020, jumping from 18% to 23% of our GDP. This jump in Social Benefits can be seen in Child Poverty rates in the United states dropping from 21% to to 18% from 2019 to 2020, and dropping further to 13.7% in 2021, now only slightly above the standard deviation for Child Poverty. While this spending is not responsible exclusively for the drop, the moderate correlation already proven here is certainly one proven, direct factor. This data also is not split up by the OECD between In Cash and In Kind spending, so a determination on which is more effective cannot be made with these points. Additionally, Childhood Poverty increased from 13.7% in 2021 to 17% in January and 16.7% in February of 2022, potentially hinting at the sharp drop being only temporary.
Increasing our Family Benefits Spending is proven to not only decrease Child Poverty rates but also make money over time, and there are no real statistical arguments against increasing Family Benefits Spending. My research is intended to help advise on the direction of this spending, and it is my recommendation that policy makers focus on direct In Cash Benefits. This research is not about general social spending, nor does it try to prove or disprove the merits of In Cash or In Kind Benefits as concepts, it does prove that In Cash Benefits are more effective for Family Benefits Spending.
Citations
Aizer, Anna, et al. “Children and the US Social Safety Net: Balancing Disincentives for Adults and Benefits for Children.” Journal of Economic Perspectives, vol. 36, no. 2, 2022, pp. 149–174., https://doi.org/10.1257/jep.36.2.149.
Goodman-Bacon, Andrew. “The Long-Run Effects of Childhood Insurance Coverage: Medicaid Implementation, Adult Health, and Labor Market Outcomes.” American Economic Review, vol. 111, no. 8, 2021, pp. 2550–2593., https://doi.org/10.1257/aer.20171671.
OECD (2023a), Social benefits to households (indicator). doi: 10.1787/423105c6-en (Accessed on 10 April 2023)
OECD (2023b), Poverty rate (indicator). doi: 10.1787/0fe1315d-en (Accessed on 10 April 2023)
OECD (2023c), Gross domestic product (GDP) (indicator). doi: 10.1787/dc2f7aec-en (Accessed on 10 April 2023)