ArticlePDF Available

Does Owning a Home Make Us More Generous?

June 2023
Journal of Housing Research

DOI:10.2139/ssrn.4472886

Authors:

Xun Bian

University of North Texas

Feifei Zhu

We examine the multifaceted influence of homeownership on charitable giving through several channels: tax deductibility, household wealth, and mobility. We find that homeowners donate substantially more than renters, and tax deductibility, wealth, and mobility are important predictors of the owner-renter gap in donations. We also show that the owner-renter difference in donations cannot be fully explained by these three channels. After controlling for an extensive list of household characteristics and the three channels, homeowners still donate approximately 20% more than renters. Our results are robust to a variety of modeling and identification strategies as well as different measures of donations. Our study further reveals that the likelihood to donate correlates inversely with mobility but is insensitive to tax deductibility and wealth. In contrast, tax deductibility and wealth are important predictors of the size of contributions. Furthermore, we show the owner-renter difference in donations varies substantially by generation cohorts.

Household Fixed-Effects Model

…

Figures - uploaded by Xun Bian

Content may be subject to copyright.

Content uploaded by Xun Bian

Content may be subject to copyright.

Does Owning a Home Make Us More

Generous?

Xun Bian

G. Brint Ryan College of Business

University of North Texas

xun.bian@unt.edu

Feifei Zhu

Fannie Mae

feifei zhu@fanniemae.com

June 7, 2023

Contact Author. G. Brint Ryan College of Business, University of North Texas. 1155 Union Cir-

cle 311160, Denton, Texas #76203-5017. Email: xun.bian@unt.edu.

Acknowledgments: The views expressed are those of the authors and not necessarily those of Fannie

Mae or the Federal Housing Finance Agency. We are grateful to an anonymous referee and the Managing

Editor (Kimberly Goodwin), whose insights have markedly improved the paper. We would also like to thank

Peng Liang, Zhenguo (Len) Lin, and the participants at the 2022 American Real Estate Society (ARES) An-

nual Meeting and the Financial Management Association (FMA) 2022 Annual Meeting for helpful feedback.

All errors are, of course, our own.

Does Owning a Home Make Us More

Generous?

Abstract

We examine the multifaceted inﬂuence of homeownership on charitable giving through

several channels: tax deductibility, household wealth, and mobility. We ﬁnd that

homeowners donate substantially more than renters, and tax deductibility, wealth,

and mobility are important predictors of the owner-renter gap in donations. We also

show that the owner-renter diﬀerence in donations cannot be fully explained by these

three channels. After controlling for an extensive list of household characteristics and

the three channels, homeowners still donate approximately 20% more than renters.

Our results are robust to a variety of modeling and identiﬁcation strategies as well

as diﬀerent measures of donations. Our study further reveals that the likelihood to

donate correlates inversely with mobility but is insensitive to tax deductibility and

wealth. In contrast, tax deductibility and wealth are important predictors of the size

of contributions. Furthermore, we show the owner-renter diﬀerence in donations varies

substantially by generation cohorts.

Keywords: charitable giving, altruism, homeownership

JEL Codes: D10, D64, R20

1 Introduction

Charitable giving stems from our desire to help and care for each other and is an important

part of our social and economic life. Americans have a long, and growing, tradition of

giving. According to the Giving USA Foundation, donations made to U.S. charities have been

growing consistently and reached 449.64 billion in 2019.1The vast majority of donations,

an estimated

309.66 billion (just a little shy of 70%) come from individuals. Therefore,

it is of paramount importance to understand how the decisions to give are made at the

household level. In this study, we examine the impact of homeownership on charitable

giving. Owning a home is arguably one of the most important parts of the American Dream.

To most of us, homeownership symbolizes ﬁnancial success and provides a sense of stability.

As famously outlined in Maslow’s hierarchy of needs (Maslow (1943)), humans must ﬁrst

satisfy physiological needs (e.g. air, food, shelter, etc.) and safety needs (e.g. job security,

economic stability, etc.), before seeking to achieve higher-level needs, such as love, social

belonging, and esteem. By satisfying our basic need for shelter and being a crucial milestone

of economic stability, homeownership could motivate us to arise at the higher stages of

need and inspire us to be more caring and altruistic. Of course, homeownership also comes

with signiﬁcant costs. Housing expenditures, such as mortgage payments, property tax,

and insurance premiums, could impose budget constraints and possibly limit discretionary

spending such as charitable donations. Additionally, unpredictable maintenance and repairs

may motivate people to increase precautionary savings and, as a result, give less. Does

owning a home make us more generous? If so, why? We attempt to answer these questions

in this study.

It is critical to note that homeownership is a complex social construct that is intertwined

with an array of demographic and economic variables. Therefore, without a doubt, the ef-

fect of homeownership on charitable giving is multifaceted. We ﬁrst look into three channels

1See https://givingusa.org/giving-usa-2020-charitable-giving-showed-solid-growth-climbing-to-449-64-

billion-in-2019-one-of-the-highest-years-for-giving-on-record/ (accessed on September 13, 2021)

through which homeownership may aﬀect charitable giving: tax deductibility,wealth, and

mobility. Owning a home brings about tax beneﬁts, promotes wealth accumulation, and

makes homeowners more likely to have stronger and long-term connections with their com-

munities. Each of those has been extensively studied as a consequence of homeownership.

However, the implications of such relations on charitable giving have been overlooked, and

we are the ﬁrst to ﬁll this gap. We show that homeowners donate substantially more than

renters, and a signiﬁcant amount of the diﬀerence is attributable to the eﬀects of homeown-

ership on tax deductibility, wealth, and mobility. We further inquire into the question: Are

homeowners more generous beyond these three reasons? The answer appears to be “yes”.

We ﬁnd that a substantial portion of the owner-renter gap on donations cannot be explained

by the three channels. To address the endogeneity surrounding the selection into homeown-

ership, we apply three modeling techniques: household ﬁxed-eﬀect models, propensity-score

matching (PSM), and matched-sample diﬀerence-in-diﬀerence (DiD) estimators. Our results

are robust to these modeling and identiﬁcation strategies. Our ﬁndings suggest that home-

ownership increases generosity, and such an eﬀect goes beyond what can be explained by tax

deductibility, household wealth, and mobility.

We further examine two distinct dimensions of charitable giving: the likelihood to donate

and the size of donations (conditional on being a donor). First, we ﬁnd that homeownership

signiﬁcantly boosts both the likelihood to donate and the size of donations. Additionally,

we ﬁnd that the likelihood to donate correlates inversely with mobility (measured by the

duration a household stayed in their community) but is rather insensitive to tax deductibil-

ity and household wealth. In contrast, tax deductibility and wealth play a crucial role in

predicting the size of donations. These results suggest that as opposed to how much to give,

which appears to be a decision inﬂuenced more heavily by ﬁnancial considerations, whether

or not to give exhibits a stronger tie to one’s connection to local communities. We further

look into giving behaviors of diﬀerent generational cohorts. We ﬁnd that for baby boomers,

owning a home increases giving to both religious and non-religious causes. In contrast, for

Generation X, such an eﬀect exists only for non-religious donations. Finally, we ﬁnd that

millennial homeowners do not give more than their renter counterparts. Our results suggest

that the eﬀect of homeownership on charitable giving is either age- or generation-dependent.

We are the ﬁrst to examine the eﬀect of homeownership on charitable giving by inquiring

into whether or not and why owning a home elicits more generosity. Our ﬁndings have impor-

tant public policy implications. In the United States, homeownership is heavily subsidized

through preferential tax treatments at the federal and state levels. For example, homeowners,

who itemize their tax returns, can deduct property tax and interest paid on home mortgages.

Additionally, capital gains from home sales are often exempt from taxation. Using data from

1990, Gyourko and Sinai (2003) estimate that tax beneﬁts granted to homeowners reached

almost

200 billion. A natural question to ask is: Why should homeowners be given such

substantial tax advantages over renters? What justiﬁes the subsidization of homeownership?

Public policies subsidizing homeownership may be defendable on the ground that owning a

home produces positive externalities, that is, social beneﬁts accrued to people other than

homeowners themselves. Previous studies, over the past two decades, show that homeown-

ership produces several kinds of external beneﬁts. First, there is evidence that living in

owner-occupied dwellings could be beneﬁcial to children. For example, children of home-

owners stay in school longer, and they have better cognitive abilities and fewer behavior

problems (Green and White (1997) and Haurin et al. (2002)). Furthermore, homeownership

beneﬁts neighbors and communities. Relative to renters, homeowners take better care of

their homes and produce a positive spillover eﬀect on neighborhood appearance and, as a re-

sult, the values of nearby properties (Galster (1983) and Harding et al. (2000)). Additionally,

homeowners are more involved community members and contribute more eﬀorts to improve

their neighborhoods (Rossi and Weber (1996) and DiPasquale and Glaeser (1999)). We join

this stream of literature by examining the external beneﬁts of homeownership through the

lens of charitable giving. Our study complements previous literature by identifying an addi-

tional avenue through which homeownership produces social beneﬁts. Our ﬁndings supply

another potential justiﬁcation for encouraging homeownership.

We also look into why homeowners donate more than renters. Intertwined with many

aspects of our social and economic life, homeownership’s inﬂuence on donation is multi-

dimensional. By separating the three motives to give, tax deductibility, household wealth,

and mobility, we hope to accomplish two important purposes. First, understanding how

much homeowners’ greater generosity can be attributed to the three channels is important

in its own right. Although tax treatments, wealth, and mobility are closely related to home-

ownership in the United States, such associations are by no means intrinsic, permanent,

and universal. For example, tax rules change over time and vary substantially from place

to place. Most recently, the Tax Cuts and Jobs Act (TCJA) of 2017 nearly doubled the

standard deduction and, as a result, substantially reduced tax advantages of homeownership

relative to renting. Furthermore, while U.S. homeowners receive considerable tax subsidies,

this is not the case in many other countries. Therefore, understanding the role tax incentives

play in aﬀecting donations enables us to forecast how giving behaviors may react to future

tax rule changes and comprehend how donations may vary geographically. Additionally, al-

though homeownership is shown to promote wealth accumulation, such an eﬀect is far from

guaranteed. There are periods during which homeowners saw their net worth decimated by

falling house prices. The 2007 ∼2009 subprime mortgage crisis is the latest example. During

this period, U.S. housing prices fell by nearly 30%, and U.S. household net worth declined

by nearly

13 trillion (20%) from its 2007 pre-crisis peak. Furthermore, the wealth growth

of homeowners varies tremendously across diﬀerent income and racial groups. For example,

low-income and minority homeowners experience much less wealth growth relative to their

high-income and white counterparts (Loving et al. (2012) and Newman and Holupka (2016)).

Therefore, examining the wealth channel provides insights into how giving behaviors could

be diﬀerent across time and demographic groups.

Second, controlling for tax deductibility, household wealth, and mobility and then ex-

amining whether or not homeowners still donate more enables us to measure the “core”

homeownership eﬀect, that is, the impact of homeownership on donation in addition to

these three channels. In the U.S., homeownership comes with “perks”. Owning a home

leads to tax beneﬁts, helps accumulate wealth, and makes homeowners more likely to bene-

ﬁt from improvements in their communities. For those reasons, it is perhaps not surprising

that homeowners donate more. First, the deductibility of mortgage interest and property tax

makes U.S. homeowners much more likely to itemize than renters. Because charitable giving

is also deductible under itemization, homeowners, on average, enjoy a lower after-tax cost of

donation than renters. Do homeowners give more simply because it is cheaper for them to

do good? If so, the association between owning a home and greater generosity is merely an

outcome of the tax rules rather than a true external beneﬁt resulting from homeownership.

Furthermore, owning a home promotes wealth accumulation, and the wealth eﬀect may in-

crease giving. However, if the owner-renter diﬀerence in donation is only attributable to their

wealth gap, then as far as motivating people to give, there is little diﬀerence between owning

a home versus holding any other assets that can appreciate in value (e.g. stocks, bonds,

cryptocurrencies, etc.). It is perhaps worth noting that in the United States, gains from

home sales are often exempt from the capital gains tax, an advantage enjoyed by few other

assets. Can such preferential tax treatment be justiﬁed if the wealth eﬀect fully explains the

owner-renter gap in donations? Finally, being less mobile, homeowners are stable residents

of their communities. With their greater ﬁnancial stakes in the community, more social

interactions with their neighbors, and stronger emotional attachments to where they live,

homeowners possess stronger incentives to help improve local communities. Homeowners’

greater generosity may be, at least partially, motivated by the fact that they are more likely

to beneﬁt from future improvements in their neighborhoods. Are homeowners more generous

only because they are more likely to be beneﬁciaries of their own generosity? To answer all

those questions, we must carefully control the three channels and examine the “core” eﬀect

of homeownership. After controlling for tax deductibility, household wealth, and mobility,

we ﬁnd homeowners still donate approximately 20% more than renters indicating a “core”

homeownership eﬀect on charitable giving.

We also contribute to the literature by carefully examining the possible causal eﬀect of

homeownership on donations. The choice of owning a home is not random, and the endo-

geneity issue surrounding who chooses to become a homeowner must be carefully addressed

before a causal inference can be made. We begin by estimating household ﬁxed-eﬀect models

controlling for a long list of economic and demographic characteristics as well as unobserved

household heterogeneity. We also conduct a propensity score matching (PSM) analysis,

which adjusts for distributional overlap in household characteristics. Our regression and

PSM analyses yield similar estimates of the homeownership eﬀect on donations. Moreover,

we apply matched-sample DiD estimators to examine the short time window around the

transitions into homeownership. We ﬁnd new homeowners start to give signiﬁcantly more

than their counterfactuals (i.e. renters who are otherwise similar) after switching from rent-

ing to owning. By applying a variety of identiﬁcation and modeling strategies, we obtain a

set of estimates that are largely consistent. Taken together, our ﬁndings suggest that the

positive correlation between homeownership and generosity is likely to be causal.

The remainder of the paper proceeds as follows: Section 2 reviews relevant previous re-

search, Section 3 provides an overview of the data used in our analysis, Section 4 describes

our empirical methodologies and presents our main ﬁndings, Section 5 extends our main anal-

ysis by exploring 1) religious versus non-religious donations and 2) generational diﬀerences

in donations. Section 6 oﬀers concluding remarks.

2 Related Literature

Our work relates to several streams of literature. First, we join the set of studies that

examine the determinants of charitable giving. Much of the attention has been focused on

the eﬀects of income and tax deductibility. In general, higher income and a reduced after-

tax price of donations lead to more giving (Feldstein and Clotfelter (1976) and Feldstein

and Taylor (1976)). Randolph (1995) points out that in response to the ﬂuctuation of

income and marginal tax rates, taxpayers adjust the timing of donations to reduce the after-

tax price of giving. Therefore, permanent and transitory income should be diﬀerentiated

when modeling the eﬀect of tax deductibility. Otherwise, the elasticity to the after-tax

price of donation would be overestimated. By separating the transitory and permanent

components of income, Auten et al. (2002) ﬁnd that permanent income changes have a

substantially larger impact on donations than transitory shocks. Additionally, previous

studies also ﬁnd that charitable giving correlates with many household demographic and

economic characteristics. For instance, donations appear to be positively associated with

age, education, and marriage. Furthermore, people of faith and households with children

give more (Glenday et al. (1986) and Schokkaert (2006)). Donations to diﬀerent causes

appear to be correlated. For example, Brown et al. (2012) ﬁnd that donations speciﬁcally

for victims of natural disasters are positively associated with other donations, with donations

to caring and needy organizations having the strongest correlation. We contribute to this

set of studies by showing that, in addition to determinants already identiﬁed in previous

literature, homeownership status also plays a critical role in shaping the decisions to give.

Furthermore, our results suggest that tax beneﬁts granted to homeowners have two separate

eﬀects on donations. First, as shown in previous studies, through the tax-deductibility

channel, tax breaks reduce the cost of donations for homeowners and motivate existing

homeowners to be more generous. Second, there could be an additional homeownership

channel, that is, preferential tax treatments motivating more people to become homeowners,

and homeownership elicits more generosity. We supply evidence that after controlling for

the after-tax cost of donation, the owner-renter diﬀerence in donation is still positive and

quite substantial.

Our study also contributes to the literature analyzing the consequences of homeownership.

This literature is vast and covers a wide range of areas.2Here, we only review previous

studies that are directly related to our analysis: the eﬀects of homeownership on 1) wealth

accumulation and 2) mobility and civic engagement. Owner-occupied housing is both a

consumption and an investment good, and house price appreciation boosts a household’s

wealth. It is widely believed that owning a home promotes wealth accumulation. Di et al.

(2007) analyze the impact of housing tenure choices on wealth accumulation, and they ﬁnd

that homeowners, especially those who owned their homes for long periods, enjoy a greater

increase in net worth. Along the same line, Turner and Luea (2009) ﬁnd that family net

worth is, in general, positively associated with the length of homeownership. The authors also

document that such a wealth-accumulation eﬀect varies by income level. Wealthy households

gain more from homeownership than their low- and moderate-income counterparts. Several

other studies conﬁrmed the income eﬀect and also ﬁnd a race eﬀect: White homeowners

experience more wealth increase than African-American homeowners (Loving et al. (2012),

and Newman and Holupka (2016)). We add to this literature by showing that house price

appreciation can aﬀect charitable giving through the wealth channel. By examining how

much of the owner-renter diﬀerence in charitable giving can be explained by the wealth gap

between owners and renters, we ﬁnd household wealth to be positively related to giving.

However, the wealth gap between owners and renters can only explain a rather small portion

of their diﬀerence in giving.

Empirical evidence overwhelmingly shows that relative to renters, homeowners are more

stable residents in their communities. As a result, homeowners are thought to be more en-

gaged members of their neighborhoods. Consistent with this notion, Rossi and Weber (1996)

2For a more comprehensive review of consequences of homeownership, see Dietz and Haurin (2003).

ﬁnd that homeowners are more likely to be members of community improvement groups. Di-

Pasquale and Glaeser (1999) show that homeowners are more likely to vote in local elections

and solve local problems. However, using an exogenous instrument based on a randomized

trial, Engelhardt et al. (2010) ﬁnd little evidence that switching from renting to owning

increases civic or neighborhood involvement. We look at civic engagement from a diﬀerent

angle through charitable giving. By inquiring into how much of homeowners’ generosity is

due to their longer, and most likely stronger, connections with their communities, we show

that staying in a community longer increases giving by boosting the likelihood to donate.

However, a longer residency seems to have a limited impact on how much people give. In

general, our ﬁndings are consistent with previous studies that ﬁnd homeownership increases

civic engagement.

3 Data

Our data comes from the Panel Study of Income Dynamics (PSID), a longitudinal survey

on a representative sample of U.S. households administrated by the Institute for Social

Research of the University of Michigan. The PSID survey was conducted annually between

1968 and 1997, and it was then shifted to a biennial frequency in 1999. For our analysis,

the PSID data possesses several important advantages. First, it is one of the few datasets

that contain information on charitable giving. Furthermore, in addition to information

on donations and homeownership status, the PSID also contains a wealth of household

demographic and economic characteristics. This is particularly important to us because to

identify the eﬀects of tax deductibility, wealth, and mobility, variables must be constructed

to measure those factors, and the PSID data provides the necessary ingredients for us to

do that. Additionally, given the endogenous nature of homeownership choices, it is critical

to control for relevant household characteristics to minimize potential omitted variable bias.

Thanks to the richness of the PSID data, we can include a long list of control variables in

our analysis. Another crucial advantage of the PSID is that it tracks the same household

over time, and this allows us to do two things. First, we are able to control for unobserved

household heterogeneity by estimating household ﬁxed-eﬀect models. Second, it also enables

us to zero in on the short time window during which a household switched from renting to

owning with our matched-sample diﬀerence-in-diﬀerence (DiD) analysis.

Our donation data comes from the 2003 ∼2019 PSID surveys. Households are asked how

much they donated in the last year towards the eleven diﬀerent charitable causes: religious,

combo, needy, health, education, youth, culture, community, environment, world peace,

and others.3Our dependent variable, donationi

t, is calculated as the summation of the

eleven categories of donation of household ireported on survey year t. PSID also collects

information on homeownership status by asking whether or not a household owns its current

residence. We use the indicator variable homeowneri

tto capture the homeownership status

of household ireported at year t.4

One of the goals we attempt to accomplish in our study is to examine the three diﬀerent

channels: tax deductibility, wealth, and mobility, through which homeownership may impact

donations. To do that, we must measure and control those factors in our models. As shown

by previous studies, charitable donations are sensitive to the after-tax price of donations.5In

the U.S., taxpayers who itemize can deduct charitable donations from their federal and state

income taxes. Therefore, for itemizers, the price of donation is approximately one minus

the household’s marginal tax rate. For non-itemizers, the price of donation is simply one,

3The PSID begins to collect information on donations in 2001. However, in the initial year, the dollar

values were reported only for the ﬁrst ﬁve categories. For the other six categories, the PSID only asked

whether or not donations were made, to which the respondents answered “yes” or “no”. Since 2003, dollar

values are reported on all eleven categories. For our donation variable to have a consistent interpretation

across time and also to better capture all donations, we choose 2003 to be the starting point of our analysis.

The 2019 survey is the latest PSID survey at the time when we conduct our research.

4PSID classiﬁes mobile home owners as homeowners. However, most mobile home owners pay rent for

their lots. Due to the idiosyncrasy of their ownership arrangement, we eliminate from our sample the small

fraction of mobile home owners.

5Following previous literature, we henceforth refer to the after-tax price of donations as the price of

donations.

because, without deduction, each dollar of donation costs one dollar. Two challenges must

be overcome in order to estimate a household’s price of donations. First, although the PSID

collects information on itemization, it contains no information about a household’s marginal

tax rate. Following Auten et al. (2002) and Brown et al. (2012), we use the National Bureau

of Economic Research (NBER) TAXSIM program to estimate the marginal federal and state

income tax rates. Speciﬁcally, we follow the method described in Kimberlin et al. (2014) to

estimate the input parameters (e.g. marital status, ages of taxpayers, number of dependents,

wage and salary, other income, itemization, etc.) from the PSID data, and we then feed the

estimated parameters into the TAXSIM program to obtain the estimated marginal federal

and state income tax rates for each household. Another issue is that the decision to itemize

is not exogenous and could be inﬂuenced by the level of charitable donation. For example,

some taxpayers would not want to itemize if they did not donate. In this case, charitable

donations could alter the decision to itemize. Clotfelter (1980) suggests that this endogeneity

issue could potentially bias the estimated price of donations. Following Clotfelter (1980),

Auten et al. (2002), and Brown et al. (2012), we do two things to address this issue. We

ﬁrst eliminate from our sample “endogenous itemizers” – deﬁned as those who would not

have chosen to itemize in the absence of charitable donation.6Second, when we estimate

marginal tax rates, we set charitable donation to be zero to obtain the “ﬁrst-dollar price” of

donation, which does not depend on the amount of charitable donation. Randolph (1995)

and Auten et al. (2002) suggest that permanent and transitory income may have distinct

eﬀects on donation. To separate permanent and transitory income, we follow Wilhelm et al.

(2008) and estimate the permanent component of family income of household iat year t,

permanentincomei

t, by averaging family income over the recent past (using up to four surveys

6Following Clotfelter (1980), Auten et al. (2002), and Brown et al. (2012), we deﬁne “endogenous item-

izers” as those who would not have chosen to itemize in the absence of charitable donation. To identify

“endogenous itemizers”, we begin with the sample of itemizers (as reported in the PSID). We ﬁrst feed

their information into the TAXSIM program to determine whether or not they should itemize. We then set

their charitable donations to zero and reestimate whether or not the household would have itemized without

charitable donations. A household is an “endogenous itemizer” if they reported being an itemizer and are

predicted to itemize, but only when donations are included among itemized deductions.

depending on whether or not the household was in the panel over that period). To account

for the possible impact of income variability, we also control for the standard deviation of

permanent income.

Wealth can be correlated with homeownership for at least two reasons: self-selection and

wealth accumulation. First, wealthy families may be more likely to become homeowners in

the ﬁrst place. The selection of wealthy families into homeownership implies the positive

relationship between homeownership and donation could be spurious. Additionally, as shown

by previous studies, owning a home promotes wealth accumulation. The wealth module of the

PSID contains extensive information on family assets and liabilities. Using that information,

the PSID calculates for each household their net worth. Speciﬁcally, this imputed value is

constructed as housing equity plus the summation of seven asset categories net of debt

(summation of eight debt categories).7We control for household net worth in some of our

speciﬁcations to disentangle the wealth eﬀect on donation.

There is a considerable body of literature focusing on the eﬀect of homeownership on

mobility. The consensus is that homeowners are less mobile than renters. Reduced mobility,

which could be equated with stability, provides incentives to improve one’s community.

Given a lot of donations may be made locally, we hypothesize that the positive association

between homeownership and mobility can boost donations. We employ two variables to

capture household mobility. First, we look at how many times a household reported moving

during the past six years and record it using the variable # of moves. Households with

more frequent relocations in the past may reasonably expect themselves to be more mobile

in the future. Additionally, we also construct the variable years since move, computed as

the number of years since a household’s most recent move. Staying in the same community

7Housing equity is calculated as house value minus the total outstanding balance of mortgages. The seven

asset categories are farm or business, checking and saving accounts, other real estate assets, stocks, vehicles,

annuity and IRA, and other assets. The eight debt categories are farm or business debt, other real estate

debt, credit card debt, student loans, medical debt, legal debt, loans from relatives, and other debt.

longer, a household is more likely to establish a stronger connection with neighbors, friends,

and local organizations and, as a result, more likely to give. To examine this channel, we

control for # of moves and years since move in some of our speciﬁcations.

Following previous literature, we also include in our models an array of economic and

demographic household characteristics that can inﬂuence donations. Speciﬁcally, we control

for household income, age of the household head (both the linear and the squared terms),

size of the household (measured as the number of persons), the number of children, sex of

the household head, employment status, marital status, race, levels of education, religious

preference, and self-reported health status. Table 1 presents the descriptive statistics of

these variables. Sample means and standard deviations are reported for our full sample as

well as the homeowner and renter subsamples. A simple mean comparison reveals that on

average, homeowners donate approximately four times more than renters on an annual basis

(

1,741 vs.

415). A similar pattern emerges when looking at religious and non-religious

donations separately. Furthermore, we deﬁne donori

tas a binary variable indicating whether

or not household imade a donation, regardless of the amount, during year t. Table 1

shows that homeowners are twice more likely than renters to be a donor in a given year.

Table 1 also reveals critical diﬀerences in household characteristics between owners and

renters. For example, homeowners, as expected, have a lower price of donations than renters.

Additionally, owning a home is associated with higher levels of income, greater net worth,

older age, higher education attainment, and lower mobility. Our empirical analysis will

control for these diﬀerences when estimating the eﬀect of homeownership on donations.

4 Empirical Analysis

4.1 Regression Analysis

We begin by estimating the following household ﬁxed-eﬀect model to examine the rela-

tionship between donation reported in survey t+ 1 and the homeownership status reported

in survey t. To help establish causality, we make sure our homeownership variable predates

the donation. Speciﬁcally, our regression model takes the following semi-log form:

ln(donationi

t+1) = β0+β1homeowneri

t+β2Zi

t+β3Yt+β4Si

t+β5Hi+εi

t(1)

where the natural logarithm of household donation reported in survey t+ 1 is regressed on

homeownership status, homeowneri

t, together with an array of control variables Zi

t.8In all of

our speciﬁcations, we include in Zi

tthe set of demographic and economic variables described

in the previous section. Additionally, to account for unobserved time-varying and state-

speciﬁc variations of charitable, we also control for survey-year ﬁxed eﬀects, Yt, and state

ﬁxed eﬀect, Si

t. Finally, there could be unobserved attributes that make certain households

more likely to self-select into homeownership and, at the same time, lead them to donate

more. To control for unobserved household-level heterogeneity, we include in our regression

models household ﬁxed eﬀects, Hi.

We estimate ﬁve speciﬁcations. In the ﬁrst one, we only include in Zi

thousehold de-

mographic and economic characteristics, survey-year, state, and household ﬁxed eﬀects but

exclude variables related to tax deductibility, wealth, and mobility. We then incrementally

add variables related to the three channels and further estimate three additional speciﬁca-

tions (Speciﬁcations [2] - [4]). Adding variables related to the three channels step-by-step

serves two important purposes. First, it allows us to study whether or not each of those

channels is an important determinant of charitable giving and its relative signiﬁcance in

8Donations are zeros for some household-year observations, and a logarithmic transformation would render

those undeﬁned. Following previous literature, we add

1 to the donation before taking the natural log.

explaining the owner-renter diﬀerence in donations. Furthermore, we are particularly inter-

ested in if homeownership status can still predict donations after-tax deductibility, wealth,

and mobility are controlled. If so, beyond the three indirect channels, ownership must pos-

sess some additional eﬀects on charitable giving. Additionally, we also estimate an additional

speciﬁcation (Speciﬁcation 5) looking at donation as a percentage of income. Redeﬁne our

dependent variable as

%donationi

t+1 =donationi

t+1

incomei

t+1

×100,(2)

we re-estimate our baseline regression to see whether or not homeowners give a larger fraction

of their income.

Table 2 presents regression coeﬃcients and t-values estimated from Equation (1). Survey-

year, state, and household ﬁxed eﬀects are included in all speciﬁcations but suppressed in

the table for brevity. Several observations can be made from Table 2. First, homeowneri

consistently exhibits a strong positive correlation with donation across our four speciﬁcations

(all signiﬁcant at the 1% level). However, its marginal eﬀect decreases substantially as we

add variables measuring tax deductibility, wealth, and mobility into the model. For example,

with only household characteristics plus survey-year, state, and household ﬁxed eﬀects being

controlled, our ﬁrst speciﬁcation suggests that ceteris paribus, homeowners, on average, do-

nate approximately 30% more than renters. As we add variables related to tax deductibility,

wealth, and mobility, the coeﬃcient on homeowneri

tgradually decreases to 0.213 in Spec-

iﬁcation 4, suggesting that after the three channels are controlled for, homeowners donate

just a little over 20% more than renters. Incrementally adding tax deductibility, wealth, and

mobility to our model reveals that these three channels can explain approximately 10 per-

centage points, or one-third, of the variation that were previously captured by homeowneri

Looking at % donation, we ﬁnd that although the point estimate suggests a positive rela-

tionship between homeownership and giving a larger fraction of household income, such a

relationship appears to be statistically insigniﬁcant. Therefore, there is no evidence that

homeowners give a larger fraction of their income. This new speciﬁcation also reveals some

other patterns. For example, the standard deviation of permanent income and wealth are

much stronger predictors of % donation than of ln(donation), suggesting that households

with more stable income and greater wealth donate a larger fraction of their income.

The estimated marginal eﬀects of tax deductibility, wealth, and mobility all have expected

signs. Congruent with previous literature, we ﬁnd that a reduced price of donations incen-

tivizes people to give more (signiﬁcant at the 1% level in Speciﬁcations [2], [3], and [4]).

Consistent with Auten et al. (2002), our point estimates suggest that the permanent compo-

nent of household income has a positive eﬀect on donation, and the variability of permanent

income, measured by its standard deviation, appears to reduce charitable giving. However,

neither of them is statistically signiﬁcant at conventional levels. As expected, wealthy house-

holds give more. Additionally, we ﬁnd years since move is an important predictor of the

amount of donation (signiﬁcant at the 1% level). Speciﬁcally, each additional year staying

in the same community increases donations by approximately 0.9%. Additionally, coeﬃ-

cients of our controlled background variables generally have expected signs. Consistent with

virtually all previous studies on charitable giving, higher income is associated with more

donations. The coeﬃcients on age and age2point to an inverted U-shaped relationship of

age on donations. We believe this pattern could be, at least in part, due to the positive eﬀect

of having school-age children on donation (discussed in the next paragraph). Middle-aged

households are more likely to have school-aged children. Consistent with this conjecture,

the eﬀect of age and age2becomes less signiﬁcant when we later control for the age of the

youngest child. Consistent with Glenday et al. (1986) and Schokkaert (2006), we also ﬁnd

that donations increase with the number of children. Furthermore, we ﬁnd married couples

donate more than their single, divorced, and separated counterparts. Finally, poor health

substantially reduces donations.

In our baseline regressions, We control for the number of children. It is possible that

children’s age may also be a crucial determinant of charitable giving. Unfortunately, our

dataset does not contain information about the ages of all children in a household. How-

ever, the PSID does survey families regarding the age of their youngest child. Using this

information, we estimate two additional speciﬁcations using the sample of households with

children. First, we control the age of the youngest child directly. We realize this speciﬁca-

tion forces a linear relationship between charitable giving and the age of the youngest child.

Therefore, we also estimate a more ﬂexible speciﬁcation by including in our regression three

age groups. Following the Centers for Disease Control and Prevention (CDC)’s deﬁnition

of child development milestones, we classify the youngest child of a household into 1) in-

fant/toddlers/preschoolers (age 0-5) (the reference group), 2) middle childhood (age 6-11),

and 3) young teens/teenagers (age 12-17).9Results from both speciﬁcations are reported

in Table A2. From the ﬁrst speciﬁcation, we observe that having charitable givings are,

generally speaking, positively associated with the age of the youngest child. The second

speciﬁcation further reveals that such an eﬀect is primarily driven by households whose

youngest child is in middle childhood (age 6-11).

4.2 The Likelihood to Donate and the Size of Donation

We further explore two distinct dimensions of charitable giving: the likelihood to donate

and the size of donation (conditional being a donor). Studying the likelihood to donate and

the size of donations separately is important and interesting for several reasons. First, while

the size of contributions may be constrained by disposable income, whether or not to donate

is more likely to be a free choice and possibly a better indicator of people’s willingness to help

others. A gift, no matter how small, is an act of kindness and caring. As a result, relative

to how much to give, the decision of whether or not to give may be less sensitive to ﬁnancial

considerations such as income, wealth, and tax deductibility. Second, the act of giving

9See https://www.cdc.gov/ncbddd/childdevelopment/positiveparenting/index.html (accessed on

January 22, 2023)

indicates awareness and, more importantly, a willingness to take action for a charitable

cause. It is possible that donors, in addition to giving money, are also more likely than

non-donors to engage in altruistic and socially conscious behaviors (e.g. volunteering and

recycling). In this case, holding the total amount of donation constant, having more people

donate, though each makes only a small contribution, would arguably be a more desirable

social outcome than having just a few very generous donors. For these reasons, we model

the likelihood to donate and the size of donations separately.

First, we want to know whether or not owning a home promotes the act of giving (re-

gardless of how much they give). In other words, relative to renters, are homeowners more

likely to be donors? We use the variable donori

tto indicate whether or not a household

donated, regardless of how much, in a particular year, and we model the likelihood to give

with a logit regression. Second, we are also interested in conditional on being a donor, do

homeowners donate more than renters? Because donors are by no means a random sample

of all households, we apply the two-stage Heckman sample selection model (Heckman (1974)

to correct for potential selection bias. Using our sample of all households, we estimate a

ﬁrst-stage selection model of whether or not a household donates a non-zero amount via a

probit regression. We then calculate the inverse Mills ratio (λi

t) from our selection model and

include that in our second-stage regression. To help with identiﬁcation, we exclude from our

second-stage regression two variables that are predictors of the likelihood of being a donor

but insigniﬁcantly associated with the size of donation: # of moves and years since move.

The rationale is that whether or not to donate depends more on a family’s connection with

their community. On the other hand, conditional on being a donor, how much to give is

more likely to be inﬂuenced by the price of donations and income.

The left-hand side panel of Table 3 reports results from our logit model examining the

likelihood to donate. Logit coeﬃcients and t-values are reported. The right-hand side panel

shows the results of our second-stage regression from the Heckman model, which examines

the size of donations. Several interesting observations can be made from Table 3. First, our

results indicate that the likelihood to donate and the size of donations are inﬂuenced by tax

deductibility and mobility in very diﬀerent ways. For example, price of donation is inﬂu-

ential to the size of donations (signiﬁcant at the 5% level) but insigniﬁcant in predicting the

likelihood to donate. On the other hand, years since move, which measures a household’s

duration of residency in their current community, is critical in determining the likelihood

to donate but insigniﬁcant in predicting the amount of contribution. Taken together, our

results suggest that how much to donate is more likely to be a decision made based on the

price of donations and income. On the other hand, whether or not to donate depends much

more on a household’s connection with their community. Furthermore, we also ﬁnd that even

after controlling for tax deductibility, wealth, and mobility, homeownership substantially in-

creases both the likelihood to donate and the size of contribution (both are signiﬁcant at the

1% level). Our estimated odds ratio suggests that owning a home raises the relative odds of

being a donor by 25.3%. Conditional being a donor, homeowners donate 16.6% more than

renters.

4.3 Propensity Score Matching (PSM) Analysis

Our regression analysis indicates that homeowners donate more than renters, and such

an eﬀect is beyond what can be fully explained by tax deductibility, wealth, and mobility.

Although we controlled for a long list of covariates and unobserved heterogeneities through

ﬁxed eﬀects, it is still possible that our regression models could suﬀer from potential model

misspeciﬁcation errors and omitted variable biases. Therefore, we also conduct a propensity

score matching (PSM) analysis, which we believe possesses several important advantages and

complements our regression analysis. First, there could be a lack of distributional overlap

between homeowners (the treated group) and renters (the control group) on observable char-

acteristics, and this can cause possible biases on the estimated marginal eﬀects. The PSM

approach can detect the lack of covariate distribution between our owner and renter samples

and adjust the distribution accordingly. Second, the PSM analysis provides an alternative

way to account for observable household heterogeneity and, as a result, mitigate potential

endogeneity concerns. By matching homeowners and renters along observable dimensions,

the PSM makes the variable of interest (i.e. homeownership) the only diﬀerence between

the treated and control groups (Rosenbaum and Rubin (1983)). In addition to accounting

for observable factors, several previous studies suggest that when an extensive number of

covariates are included, a PSM analysis minimizes the systematic variation in donations due

to unobservable diﬀerences between the two groups and, therefore, approximates a random-

ized control trial (Shadish et al. (2008), Holupka and Newman (2012), Newman and Holupka

(2016)).

We conduct our PSM analysis in several steps. We ﬁrst estimate the propensity score of

each household being a homeowner via the following logit model.

homeowneri

t=γ0+γ1Zi

t+γ2Si

t+ηi

t(3)

tincludes 1) all household economic and demographic variables we controlled in our regres-

sion models, 2) the variables related to tax deductibility, wealth, and mobility, and 3) state

ﬁxed eﬀects. Propensity scores are obtained by calculating the ﬁtted values from Equation

(2). Dehejia and Wahba (2002) suggest that relative to matching with replacement, matching

without replacement improves the precision of the estimates. We follow this recommendation

and match, without replacement, each homeowner household-year observation with a renter

household-year observation from the same year by applying the one-to-one nearest-neighbor

matching method. By applying a year-speciﬁc matching, our matching method accounts

for unobserved time-varying factors that aﬀect homeownership. To ensure only high-quality

matches are used for our estimation, we follow Austin (2009) and Strawinski (2013) and

impose a caliber of 1%, that is, we require the propensity scores for each matched pair to be

within 1.0% of each other. Our procedure produces a matched sample of 20,434 observations

(i.e. 10,217 owner-renter pairs).

We follow Rosenbaum and Rubin (1985) and Lee (2013) and conduct three diagnostic

tests to check if the overlapping and balancing conditions are met by our matched sample.

First, if our matching is successful, the treated and control groups should have similar distri-

butions of propensity scores, that is, they possess a similar estimated likelihood of becoming

homeowners. As shown in Table A3 in the appendix, the distributions of propensity scores of

the treated and control groups are almost identical to each other. Not only the treated and

control groups are very similar in their means and standard deviations, but the medians and

upper and lower quartiles are extremely close. Second, if our homeowners and their matched

renters are indeed counterfactuals of each other, we would expect the two groups to be, on

average, similar along observable dimensions. Therefore, we check if household characteris-

tics, on which our matching was based, are similar between the treated and control groups.

As reported in column [2] of Table A4 in the appendix, after the matching, the diﬀerences

in average household characteristics are very small and mostly statistically insigniﬁcant.10

Finally, we examine whether or not the variation of homeownership in our matched sample

can still be explained by controlled household characteristics. The rationale is that after

matching, the treated and control groups should be similar enough such that their diﬀerence

in homeownership choice is not attributable to observable characteristics. To ensure this

condition is satisﬁed, we re-estimate Equation (3) using our matched sample. As shown in

column [1] of Table A4 in the appendix, the vast majority of our household characteristics,

after matching, are statistically insigniﬁcant in explaining homeownership status. Addition-

ally, this regression yields a small pseudo R2of 0.6%, and a likelihood-ratio test produces

ap-value of 0.701. Both indicate that after matching, observable household characteristics

can no longer explain the choice of homeownership. Collectively, our diagnostic tests give us

conﬁdence that our matching is successful.

10Two-sample t-tests are conducted on continuous variables, and two-sample χ2-tests are conducted on

categorical variables.

We estimate the average treatment eﬀect by regressing ln(donationi

t+1) on homeowneri

using our matched sample controlling for household characteristics, survey-year, state, and

household ﬁxed eﬀects. The results of our PSM analysis are reported in Table 4. We yield an

estimated treatment eﬀect of 0.237 (signiﬁcant at the 1% level), suggesting that on average,

homeowners donate 23.7% more than their otherwise similar renter counterparts. Because

we included variables related to tax deductibility, wealth, and mobility when estimating

propensity scores, our treatment eﬀect estimate is comparable to the estimated regression

coeﬃcient when the three channels are controlled in the model (Speciﬁcation 4 in Table 2).

Comparing the two, we ﬁnd the estimated marginal eﬀect from our PSM analysis is very

similar to that produced by our household ﬁxed-eﬀect regression (0.237 vs. 0.213). We

further estimate the treatment eﬀects on the likelihood to donate and the size of donations

separately. For the likelihood to donate, we estimate a logit model, using our matched sam-

ple, regressing donori

t+2 on homeowneri

tand other household characteristics. For the size of

donations, we focus only on donors in our matched sample and regress ln(donation)i

t+2 on

homeowneri

tand other household characteristics. Estimated treatment eﬀects are positive

and highly signiﬁcant (at the 1% level) for both. Our PSM results suggest that homeown-

ership raises the relative odds of being a donor by 35.3%. Conditional on being a donor,

homeowners donate 12.1% more than renters.

4.4 Matched-Sample Diﬀerence-in-Diﬀerence Estimation

We further analyze the eﬀect of homeownership on charitable giving under a matched-

sample diﬀerence-in-diﬀerence (DiD) framework.11 Our matched-sample DiD analysis pos-

sesses several important advantages. First, while people self-select into homeownership, the

timing of becoming homeowners is arguably more random and, thus, less susceptible to po-

tential endogeneity problems. Second, by focusing on donations immediately after becoming

11Newman and Holupka (2016) applied a similar method to study the eﬀect of homeownership on wealth

accumulation. Additionally, Barker and Miller (2009) used the diﬀerence-in-diﬀerence method to study the

eﬀect of homeownership on child welfare.

homeowners, our DiD estimates capture the owner-renter diﬀerence on donations correlated

with the change of homeownership. In other words, our DiD estimators are immune to time-

invariant confounders and time-variant confounders that are group-invariant. Third, both

the wealth-accumulation eﬀect and the mobility eﬀect of homeownership take time to build.

By focusing on donations made at the initial stage of homeownership, our DiD estimators

remove the eﬀects from these two channels.

Similar to our PSM analysis, we deﬁned homeownership as the treatment. However,

instead of looking at all homeowners, we focused on households who switched from renting

to owning at year t. Speciﬁcally, our treated group includes observations reported being a

“renter”, “owner”, and “owner” respectively at time t−1, t, and t+ 1.

t−1t t + 1

Treated group Renter Owner Owner

Control group Renter Renter Renter

For each observation in our treatment group, we apply the propensity-score matching pro-

cedure described in Section 4.3 to match it with a similar renter counterpart from the same

year, who reported being a “renter” at time t−1, t, and t+ 1. Our matching procedure aims

to create a matched sample mimicking a randomized controlled trial in which a new home-

owner, who switched from renting to owning at t, is matched with a renter who possesses

a similar propensity to become a homeowner but did not make the switch. To ensure high

matching quality, we conduct the same diagnostic tests described in the previous section to

ensure the overlapping and balancing conditions are met.12 We then estimate the following

household panel regression using our matched sample:

ln(donationi

t) = δ1treatment ×posti

t+δ2Hi+δ3Yt+εi

t(4)

where treatment indicates whether or not an observation belongs to the treatment group,

12Results of our overlapping and balancing tests are available upon request.

and posti

tidentiﬁes the post-treatment period. Therefore, treatment ×posti

ttakes the value

of 1 if household ibecame homeowners at year tand stayed homeowners at year t+ 2. The

estimated δ1is our DiD estimator of the impact on charitable giving from the change of

ownership status (i.e. transitioning from owning to renting at time t). Additionally, we

also control for survey-year and household ﬁxed eﬀects. It is important to note that simple

standard DiD models often only control for treatment and post-treatment ﬁxed eﬀects, that

is, treatment and posti

tare included in Equation (3) in addition to their interaction term.

We are able to go beyond that and control for both survey-year and household ﬁxed eﬀects

because 1) in our treatment group, households became homeowners in diﬀerent years, and 2)

some households switched between the treated and the control groups. For example, a renter

household previously in our control group may later switch to owning and enter our treated

group. The standard treatment and post-treatment ﬁxed eﬀects combination is nested within

our survey-year and household ﬁxed eﬀects combination.13 By controlling for survey-year

and household ﬁxed eﬀects, we achieve two important purposes. First, by allowing average

donation to vary by survey year and household, our model is more ﬂexible and, as a result,

has a better goodness-of-ﬁt. Second, with time-varying and household-speciﬁc variations

being absorbed by survey-year and household ﬁxed eﬀects, our estimated homeownership

eﬀect is more conservative than that from the simple DiD speciﬁcation.

Results from our matched-sample DiD analysis are presented in Table 5. The estimated

coeﬃcient of T reatment ×P ost is 0.339, suggesting that switching from renting to owning

is associated with a 33.9% increase in donation (statistically at the 1% level). We further

examine the likelihood to donate and the size of donations using our matched-sample DiD

method. To model the likelihood to donate, we replace ln(donationi

t) in Equation (3) with

the indicator variable donori

tand estimate a linear probability model. To model the size of

donations (conditional on being a donor), we re-estimate Equation (3) using only donors in

13Gormley and Matsa (2011) used a similar approach to study corporate responses to the liability risk

from workers’ exposure to newly identiﬁed carcinogens. As a robustness check, we also estimated the simple

DiD speciﬁcation with treatment and post indicators. The results are similar.

our matched sample. Our results suggest that switching from renting to owning increases

the likelihood of being a donor by 3.6%, and conditional on being a donor, homeowners

donate 25.3% more than renters (both are signiﬁcant at the 1% level). In summary, results

from our matched-sample DiD analysis are consistent with our regression and PSM analyses.

Switching from renting to owning is associated with a signiﬁcant increase in giving after the

change, and such an increase is a result of both a greater likelihood to donate as well as an

increased size of contributions.

5 Further Explorations

5.1 Religious vs. Non-Religious Donations

We further explore several interesting dimensions of charitable giving and their relations

to homeownership. In this section, we separate religious and non-religious donations to see

if they are impacted by homeownership diﬀerently. Doing so helps further pinpoint the

channels through which homeownership aﬀects charitable giving. Religious donation is con-

sistently the largest category of charitable giving. According to the Giving USA Foundation,

religious congregations received the largest share, approximately 32%, of all charitable giv-

ing made to U.S. charities. The signiﬁcance of religious donations is even more evident in

our data when focusing only on donations made by households. During our sample period,

religious donations, on average accounts for 48% of total household giving each year. It is

possible that homeowners might just be more religious than renters, and, as a result, they

give more to religious causes. Although we control for religious preference, this may not fully

capture the intensity of religiosity. In this case, we could mistake religiosity for generosity. If

being more religious is the only reason why homeowners are more generous than renters, we

should observe a positive relationship only between homeownership and religious donations.

The association between homeownership and non-religious donations could be non-existent

or even negative. Being more religious may lead homeowners to donate less to other causes

if religious and non-religious donations are subject to the same budget constraint. Alter-

natively, if homeownership increases generosity in general, its impact should spread across

both religious and non-religious donations.

We re-estimate our household ﬁxed-eﬀect models separately for religious and non-religious

donations, and the results are reported in Table 6. We see that homeownership is associated

with greater levels of both religious and non-religious donations (both are signiﬁcant at the 1%

level), and the marginal eﬀects are quite similar between the two types (14.0% versus 13.4%).

Examining the coeﬃcients on variables related to tax deductibility, wealth, and mobility, we

ﬁnd the three channels appear to inﬂuence religious and non-religious donations similarly

except for the fact that religious donations appear to be more sensitive to the variability of

permanent income. We further conduct the propensity-score matching analysis and matched-

sample DiD analysis separately for religious and non-religious donations. The results are

respectively reported in Tables 7 and 8. Consistent with results from our regression analysis,

the estimated treatment eﬀects from our PSM method are both signiﬁcantly positive and

similar in their magnitudes (with the marginal eﬀect being slightly stronger on non-religious

donations). Our matched-sample DiD estimates are also in line with our regression and PSM

results. Switching from renting to owning is associated with respectively an 18.9% and 26.9%

increase in religious and non-religious donations (both are statistically at the 1% level).

We also take a more detailed look at non-religious donations. Instead of combining non-

religious donations into a single category, we re-estimate our baseline regression separately

for each donation category. Results are reported in Table A1 in the Appendix. Nine out of

ten categories of non-religious donations exhibit a positive relationship with homeownership

(as indicated by their positive coeﬃcients). Donations towards needy (signiﬁcant at the

1% level), culture (signiﬁcant at the 5% level), and environmental causes (signiﬁcant at the

5% level) exhibit statistically signiﬁcant associations with homeownership. In general, our

by-category analysis conﬁrms our ﬁnding that homeownership increases charitable donations.

5.2 The Eﬀect of Age/Generation on Owner-Renter Diﬀerence

The eﬀect of homeownership on donation could vary through one’s life cycle. In this

subsection, we examine the inﬂuence of homeownership on charitable giving for diﬀerent

generation cohorts. To deﬁne generation cohorts, we adopt the following deﬁnitions used by

Pew Research Center (Dimock (2019)).

Generation Cohort Deﬁnition

Boomers (and older) born on or before 1964

Generation X born between 1965 and 1980

Millennials (and younger) born on or after 1981

Based on the age of the household head, we stratify our full sample into three generation-

cohort subsamples. We then re-estimate our regression models separately for “boomers”,

“Generation X”, and “millennials”.14

Results are reported in Table 9. Looking at total donations, we ﬁnd a substantial owner-

renter gap in donations for the two older generation cohorts: boomers and Generation X,

and the estimated marginal eﬀects appear to be similar between these two cohorts. Specif-

ically, homeowners in the boomers and the Generation-X groups donate respectively 22.6%

and 24.9% more than renters from the same cohort (both are signiﬁcant at the 1% level).

However, millennial homeowners do not give signiﬁcantly more than their renter counter-

parts. Separating religious versus non-religious donations further reveals some interesting

diﬀerences between generation cohorts. First, our results suggest that relative to renters in

the same group, millennial homeowners do not give signiﬁcantly more to either religious or

non-religious causes. Furthermore, Generation-X homeowners donate 23.4% more to non-

religious causes (signiﬁcant at the 1% level). However, such an eﬀect is insigniﬁcant for

religious donations. In contrast, homeowners in the boomer group give more to both reli-

gious and non-religious causes. Relative to renters in the same age cohort, they make 22.4%

14Henceforth, we use the term “boomers” to mean boomers and older. Similarly, we use the term “mil-

lennials” to mean millennials and younger.

more religious donations and 13.5% more non-religious donations.

Results from our PSM analysis (reported in Table 10) are consistent with our regres-

sion results. The owner-renter gap in donations continues to be small and insigniﬁcant

for the millennial group. Both boomers and Generation-X homeowners donate more than

renters. However, Generation X homeowners’ greater generosity seems to be primarily di-

rected towards non-religious causes, while boomer homeowners give more to both religious

and non-religious causes. Turning to our DiD estimates (reported in Table 11), consistent

with results from our regression and PSM analysis, we ﬁnd no signiﬁcant diﬀerence in dona-

tion between millennial owners and renters. Furthermore, our DiD estimates also show that

relative to renters in the same cohort, boomers and Generation X homeowners both donate

more to non-religious causes. There are two possible interpretations of our ﬁndings. First,

what we document here could be a life-cycle eﬀect. As people age, their giving behaviors

will change, and younger generations will eventually behave more like older generations. For

example, millennial homeowners may not give more (relative to renters) now but may be-

come more generous as they age. Alternatively, we could have identiﬁed some generational

diﬀerences, that is, diﬀerent generation cohorts possess distinct tastes for charitable giving.

In this case, millennial homeowners may not give more even as they become older. In this

case, homeonwership could be losing its charm of eliciting generosity from millennials. Both

are plausible interpretations, and data on charitable giving spanning a generation’s entire

life cycle is needed to distinguish them empirically.

6 Concluding Remarks

In this study, we examine the multifaceted inﬂuence of homeownership on charitable giving

through several channels: tax deductibility, household wealth, and mobility. We ﬁnd that

homeowners donate more than renters, and the three channels are important predictors of

the owner-renter gap in donations. We also show that the owner-renter diﬀerence cannot be

fully explained by these three channels. After controlling for an extensive list of household

characteristics and the three channels, homeowners still donate approximately 20% more

than renters. To address the endogeneity surrounding the selection into homeownership,

we apply three modeling techniques: household ﬁxed-eﬀect models, propensity-score match-

ing (PSM), and matched-sample diﬀerence-in-diﬀerence (DiD) estimators. Our results are

robust to these modeling and identiﬁcation strategies. Our ﬁndings suggest that homeown-

ership increases generosity, and such an eﬀect goes beyond what can be explained by tax

deductibility, household wealth, and mobility.

We further examine two distinct dimensions of charitable giving: the likelihood to donate

and the size of donations. We show that owning a home signiﬁcantly boosts both the

likelihood to donate and the size of donations. Additionally, we ﬁnd that the likelihood to

donate correlates inversely with mobility but is insensitive to tax deductibility and wealth.

In contrast, tax deductibility and wealth are important predictors of the size of donations.

We further look into giving behaviors of diﬀerent generational cohorts. We ﬁnd that for

baby boomers, owning a home increases giving to both religious and non-religious causes.

In contrast, for Generation X, such an eﬀect exists only for non-religious donations. Finally,

we ﬁnd that millennial homeowners do not give more than their renter counterparts.

The importance of our ﬁndings is twofold. First, by documenting that the owner-renter gap

in charitable donations cannot be fully explained by tax deductibility, household wealth, and

mobility, our ﬁndings suggest that at least a part of the greater generosity of homeowners

(about 20% according to our estimation) is directly linked to homeownership itself. This

means homeownership raises charitable giving not just due to its fringe beneﬁts (i.e. tax

deductions). In other words, the linkage between homeownership and charitable giving is

rooted deeper than just the tax treatment of housing. Our ﬁndings imply that anything

(other than just tax rules) that aﬀects housing tenure choices, such as housing and mortgage

market conditions, household consumption and risk preferences, and racial disparities, can

all inﬂuence charitable giving through the homeownership channel.

Second, charitable givings are positively associated with social capital. For example, the

United States Congress Joint Economic Committee (JEC) publishes a county-level social

capital index, and one important component of the JEC Social Capital Index is the per-

centage of the adult population (i.e. 25 and older) who made donations in the past year.15

Communities with more people donating are better communities. Abundant social capital

in a community is linked to improved intergenerationally mobility (Chetty et al. (2014)),

more innovations (Hasan et al. (2020)), lower mortgage delinquency rates (Li et al. (2020)),

and fewer infections during the COVID-19 pandemic (Makridis and Wu (2021)). Our study

documents a positive relationship between homeownership and the likelihood to donate, a

crucial indicator of social capital. Therefore, it is perhaps reasonable to think that home-

ownership rates could be a critical determinant of community social capital and all social

outcomes inﬂuenced by it.

15see https://www.jec.senate.gov/public/index.cfm/republicans/socialcapitalproject. Ac-

cessed on March 5, 2023.

References

Austin, P. C. (2009). Balance diagnostics for comparing the distribution of baseline covariates

between treatment groups in propensity-score matched samples. Statistics in medicine,

28(25):3083–3107.

Auten, G. E., Sieg, H., and Clotfelter, C. T. (2002). Charitable giving, income, and taxes:

An analysis of panel data. American Economic Review, 92(1):371–382.

Barker, D. and Miller, E. (2009). Homeownership and child welfare. Real Estate Economics,

37(2):279–303.

Brown, S., Harris, M. N., and Taylor, K. (2012). Modelling charitable donations to an un-

expected natural disaster: Evidence from the us panel study of income dynamics. Journal

of Economic Behavior & Organization, 84(1):97–110.

Chetty, R., Hendren, N., Kline, P., and Saez, E. (2014). Where is the land of opportunity?

the geography of intergenerational mobility in the united states. The Quarterly Journal

of Economics, 129(4):1553–1623.

Clotfelter, C. T. (1980). Tax incentives and charitable giving: Evidence from a panel of

taxpayers. Journal of Public Economics, 13(3):319–340.

Dehejia, R. H. and Wahba, S. (2002). Propensity score-matching methods for nonexperi-

mental causal studies. Review of Economics and statistics, 84(1):151–161.

Di, Z. X., Belsky, E., and Liu, X. (2007). Do homeowners achieve more household wealth in

the long run? Journal of Housing Economics, 16(3-4):274–290.

Dietz, R. D. and Haurin, D. R. (2003). The social and private micro-level consequences of

homeownership. Journal of urban Economics, 54(3):401–450.

Dimock, M. (2019). Deﬁning generations: Where millennials end and generation z begins.

Pew Research Center, 17(1):1–7.

DiPasquale, D. and Glaeser, E. L. (1999). Incentives and social capital: Are homeowners

better citizens? Journal of urban Economics, 45(2):354–384.

Engelhardt, G. V., Eriksen, M. D., Gale, W. G., and Mills, G. B. (2010). What are the social

beneﬁts of homeownership? experimental evidence for low-income households. Journal of

urban Economics, 67(3):249–258.

Feldstein, M. and Clotfelter, C. (1976). Tax incentives and charitable contributions in the

united states: A microeconometric analysis. Journal of Public Economics, 5(1-2):1–26.

Feldstein, M. and Taylor, A. (1976). The income tax and charitable contributions. Econo-

metrica: Journal of the Econometric Society, pages 1201–1222.

Galster, G. C. (1983). Empirical evidence on cross-tenure diﬀerences in home maintenance

and conditions. Land Economics, 59(1):107–113.

Glenday, G., Gupta, A. K., and Pawlak, H. (1986). Tax incentives for personal charitable

contributions. The Review of Economics and Statistics, pages 688–693.

Gormley, T. A. and Matsa, D. A. (2011). Growing out of trouble? corporate responses to

liability risk. The Review of Financial Studies, 24(8):2781–2821.

Green, R. K. and White, M. J. (1997). Measuring the beneﬁts of homeowning: Eﬀects on

children. Journal of urban economics, 41(3):441–461.

Gyourko, J. and Sinai, T. (2003). The spatial distribution of housing-related ordinary income

tax beneﬁts. Real Estate Economics, 31(4):527–575.

Harding, J., Miceli, T. J., and Sirmans, C. (2000). Do owners take better care of their

housing than renters? Real Estate Economics, 28(4):663–681.

Hasan, I., Hoi, C.-K. S., Wu, Q., and Zhang, H. (2020). Is social capital associated with

corporate innovation? evidence from publicly listed ﬁrms in the us. Journal of Corporate

Finance, 62:101623.

Haurin, D. R., Parcel, T. L., and Haurin, R. J. (2002). Does homeownership aﬀect child

outcomes? Real Estate Economics, 30(4):635–666.

Heckman, J. (1974). Shadow prices, market wages, and labor supply. Econometrica: journal

of the econometric society, pages 679–694.

Holupka, S. and Newman, S. J. (2012). The eﬀects of homeownership on children’s outcomes:

Real eﬀects or self-selection? Real Estate Economics, 40(3):566–602.

Kimberlin, S., Kim, J., and Shaefer, L. (2014). An updated method for calculating income

and payroll taxes from psid data using the nber’s taxsim, for psid survey years 1999

through 2011. Unpublished manuscript, University of Michigan. Accessed May, 6:2016.

Lee, W.-S. (2013). Propensity score matching and variations on the balancing test. Empirical

economics, 44(1):47–80.

Li, L., Ucar, E., and Yavas, A. (2020). Social capital and mortgage delinquency. The Journal

of Real Estate Finance and Economics, pages 1–25.

Loving, A. C., Finke, M. S., and Salter, J. R. (2012). Does home equity explain the black

wealth gap? Journal of Housing and the Built Environment, 27(4):427–451.

Makridis, C. A. and Wu, C. (2021). How social capital helps communities weather the

covid-19 pandemic. PloS one, 16(1):e0245135.

Maslow, A. H. (1943). A theory of human motivation. Psychological review, 50(4):370.

Newman, S. J. and Holupka, C. S. (2016). Is timing everything? race, homeownership and

net worth in the tumultuous 2000s. Real Estate Economics, 44(2):307–354.

Randolph, W. C. (1995). Dynamic income, progressive taxes, and the timing of charitable

contributions. Journal of Political Economy, 103(4):709–738.

Rosenbaum, P. R. and Rubin, D. B. (1983). The central role of the propensity score in

observational studies for causal eﬀects. Biometrika, 70(1):41–55.

Rosenbaum, P. R. and Rubin, D. B. (1985). Constructing a control group using multi-

variate matched sampling methods that incorporate the propensity score. The American

Statistician, 39(1):33–38.

Rossi, P. H. and Weber, E. (1996). The social beneﬁts of homeownership: Empirical evidence

from national surveys. Housing policy debate, 7(1):1–35.

Schokkaert, E. (2006). The empirical analysis of transfer motives. Handbook of the economics

of giving, altruism and reciprocity, 1:127–181.

Shadish, W. R., Clark, M. H., and Steiner, P. M. (2008). Can nonrandomized experiments

yield accurate answers? a randomized experiment comparing random and nonrandom

assignments. Journal of the American statistical association, 103(484):1334–1344.

Strawinski, P. (2013). Controlling for overlap in matching. Technical report, Faculty of

Economic Sciences, University of Warsaw.

Turner, T. M. and Luea, H. (2009). Homeownership, wealth accumulation and income status.

Journal of Housing Economics, 18(2):104–114.

Wilhelm, M. O., Brown, E., Rooney, P. M., and Steinberg, R. (2008). The intergenerational

transmission of generosity. Journal of Public Economics, 92(10-11):2146–2156.

Table 1: Summary Statistics

Full Sample Homeowner Renter

Variables Mean Std. Dev. Mean Std. Dev. Mean Std. Dev.

Total donation 1165.888 3222.356 1741.444 3999.510 415.384 1438.825

Religious donation 450.646 1892.912 672.778 2382.284 160.995 840.775

Non-religious donation 715.242 2269.355 1068.666 2815.583 254.389 1073.820

Donor (Y/N) 0.548 0.498 0.710 0.454 0.338 0.473

Price of donation 0.936 0.119 0.900 0.137 0.984 0.063

Permanent income 62557.181 74338.206 82895.681 88091.435 36036.539 37022.578

Std. Dev. of permanent income 20363.465 48793.690 25476.297 58831.119 13696.522 29895.646

Wealth 239202.314 1109745.389 401789.645 1440846.169 27194.519 225738.558

# of moves 0.911 1.019 0.498 0.797 1.451 1.025

Years since move 11.264 13.313 15.812 14.173 5.335 9.182

Income 68173.997 95502.379 90665.235 116798.817 38846.264 41300.748

# of persons 2.655 1.492 2.817 1.410 2.443 1.567

# of children 0.827 1.181 0.779 1.112 0.890 1.262

Age 46.867 15.660 51.157 14.820 41.273 14.945

Male-headed household 0.705 0.456 0.813 0.390 0.563 0.496

Female-headed household 0.295 0.456 0.187 0.390 0.437 0.496

White 0.596 0.491 0.723 0.447 0.431 0.495

African-American 0.349 0.477 0.225 0.417 0.512 0.500

Asian-American 0.012 0.109 0.014 0.116 0.010 0.099

Other race 0.042 0.201 0.038 0.192 0.048 0.213

None, atheist 0.138 0.345 0.113 0.317 0.170 0.376

Catholic 0.197 0.397 0.228 0.420 0.155 0.362

Jewish 0.019 0.136 0.023 0.151 0.013 0.114

Protestant 0.619 0.486 0.614 0.487 0.625 0.484

Other non-Christian: Muslim, Rastafarian, etc. 0.013 0.115 0.011 0.105 0.017 0.128

Greek/Russian/Eastern Orthodox 0.002 0.047 0.002 0.050 0.002 0.042

Other 0.012 0.110 0.008 0.090 0.018 0.132

Less than high school 0.175 0.380 0.121 0.326 0.246 0.431

High school graduate 0.303 0.460 0.293 0.455 0.316 0.465

Some college 0.258 0.437 0.249 0.432 0.269 0.443

College graduate 0.157 0.364 0.196 0.397 0.106 0.308

Some post graduate 0.107 0.309 0.140 0.347 0.063 0.243

Married 0.548 0.498 0.729 0.445 0.313 0.464

Single 0.211 0.408 0.076 0.265 0.387 0.487

Widowed 0.059 0.235 0.063 0.244 0.052 0.223

Divorced or separated 0.182 0.386 0.132 0.338 0.248 0.432

Employed 0.709 0.454 0.735 0.441 0.675 0.468

Unemployed, looking for work 0.066 0.248 0.028 0.166 0.114 0.318

Unemployed, not looking for work 0.084 0.278 0.047 0.212 0.133 0.340

Retired 0.141 0.348 0.190 0.392 0.077 0.267

Health condition - excellent 0.182 0.386 0.195 0.396 0.164 0.370

Health condition - very good 0.340 0.474 0.365 0.481 0.309 0.462

Health condition - good 0.309 0.462 0.304 0.460 0.315 0.465

Health condition - fair 0.129 0.335 0.104 0.306 0.161 0.368

Health condition - poor 0.040 0.195 0.031 0.174 0.051 0.219

N 51,224 28,991 22,233

Table 2: Household Fixed-Eﬀects Model

[1] [2] [3] [4] [5]

Homeowner 0.304*** 0.236*** 0.234*** 0.213*** 0.182

(7.78) (5.54) (5.49) (4.98) (1.08)

Price of donation -0.519*** -0.520*** -0.527*** 0.092

(-4.36) (-4.36) (-4.42) (0.20)

ln(permanent income) 0.067 0.065 0.061 0.300*

(1.61) (1.58) (1.47) (1.83)

ln(Std. Dev. of permanent income) -0.012 -0.012 -0.009 -0.166***

(-0.72) (-0.73) (-0.52) (-2.50)

Wealth 0.025* 0.024* 0.304***

(1.73) (1.70) (5.35)

# of moves -0.011 0.001

(-0.59) (0.02)

Years since move 0.009*** 0.014*

(4.04) (1.69)

ln(income) 0.041*** 0.024* 0.024* 0.024* 0.064

(3.67) (1.82) (1.81) (1.82) (1.23)

# of person -0.045** -0.062*** -0.062** -0.063*** -0.181*

(-1.98) (-2.58) (-2.57) (-2.62) (-1.91)

# of children 0.081*** 0.094*** 0.094*** 0.096*** 0.296***

(3.14) (3.48) (3.47) (3.55) (2.77)

Age 0.078** 0.052 0.052 0.054 0.006

(2.42) (1.44) (1.44) (1.49) (0.04)

Age2-0.001*** -0.001*** -0.001*** -0.001*** -0.001***

(-13.21) (-11.23) (-11.26) (-11.43) (-2.81)

Female-headed household 0.840 0.858 0.856 0.846 0.333

(0.34) (0.35) (0.35) (0.35) (0.03)

African-American 0.245 0.589** 0.588** 0.571** 0.775

(1.00) (2.23) (2.22) (2.16) (0.74)

Asian-American -0.089 -0.153 -0.155 -0.208 -0.871

(-0.25) (-0.40) (-0.41) (-0.55) (-0.58)

Other race 0.127 0.151 0.150 0.140 0.138

(1.09) (1.23) (1.22) (1.14) (0.29)

Catholic -0.208 -0.302 -0.297 -0.285 1.090

(-0.68) (-0.91) (-0.90) (-0.86) (0.83)

Jewish -0.483 -0.077 -0.077 -0.161 1.589

(-0.60) (-0.09) (-0.09) (-0.19) (0.48)

Protestant -0.332* -0.283 -0.285 -0.286 0.504

(-1.80) (-1.36) (-1.37) (-1.38) (0.61)

Other non-Christian -0.126 -1.056* -1.056* -1.073* 0.003

(-0.25) (-1.76) (-1.76) (-1.79) (0.00)

Greek/Russian/Eastern Orthodox -0.859 -1.964 -1.960 -2.076 0.372

(-0.50) (-0.80) (-0.80) (-0.84) (0.04)

Other -0.687*** -0.601** -0.603** -0.604** 0.811

(-2.92) (-2.17) (-2.18) (-2.18) (0.74)

Continued

Table 2: Household Fixed-Eﬀects Model - Cont.

[1] [2] [3] [4] [5]

High school graduate -0.160** -0.139 -0.138 -0.135 -0.558

(-1.97) (-1.59) (-1.59) (-1.55) (-1.62)

Some college -0.094 -0.066 -0.066 -0.058 0.008

(-1.03) (-0.68) (-0.68) (-0.60) (0.02)

College graduate -0.241** -0.175 -0.175 -0.172 -0.540

(-2.10) (-1.43) (-1.43) (-1.41) (-1.12)

Some post graduate -0.271** -0.239* -0.239* -0.235* -0.732

(-2.14) (-1.77) (-1.77) (-1.74) (-1.37)

Single -0.383*** -0.376*** -0.377*** -0.382*** 0.124

(-5.25) (-4.40) (-4.41) (-4.46) (0.37)

Widowed -0.164 -0.182 -0.182 -0.176 -0.655

(-1.12) (-1.21) (-1.21) (-1.17) (-1.10)

Divorced or separated -0.568*** -0.587*** -0.586*** -0.574*** -0.351

(-7.96) (-7.89) (-7.89) (-7.72) (-1.19)

Unemployed, looking for work -0.024 -0.033 -0.033 -0.033 0.104

(-0.51) (-0.66) (-0.66) (-0.65) (0.52)

Unemployed, not looking for work -0.010 0.002 0.002 0.003 0.120

(-0.18) (0.04) (0.04) (0.05) (0.54)

Retired -0.058 -0.039 -0.039 -0.037 0.700

(-1.08) (-0.71) (-0.71) (-0.66) (3.20)***

Health condition - very good 0.017 0.014 0.014 0.014 -0.097

(0.50) (0.38) (0.36) (0.38) (-0.65)

Health condition - good 0.013 0.010 0.010 0.012 0.080

(0.33) (0.24) (0.24) (0.27) (0.47)

Health condition - fair -0.108** -0.114** -0.113** -0.108** 0.022

(-2.10) (-2.09) (-2.07) (-1.98) (0.10)

Health condition - poor -0.305*** -0.315*** -0.314*** -0.305*** -0.388

(-3.89) (-3.83) (-3.82) (-3.70) (-1.19)

Survey-year ﬁxed eﬀects Yes Yes Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes Yes Yes

Household ﬁxed eﬀects Yes Yes Yes Yes Yes

N 57,945 51,224 51,224 51,224 51,224

R-squared 71.54% 72.01% 72.01% 72.03% 35.19%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 3: Household Fixed-Eﬀects Model ∼Likelihood to Donate and the Size

of Donation

[1] [2]

Dependent variable Donor (Y/N) Size of donation

Logit Coef. Odds Ratio Coef.

Homeowner 0.225*** 1.253 0.166***

(3.96) (3.02)

Price of donation -0.146 0.864 -0.351**

(0.85) (-2.39)

ln(permanent income) -0.037 0.964 -0.038

(0.57) (-0.57)

ln(Std Dev of permanent income) 0.005 1.005 -0.003

(0.21) (-0.17)

Wealth 0.018 1.018 0.021

(0.49) (1.43)

# of moves -0.016 0.984

(0.62)

Year since move 0.009*** 1.009

(3.09)

Inverse Mills ratios (λ) -0.423*

(-1.94)

Household Characteristics Yes Yes

Survey-year ﬁxed eﬀects Yes Yes

State ﬁxed eﬀects Yes Yes

Household ﬁxed eﬀects Yes Yes

N 51,224 51,224

R-squared / pseudo R-squared 31.14% 72.02%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 4: Propensity-Score Matching (PSM) Estimates

Dependent variable Total donation Donor (Y/N) Size of donation

Homeowner 0.237*** 0.302*** 0.121***

(4.00) (3.73) (3.55)

N 20,434 20434 10,474

R-squared 74.96% 20.6% 80.85%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 5: Matched-Sample Diﬀerence-in-Diﬀerence (DiD) Estimates

Dependent variable Total donation Donor (Y/N) Size of donation

T reatment ×P ost 0.339*** 0.036** 0.253***

(3.47) (2.28) (3.94)

N 6,383 6,383 3,037

R-squared 71.22% 65.88% 78.34%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 6: Household Fixed-Eﬀects Model ∼Religious vs. Non-Religious

Donations

[1] [2]

Dependent variable Religious

donation

Non-religious

donation

Coef. Coef.

Homeowner 0.140*** 0.134***

(3.55) (3.28)

Price of donation -0.482*** -0.527***

(-4.40) (-4.63)

ln(Permanent income) 0.035 0.062

(0.91) (1.58)

ln(Std. Dev. of permanent income) -0.042*** 0.003

(-2.71) (0.19)

Wealth 0.016 -0.002

(1.20) (-0.16)

# of moves -0.002 -0.020

(-0.10) (-1.13)

Years since move 0.006*** 0.006***

(3.21) (2.84)

Household Characteristics Yes Yes

Survey-year ﬁxed eﬀects Yes Yes

State ﬁxed eﬀects Yes Yes

Household ﬁxed eﬀects Yes Yes

N 51,224 51,224

R-squared 73.57% 67.23%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 7: Propensity Score Matching (PSM) Estimates ∼Religious vs.

Non-Religious Donations

Dependent variable Religious

donation

Non-religious

donation

Treatment: Homeowner = Yes 0.159*** 0.167***

(3.04) (3.03)

N 20,434 20,434

R-squared 77.39% 71.19%

Propensity score estimated based on:

Tax deductibility Yes Yes

Wealth Yes Yes

Mobility Yes Yes

Household characteristics Yes Yes

State ﬁxed eﬀects Yes Yes

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 8: Matched-Sample Diﬀerence-in-Diﬀerence (DiD) Estimates ∼Religious

vs. Non-Religious Donations

Dependent variable Religious

donation

Non-religious

donation

T reatment ×P ost 0.189** 0.269***

(2.32) (2.90)

N 6,383 6,383

R-squared 75.18% 65.29%

Propensity score estimated based on:

Tax deductibility Yes Yes

Wealth Yes Yes

Mobility Yes Yes

Household characteristics Yes Yes

State ﬁxed eﬀects Yes Yes

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 9: Household Fixed-Eﬀects Model ∼Generational Diﬀerences

Panel A: Dependent Variable = Total Donation

[1] [2] [3]

Dependent variable Boomers Generation X Millennial

Homeowner 0.226*** 0.249*** 0.049

(3.26) (3.55) (0.53)

Price of donation -0.335** -0.638*** -0.971***

(-2.07) (-3.00) (-2.90)

ln(Permanent income) 0.137** -0.041 0.076

(2.21) (-0.56) (0.83)

ln(Std. Dev. of permanent income) -0.060** 0.038 0.037

(-2.47) (1.22) (1.05)

Wealth 0.028* 0.009 0.022

(1.79) (0.26) (0.25)

# of moves -0.028 0.006 -0.039

(-0.91) (0.21) (-0.97)

Years since move 0.007*** 0.011** 0.001

(2.76) (2.49) (0.09)

Household Characteristics Yes Yes Yes

Survey-year ﬁxed eﬀects Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

Household ﬁxed eﬀects Yes Yes Yes

N 24,839 16,249 10,136

R-squared 71.60% 68.28% 72.28%

Panel B: Dependent Variable = Religious Donation

[1] [2] [3]

Dependent variable Boomers Generation X Millennial

Homeowner 0.224*** 0.046 0.079

(3.32) (0.73) (1.10)

Price of donation -0.337** -0.526*** -1.008***

(-2.14) (-2.78) (-3.85)

ln(Permanent income) 0.081 0.023 -0.065

(1.35) (0.35) (-0.90)

ln(Std. Dev. of permanent income) -0.061** -0.013 -0.013

(-2.54) (-0.48) (-0.46)

Wealth 0.004 0.060* 0.079

(0.27) (1.89) (1.10)

# of moves -0.020 -0.004 0.024

(-0.68) (-0.16) (0.76)

Years since move 0.004 0.010** -0.002

(1.52) (2.55) (-0.31)

Household Characteristics Yes Yes Yes

Survey-year ﬁxed eﬀects Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

Household ﬁxed eﬀects Yes Yes Yes

N 24,839 16,249 10,136

R-squared 73.08% 70.56% 75.55%

Continued

Table 9: Household Fixed-Eﬀects Model ∼Generational Diﬀerences - Cont.

Panel C: Dependent Variable = Non-religious Donation

[1] [2] [3]

Boomers Generation X Millennial

Homeowner 0.135** 0.234*** -0.082

(2.05) (3.44) (-0.96)

Price of donation -0.523*** -0.508** -0.492

(-3.40) (-2.46) (-1.58)

ln(Permanent income) 0.124** -0.071 0.155*

(2.11) (-1.00) (1.81)

ln(Std. Dev. of permanent income) -0.053** 0.044 0.043

(-2.29) (1.46) (1.28)

Wealth 0.003 -0.025 0.041

(0.20) (-0.71) (0.48)

# of moves -0.008 -0.011 -0.063*

(-0.27) (-0.37) (-1.69)

Years since move 0.006** 0.005 0.006

(2.50) (1.11) (0.84)

Household Characteristics Yes Yes Yes

Survey-year ﬁxed eﬀects Yes Yes Yes

Household ﬁxed eﬀects Yes Yes Yes

N 24,839 16,249 10,136

R-squared 68.77% 61.89% 65.63%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 10: Propensity-Score Matching (PSM) Estimates ∼Age/Generation

Diﬀerences

Panel A: Dependent Variable = Total Donation

Boomers Generation X Millennials

Treatment: Homeowner = Yes 0.266*** 0.206** 0.046

(2.83) (2.25) (0.29)

N 9,582 7,514 3,002

R-squared 74.33% 73.65% 83.04%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

Panel B: Dependent Variable = Religious Donation

Boomers Generation X Millennials

Treatment: Homeowner = Yes 0.257*** 0.007 0.161

(2.95) (0.09) (1.31)

N 9,582 7,514 3,002

R-squared 75.89% 77.14% 85.83%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

Panel C: Dependent Variable = Non-religious Donation

Boomers Generation X Millennials

Treatment: Homeowner = Yes 0.124 0.176** 0.035

(1.47) (1.99) (0.25)

N 9,582 7,514 3,002

R-squared 72.77% 67.70% 79.94%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table 11: Matched-Sample Diﬀerence-in-Diﬀerence (DiD) Estimates ∼

Age/Generation Diﬀerences

Panel A: Dependent Variable = Total Donation

Boomers Generation X Millennials

T reatment ×P ost 0.381* 0.456*** 0.136

(1.81) (2.71) (0.68)

N 1,940 2,325 1,836

R-squared 74.05% 68.53% 73.01%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

Panel B: Dependent Variable = Religious Donation

Boomers Generation X Millennials

T reatment ×P ost 0.145 0.294** -0.067

(0.76) (2.14) (-0.43)

N 1,940 2,325 1,836

R-squared 76.09% 73.31% 75.96%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

Panel C: Dependent Variable = Non-religious Donation

Boomers Generation X Millennials

T reatment ×P ost 0.305 0.372** 0.143

(1.60) (2.36) (0.77)

N 1,940 2,325 1,836

R-squared 69.49% 63.18% 68.81%

Propensity score estimated based on:

Tax deductibility Yes Yes Yes

Wealth Yes Yes Yes

Mobility Yes Yes Yes

Household characteristics Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Appendix

Table A1: Household Fixed-Eﬀects Model ∼By Donation Categories

Combo Community Culture Education Environment Health Needy Other World Youth

Homeowner 0.019 0.026 0.043*** 0.038 0.036** 0.018 0.162*** 0.019 0.027 -0.024

(0.53) (1.55) (2.39) (1.46) (1.98) (0.62) (4.14) (0.79) (1.47) (-0.99)

Price of donation -0.564*** 0.063 0.121*** -0.048 0.026 -0.205*** -0.202* -0.011 -0.032 0.086

(-5.56) (1.34) (2.42) (-0.66) (0.50) (-2.51) (-1.85) (-0.16) (-0.63) (1.29)

ln(permanent income) 0.008 0.002 0.033* 0.026 0.019 0.073*** 0.043 0.014 0.009 0.019

(0.23) (0.12) (1.93) (1.02) (1.08) (2.56) (1.15) (0.59) (0.53) (0.84)

ln(Std Dev of permanent income) -0.029** 0.002 0.004 0.001 -0.001 -0.018 0.001 0.005 0.000 -0.002

(-2.00) (0.33) (0.51) (0.08) (-0.14) (-1.53) (0.07) (0.54) (0.05) (-0.24)

Wealth -0.006 -0.006 0.030 0.001 0.006 0.002 -0.013 0.012 0.000 0.014*

(-0.46) (-1.00) (4.96) (0.14) (1.04) (0.19) (-0.97) (1.48) (-0.08) (1.78)

# of moves 0.006 0.002 0.001 -0.006 0.013 -0.010 -0.007 -0.027*** -0.006 0.010

(0.36) (0.34) (0.13) (-0.54) (1.59) (-0.76) (-0.40) (-2.60) (-0.79) (1.01)

Year since move 0.001 0.000 0.002** 0.001 0.002*** 0.003* 0.004* 0.003*** 0.000 -0.002

(0.81) (0.15) (2.02) (0.81) (2.40) (1.84) (1.89) (2.51) (0.35) (-1.63)

Household Characteristics Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Survey-year ﬁxed eﬀects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

State ﬁxed eﬀects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Household ﬁxed eﬀects Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

N 51,224 51,206 51,203 51,181 51,190 51,106 50,942 51,164 51,192 51,173

R-squared 51.67% 33.62% 58.17% 55.05% 53.44% 52.26% 49.41% 30.45% 40.58% 42.01%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table A2: Household Fixed-Eﬀects Model ∼Children’s Age

[1] [2]

Homeowner 0.175*** 0.177***

(2.61) (2.64)

Price of donation -0.551*** -0.551***

(-2.72) (-2.71)

ln(permanent income) 0.103 0.099

(1.32) (1.26)

ln(Std Dev of permanent income) -0.003 -0.004

(-0.11) (-0.12)

Wealth 0.011 0.013

(0.27) (0.31)

# of moves -0.017 -0.018

(-0.60) (-0.65)

Year since move 0.003 0.003

(0.63) (0.65)

Age of youngest child 0.018***

(2.64)

Middle childhood (youngest child) 0.098**

(1.96)

Young teens/teenagers (youngest child) 0.117

(1.59)

Household Characteristics Yes Yes

Survey-year ﬁxed eﬀects Yes Yes

State ﬁxed eﬀects Yes Yes

Household ﬁxed eﬀects Yes Yes

N 21,694 21,694

R-squared 75.18% 75.17%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

Table A3: Distribution of Estimated Propensity Scores

# Mean Std. Dev. Min Q1 Median Q3 Max

Renter 10217 0.576 0.257 0.000 0.376 0.596 0.791 1.000

Owner 10217 0.576 0.257 0.000 0.376 0.597 0.791 1.000

Table A4: Tests of Overlapping and Balancing Conditions

[1] [2]

Variable Coef. tstat. / χ2stat.

Price of donation -0.014 0.62

(0.11)

Permanent income 0.028 0.33

(0.74)

Std. Dev. of permanent income 0.002 -0.16

(0.14)

Wealth 0.093** -2.56

(2.45)

# of moves -0.004 -0.82

(0.24)

Years since move -0.003** 2.49

(2.16)

ln(income) -0.010 0.79

(0.45)

# of persons -0.052** 0.89

(2.16)

# of children 0.065** -0.46

(2.35)

Age -0.005 -0.47

(0.82)

Age20.000 -0.72

(0.85)

Female-headed household 0.029 2.42

(1.28)

African-American 0.034 0.64

(0.76)

Asian-American -0.057

(0.59)

Other race 0.009

(0.16)

Catholic -0.026 3.79

(0.43)

Jewish 0.003

(0.03)

Protestant -0.0005

(0.01)

Other non-Christian 0.159

(1.37)

Greek/Russian/Eastern Orthodox -0.231

(0.87)

Other 0.055

(0.44)

Continued

Table A4: Tests of Overlapping and Balancing Conditions - Cont.

[1] [2]

Variable Coef. tstat. / χ2stat.

High school graduate 0.023 0.84

(0.89)

Some college -0.015

(0.58)

College graduate 0.013

(0.38)

Some post graduate -0.033

(0.82)

Single 0.021 1.91

(0.57)

Widowed -0.073

(1.34)

Divorced or separated 0.014

(0.44)

Unemployed, looking for work -0.034 6.84

(0.64)

Unemployed, not looking for work 0.068

(1.54)

Retired 0.026

(0.53)

Health condition - very good -0.025 1.98

(0.89)

Health condition - good -0.012

(0.44)

Health condition - fair 0.026

(0.73)

Health condition - poor 0.016

(0.27)

Intercept 0.288

(1.58)

N 20434

pseudo R-squared 0.6%

t-values in parentheses.

*** p<0.01, ** p<0.05, * p<0.1

ResearchGate has not been able to resolve any citations for this publication.

How social capital helps communities weather the COVID-19 pandemic

Article

Full-text available

Jan 2021
PLOS ONE

Why have the effects of COVID-19 been so unevenly geographically distributed in the United States? This paper investigates the role of social capital as a mediating factor for the spread of the virus. Because social capital is associated with greater trust and relationships within a community, it could endow individuals with a greater concern for others, thereby leading to more hygienic practices and social distancing. Using data for over 2,700 US counties, we investigate how social capital explains the level and growth rate of infections. We find that moving a county from the 25th to the 75th percentile of the distribution of social capital would lead to a 18% and 5.7% decline in the cumulative number of infections and deaths, as well as suggestive evidence of a lower spread of the virus. Our results are robust to many demographic characteristics, controls, and alternative measures of social capital.

Social Capital and Mortgage Delinquency

Article

Full-text available

Apr 2022
J R Estate Finance Econ

This study offers a simple theoretical model and empirical evidence to address the impact of social capital on mortgage delinquency. Social capital includes the norms, values, trust, and information common to a social network, which enable cooperative and shared actions. Using a new county-level dataset between 1999 and 2011 for the U.S, we find new evidence to show that social capital significantly affects the likelihood of mortgage delinquency. In particular, we find that a one-standard-deviation increase in social capital leads to a 0.13 standard deviation decrease in mortgage delinquency. The effect of social capital remains significant after controlling for location fixed effects and addressing endogeneity. The primary explanation is that social norms or trust could limit opportunistic behavior among homeowners and negatively affect strategic default activities. We also find that the impact of social capital on mortgage delinquency increased after the recent financial crisis. Furthermore, we show that the impact of social capital is more pronounced when the default is more likely to be strategic. Our findings have important implications for players in the mortgage industry and for policymakers in that cooperative and shared actions can play an important role in the mortgage default process. Thus, the assessment of default risk should consider social capital in addition to the factors already documented in the literature.

The Central Role of the Propensity Score in Observational Studies For Causal Effects

Article

Full-text available

Apr 1983
BIOMETRIKA

The propensity score is the conditional probability of assignment to a particular treatment given a vector of observed covariates. Both large and small sample theory show that adjustment for the scalar propensity score is sufficient to remove bias due to all observed covariates. Applications include: (i) matched sampling on the univariate propensity score, which is a generalization of discriminant matching, (ii) multivariate adjustment by subclassification on the propensity score where the same subclasses are used to estimate treatment effects for all outcome variables and in all subpopulations, and (iii) visual representation of multivariate covariance adjustment by a two- dimensional plot.

Is Social Capital Associated with Corporate Innovation? Evidence from Publicly Listed Firms in the U.S.

Article

Jan 2020

Is social capital associated with corporate innovation? Evidence from publicly listed firms in the U.S

Article

Apr 2020
J CORP FINANC

We find that social capital in U.S. counties, as captured by strength of social norms and density of social networks, is positively associated with innovation of firms headquartered in the county, as captured by patents and citations. This relation is robust in fixed-effect regressions, instrumental variable regressions with a Bartik instrument, propensity score matching regressions, and a difference-in-differences design that isolates the effects of over time variations in social capital due to corporate headquarter relocations. Strength of social norms plays a more dominant role than density of social networks in producing these empirical regularities. Cross-sectional evidence indicates the prominence of the contracting channel through which social capital relates to innovation. Additionally, we find that social capital is also positively associated with trademarks and effectiveness of corporate R&D expenditures.

Propensity Score Matching and Variations on the Balancing Test

Article

Jan 2006

Wang-Sheng Lee

A dynamic theory of human motivation

Article

Jan 1958

A.H. Maslow

Is Timing Everything? Race, Homeownership and Net Worth in the Tumultuous 2000s

Article

Sep 2015
R Estate Econ

We use the Panel Study of Income Dynamics to estimate how net worth was affected among low- and moderate-income households who became first-time homebuyers at different points during the volatile 2000s. We address selection using propensity score matching and estimating difference-in-difference models, and use quantile regressions to account for the skew in net worth outcomes. Results highlight the significance of race in the relationship between first-time home buying and net worth during the decade. Although timing was critical to the short-term trajectory of net worth for whites, total net worth declines for black first-time homebuyers regardless of economic climate. The most dramatic differences between black and white new homebuyers is their neighborhood locations, with blacks purchasing in predominantly black neighborhoods with lower housing prices and price appreciation, and lower and declining rates of homeownership.

Where is the Land of Opportunity? The Geography of Intergenerational Mobility In The United States

Article

Dec 2014

We use administrative records on the incomes of more than 40 million children and their parents to describe three features of intergenerational mobility in the United States. First, we characterize the joint distribution of parent and child income at the national level. The conditional expectation of child income given parent income is linear in percentile ranks. On average, a 10 percentile increase in parent income is associated with a 3.4 percentile increase in a child’s income. Second, intergenerational mobility varies substantially across areas within the United States. For example, the probability that a child reaches the top quintile of the national income distribution starting from a family in the bottom quintile is 4.4% in Charlotte but 12.9% in San Jose. Third, we explore the factors correlated with upward mobility. High mobility areas have (i) less residential segregation, (ii) less income inequality, (iii) better primary schools, (iv) greater social capital, and (v) greater family stability. Although our descriptive analysis does not identify the causal mechanisms that determine upward mobility, the publicly available statistics on intergenerational mobility developed here can facilitate research on such mechanisms. JEL Codes: H0, J0, R0.

The Social Benefits of Homeownership: Empirical Evidence from National Surveys

Article

Jan 1996

Claims that homeownership is beneficial to both owners and society have not been examined empirically. This article explores evidence from the General Social Survey and the National Survey of Families and Households, supple-mented by data from the American National Election Studies and several small but highly relevant researches, to determine whether owners and renters differ in a variety of ways. Strong differences in demographics were found. Small differences were found in some respects: Owners tended to be higher in life satisfaction and self-esteem and more likely to be members of community improvement groups. On a wide variety of social issues—from political partisanship to ethnocentric views—owners and renters were essentially alike. Thus, it appears that the claims for some social and individual benefits from homeownership are supported, but only weakly. Recognizing that the exploratory analyses pre-sented are not definitive, several strategies for more definitive future research are recommended.

Does Owning a Home Make Us More Generous?

Abstract and Figures

Recommended publications

Does Owning a Home Make Us More Generous?

Are Homeowners More Likely to Donate?

Does homeownership promote wealth accumulation?

Holding onto the Past: Previous Homes, Post-Move Housing Consumption, and the Great Recession