1

Comments Welcome

Understanding Food Insecurity during the Great Recession: Final Report for the Russell Sage Foundation Award 92-12-09

Patricia M. Anderson Dartmouth College

Kristin F. Butcher Wellesley College

Hilary W. Hoynes

University of California, Berkeley

Diane Whitmore Schanzenbach Northwestern University

Note: We thank Mary Zaki for excellent research assistance. We also thank Joshua

Gliken, Patrick Gould, Grace Ma, Nicholas Paine, Jamie Song and Linh Vu for excellent

research assistance under the auspices of the James O. Freedman Presidential Scholar

Program at Dartmouth College.

2

Executive Summary

This project documents a sharp increase in food insecurity during the Great

Recession. We find that the increase in food insecurity leads the increase in poverty that

accompanied the recession, and that there was an increase in food insecurity even after

controlling for a household’s income relative to the poverty ratio.

We explore components of the food insecurity measure, and find that the increase is

not only driven by an increase in anxiety about resources, but also among more severe

outcomes such as increases in skipping meals among adults and children. The increase in

food insecurity occurred across a wide range of the income distribution, above income

levels that are eligible for food stamp benefits and up to 300 percent of the poverty line. We

find that the factors that had previously been protective against food insecurity – such as

having completed college or owning a home – have lost some of their predictive power.

Importantly, we also find that while the increase in the unemployment rate can

explain the increase in food insecurity, the relationship works through more than just the

decline in income. It appears that the unemployment rate may be serving as a proxy for

other factors in the macro economy that were particularly harmed during the Great

Recession, and declines in rates of leverage and housing prices appear to play an important

role in the explanation.

Across a wide range of evidence, we conclude that the best explanation is that the

overall rate of consumption relative to income appears to have dropped during this period.

The pattern is consistent with explanatory factors that include psychological factors, an

increase in saving, and a loss of credit.

3

Introduction

Food insecurity rates have closely tracked poverty rates in the United States in

recent years, and both have trended up since 2000. With the onset of the recession in 2008,

however, food insecurity rates spiked and are far above the level predicted by the poverty

rate, as seen in Figure 1. In 2007, 11 percent of households were deemed food insecure,

and by 2008 that had climbed to over 14 percent, and was higher than the percent of

households in poverty. By 2010, the poverty rate caught up and rose to 15.1 percent, and

the food insecurity rate remained above its historical levels at 14.5 percent.1 Both the

poverty and food insecurity rates have remained unchanged through 2012.

Food insecurity, and the policies that combat it, are critical issues. Access to

adequate amounts of healthy food is of primary importance to health and well-being, and

food insecurity may have lasting consequences, particularly for children. Recent work in

developmental biology (summarized in Gluckman and Hanson 2005) has shown that

inadequate availability of nutrition during certain critical periods of a child’s development

causes irreversible damage to brain functioning and future health.2 Even with this

increased awareness of the importance of resources during childhood, the fraction of

children living in food-insecure households has increased from 16.9 percent in 1999 to

23.2 percent in 2009. This has declined somewhat in recent years, with the 2012 rate back

down to 20.0. Over that same time period, the fraction of children living in households with

1 See http://www.ers.usda.gov/briefing/foodsecurity/stats_graphs.htm#food_secure and http://www.census.gov/hhes/www/poverty/data/incpovhlth/2010/index.html 2 In recent work, Almond, Hoynes & Schanzenbach show that the introduction of Food Stamps in the 1960s and 1970s improved infant health (2011) and also improved adult health outcomes among the cohort that were exposed to the program in early childhood (Hoynes, Schanzenbach and Almond 2012). These findings demonstrate that food insecurity is critical and that policy intervention can be effective in addressing it. See also Gundersen and Oliveira (2001), Gundersen and Ziliak (2003), and Borjas (2004).

4

very low food security among children almost doubled, from 0.7 percent in 1999 to a peak

of 1.3 percent in 2009, to 1.2 percent in 2012.

The poverty indicator is only a blunt proxy for the resources available to a family,

and the rate of food insecurity typically declines in relation to a family’s income level on

both sides of the poverty line. Given this relationship, some of the increase in the rate of

food insecurity may be due to a decline in income that for many families accompanied the

recession. Figure 2 presents rates of food insecurity by a family’s income-to-poverty ratio,

separately for the time period before and after the onset of the Great Recession.3 If

diminished resources explained the increase in food insecurity rates, but the relationship

between income and food insecurity remained the same, then the two lines would overlap.

In such a case, the overall rate of food insecurity would have increased because there are

more families with low levels of income relative to poverty, but the relationship between

income and food insecurity would remain unchanged. Instead, we see that after holding

constant a family’s income relative to poverty, rates of food insecurity increased by an

average of 5 percentage points during the Great Recession.

This project attempts to understand what has been different about the Great

Recession and its relationship to food insecurity. Prior research has described how

households with different food security status compare in terms of background

characteristics, total food spending, and dietary intake. This report documents the work we

have done to understand the patterns of food insecurity during the Great Recession. We

have combined evidence from many data sources including the Current Population Survey

3 Because of limited sample sizes, we present rates of food insecurity averaged across 20-point bins (e.g. 0-20%; 21-40%; 41-60%) until 200 percent of the poverty line, then average rates for 200-250 percent and 250-300 percent of the poverty line.

5

Food Security Supplement (CPS-FSS), the National Health and Nutrition Examination

Survey (NHANES) and the American Time Use Study (ATUS) to document the economic,

behavioral and demographic characteristics of food insecure households and investigate

how these patterns have changed during the Great Recession.

We summarize the entire body of work in this report, and are in the process of

finalizing various manuscripts to submit for publication. The report is broken into 6

sections. Section 1 describes the measurement of food insecurity, and documents the

increase in the overall measure and its component parts during the Great Recession.

Section 2 documents the fluctuations in aggregate economic indicators during the Great

Recession, and tests whether those are differentially related to food insecurity during this

time period. Section 3 examines whether characteristics of food insecure households have

changed during the Great Recession. Section 4 documents changes in individuals’ behavior

in terms of time use, consumption nutrient intake patterns during the Great Recession.

Section 5 investigates the potential roles of macroeconomic conditions in predicting food

insecurity outcomes, including overall prices, housing prices, unemployment duration, and

debt-to-income leverage rates. Section 6 concludes and describes our plans for future

research.

I. The Increase in Food Insecurity

a. Defining Food Insecurity

Food security status is measured through a battery of questions asked during the

December Current Population Survey as part of the Food Security Supplement (CPS-FSS).

There are 10 questions asked of all households, and an additional 8 questions asked of

6

households with children. Table 1 shows the questions included in the battery (Bickel et al.

2000). There are four kinds of questions: those that capture anxiety or perception that the

food budget or supply is inadequate in quantity. There are also questions that capture

whether food is perceived to be inadequate in quality. A group of questions are more

quantitative in nature, asking about instances where food intake was reduced or weight

loss occurred associated with reduced food intake. One set of these questions pertains to

adults and the other to children in the household.

Answering more of these questions affirmatively indicates a more severe degree of

food insecurity. For example, for the battery of questions in the Current Population Survey,

“very low food security among children” is equal to 1 if 5 or more of the 8 child-centered

food security questions are answered affirmatively (Nord et al. 2009), and zero otherwise.

The standard battery of questions, and the screening or skip logic for the battery, has

changed over time as thinking about food security and hunger has changed, although for

the CPS-FSS the battery and screening has been consistent since 1998.

Figures 3 and 4 show the percent of households answering affirmatively to each of

these items from 1998 through 2010. The vertical line indicates 2008, the beginning of the

Great Recession. Figures 3a and 3b present results for the items that refer to adult food

consumption and Figures 4a and 4b are for items that specifically refer to children’s food

consumption. Some components of food insecurity, for example being worried that food

will not last, are much more common than others, for example children going a whole day

without eating. The higher prevalence items are in Figure 3a and 4a, and the lower

prevalence items are in 3b and 4b, so that the time patterns are not obscured by the scales.

7

b. Increase in Food Insecurity during the Great Recession

It is clear looking across all the items that food insecurity is higher in worse

economic periods. Some of the items appear more variable than others, but all have

increased during the period of the Great Recession. This, of course, is not unexpected. The

“Great Recession” is the worst economic contraction since the Great Depression. To the

extent that there are more individuals in poverty, we would expect that there would be

more food insecure individuals as well. However, Figure 1 suggests that food insecurity is

high both because the poverty rate increased and because for a given poverty rate, there

are more people reporting they are food insecure. Furthermore, Figure 2 shows that even

conditional on income-to-poverty levels, food insecurity rates climbed during the Great

Recession.

The level and increase in food insecurity is not uniform across the nation. Figures 5a

and 5b show state average food insecurity rates before and after the Great Recession,

respectively. Rates of food insecurity were generally higher in the South and West prior to

the recession. Most states saw an increase during the recession, as indicated by the darker

shading in most states comparing panel a to panel b. Figure 5c shows the distribution of the

increase in food insecurity. The increased rates do not appear to be particularly

concentrated in any particular region of the country, and the state of Nevada had the

largest increase during the recession. Below, we use geographic variation in the depth of

the recession in an attempt to better understand the causes of the sharp increase in food

insecurity.

8

II. Economic Indicators during the Great Recession

In this section we examine aggregate economic indicators over time to look for ways

in which the Great Recession is different from other periods, including its differences from

other recessionary time periods. Are there differences that, intuitively, would suggest that

food insecurity would be especially high during this recession?

Figures 6a-6e present data from the Federal Reserve Economic Data (FRED) series.4

As is well-known, Figure 6a shows the very high and sustained unemployment rates during

this recession. Although peak rates were higher in the 1980 recession, they declined more

rapidly. The Great Recession is also distinguished by very long unemployment durations.

Figure 6b shows that median duration of unemployment has been dramatically higher than

in earlier downturns.

Price fluctuations have also been severe during the Great Recession. Figure 6c

shows the decline in housing prices that has been a focal point of the Great Recession.

While housing prices and unemployment have been thoroughly discussed during this

recession, food prices have received relatively less attention. Inflation overall has been low,

as is often the case in periods of economic contraction, however, food prices have been

relatively high. Figure 6d shows the CPI for food scaled by the “core” CPI, the price index

less prices for food and energy, which tend to be more volatile. Energy prices scaled by core

CPI (not reported) show a similar pattern. Thus, although core inflation has been low, price

levels and growth for food and energy have been relatively high compared to all earlier

periods for which we have data, except for the downturn in 1980. Energy, while not

directly related to food consumption, is a necessity and tends to be inelastically demanded

4 http://research.stlouisfed.org/fred2/

9

and to take up a larger share of the budgets of low-income families. Thus high energy prices

may have in important indirect effect on food security, while high food prices, obviously,

can directly affect a family’s ability to secure adequate amounts of food.

Finally, Figure 6e presents data on Consumer Sentiment. The Great Recession saw a

precipitous drop in consumer sentiment to levels on par with the recession of the early

1980s.

In sum, the Great Recession is a period of high unemployment rates with very high

unemployment duration. At the same time, many of those who owned homes saw the

prices of their homes plummet (and thus their ability to use the equity in their homes to

weather a period of sustained unemployment fall). In addition, prices for necessities like

food and energy were relatively high, leading to a particularly strained budget. Taken

together, these macroeconomic facts may be contributors to the increase in food insecurity

during the Great Recession.

III. Differences in Characteristics of Food Insecurity Households during the

Recession

Figure 2 above suggests that food insecurity is not only higher than in previous

periods, but that its relationship with income has changed. In this section we more formally

test whether the relationship between food insecurity and other measures of household

characteristics or aggregate economic health have changed.

10

a. Changing Relationship between Economic Indicators and Food

Insecurity

In this section we first use the CPS-FSS data and data on poverty and unemployment

to examine the relationship between food insecurity and macro economic indicators to

determine the extent to which the relationship between FI and the macro economy appears

to be different during the Great Recession.

First, we use the U.S. annual poverty rate to predict overall food insecurity or one of

its components in a linear regression framework. We then test whether the relationship

between poverty and food insecurity is different during the Great Recession by taking the

residuals from that regression and running them on an indicator for the Great Recession

years (2008 forward). If that indicator is statistically significant, it suggests that the

relationship between poverty and food insecurity during the Great Recession is, indeed,

“different” from other periods. We follow this two-step process to ensure that we allow the

macroeconomic indicator (the poverty rate in this case) to have the first crack at explaining

food insecurity, before the more flexible control for the later years.

We then repeat the exercise with other measures that capture economic conditions:

the U.S. unemployment rate, plus the state-level unemployment and poverty rates. Our

primary outcome is an indicator of 12-month food insecurity status, but we also separately

investigate each of the 18 elements that go into constructing food insecurity measures for

adults and children and report a subset of these findings. These regressions are run using

pooled data from 1998-2010, merged with measures of unemployment and poverty.5 We

use one observation per household, weighted by the household supplement weight, and

5 Unemployment measures come from the BLS, LAUS series (http://www.bls.gov/lau/). The poverty measures come from the Census Bureau’s SAIPE series (http://www.census.gov/did/www/saipe/).

11

adjust the standard errors to allow for arbitrary forms of heteroskedasiticty and for

correlation within calendar year across households.

Consider the first column of Table 2. This indicates that, indeed, as the U.S. poverty

rate increases, so does the food insecurity rate. A one-point increase in the poverty rate

increases the likelihood that a household is food insecure by 1.14 percentage points.

Column 2 then uses the residual from this regression as the dependent variable and

regresses that against a dummy for the Great Recession. The statistically significant

(p=5.6%) coefficient of 0.0139 indicates that food insecurity is about 1.4 percentage points

higher than expected during the Great Recession. In other words, the fact that poverty

increased during the Great Recession explains a portion of why food insecurity is higher

during this time period. However, there is an additional 1.4 percentage points of “excess”

food insecurity during this period after accounting for the historical relationship between

poverty and food insecurity.

The third and fourth columns repeat that exercise, but use the U.S. unemployment

rate as the explanatory variable. Recall that unemployment is particularly high during this

period. When we run the residual from this regression against the indicator for the Great

Recession, we no longer find that the Great Recession has higher than expected food

insecurity, given the high unemployment rate. The coefficient on the dummy variable in

this regression is much smaller (0.005 vs. 0.014) and is not statistically different from zero

at anything close to standard significance levels.6

In column 5 we run a “horse race” between U.S. poverty and U.S. unemployment

rates in terms of “explaining” food insecurity. Once the unemployment rate is included,

6 Note, though, that the 95% confidence interval for this estimate includes not only zero, but also values larger than the 0.014 estimate found when using poverty in the first stage.

12

poverty rates are no longer statistically significantly correlated with food insecurity. The

relationship between the Great Recession indicator and the residual of this regression is

nearly identical to when unemployment is entered on its own. Columns 7-10 repeat this

exercise using state-level data poverty rates in the regression. In other results, not shown

here, we have included other local measures of poverty and unemployment, and even

measures of whether the household itself is poor. No matter which of these other measures

are included in the regression, national unemployment rates are always significantly

related to food insecurity rates, and once they are included, the Great Recession indicator is

no longer significantly related to the residual.

We performed identical exercises for each sub-component of food insecurity, and

find results that are largely consistent with those described for the overall measure of food

insecurity. In general, the higher prevalence food insecurity items are those that are

aberrantly high given the poverty rates during the Great Recession. Some of the very low

prevalence outcomes, such as children going a whole day without eating, are very difficult

to explain with either poverty or unemployment rates during any period. For the sake of

brevity, we show results from the specifications like those in the top panel of Table 2 for

only the highest and lowest prevalence item for adults and children. Thus, the top panel of

Table 3 shows the results for an indicator that the household respondent reported that he

or she was often “worried whether our food would run out before we got money to buy

more,” the highest prevalence item in the food security battery. The second panel uses as a

dependent variable an indicator for whether the respondent and other adults in the

household ever did not eat for a whole day. The top panel shows that households report a

higher level of worry about food insecurity than one might expect given the poverty rate,

13

but once the national unemployment rate is included in the regression, the Great Recession

is not a period of “excess” worry. For the lower prevalence indicator of more severe food

insecurity, the Great Recession does not appear to be aberrantly high given either the

poverty or the unemployment rate. As column 5 shows, even when poverty rates are held

constant, the unemployment rate continues to have a statistically significant association

with these items of food insecurity. Similarly, for children’s food insecurity items in Table 4,

the Great Recession has aberrantly high rates of relying “on a few kinds of low-cost food to

feed the children because we were running out of money to buy food” given the poverty

rate. However, it is not aberrantly high given the unemployment rate. Finally, panel 2 of

Table 4 shows that neither variation in the national poverty rate, nor variation in the

national unemployment rate, can explain the incidence of children going a whole day

without eating.

We find that a strong relationship between the unemployment rate and food

insecurity holds both before and after the Great Recession, and that the high

unemployment rate during the Great Recession predicted food insecurity before it predicts

poverty. That is, food insecurity appears to be striking individuals who do not, or do not

yet, meet the official criteria for being poor. On the other hand, the increase in food

insecurity conditional on income in Figure 2 suggests that the relationship between

unemployment and food insecurity is not solely working through a decrease in income. It is

clear, then, that further investigation into the role of the Great Recession on food insecurity

is necessary. We continue this investigation in the next subsection, where we describe the

ways in which the characteristics of the food insecure have changed over this time period.

14

b. Changing Relationship between Household Characteristics and Food

Insecurity

We next examine whether characteristics – including demographics, employment

outcomes, and safety net program participation – of food insecure and secure households

have changed during the Great Recession. This is a purely descriptive exercise that is

intended to give insights into the role played by the Great Recession in raising food

insecurity to such high levels.

In columns (1) through (4) of Table 5 we present means for a series of demographic

characteristics for both the food secure and the food insecure, separately for the time

periods before (1998-2007) and after (2008-10) the Great Recession. In column (5) we

show the relative change in means during the Great Recession for the food insecure versus

the food secure. This relative change is computed by first subtracting the mean (of a given

variable) for the food insecure prior to the Great Recession from the mean for the food

insecure during the Great Recession. We then calculate this same pre/post-Great Recession

difference for the food secure and subtract it from the previously calculated difference for

the food insecure. The last column shows the p-value indicating whether this relative

change is statistically meaningful.

Take female-headed households as an example. Among food secure households,

19.8 percent of households were headed by a female prior to the Great Recession, and the

percent edged up to 19.9 percent during the Great Recession. Among food insecure

households, 39.1 percent of households were headed by a female prior to the Great

Recession. During the Great Recession, more households became food insecure, driving the

percent headed by a female down to 37.2 percent. The change in rates of female-

15

headedness among food insecure relative to food secure households is -0.020 (p=0.002),

indicating that food insecure households are statistically significantly less likely to be

headed by a female during the Great Recession. In other words, food insecurity became

slightly less concentrated among the female-headed household population during the Great

Recession.

Turning to educational attainment, as expected, the food secure are less likely to be

high school dropouts than are the food insecure. However, the proportion of the food

insecure with the lowest levels of education fell, relatively, during the recession. The

proportion of the food insecure with a college degree or higher also fell, relatively, during

the recession. The other two remaining educational categories, some college and a high

school degree, increased their share among the food insecure. Thus, it appears that food

insecurity increased relatively more during the Great Recession for middle education

individuals.

Like the lower education groups, African Americans comprise a larger share of the

food insecure, but their relative share among the food insecure fell during the Great

Recession. Similarly, disabled individuals comprise about 3.5 to 4 percent of the food

secure group, while disability is much more common among the food insecure at 12 to 13

percent; again, the relative share of this group among the food insecure declined during the

Great Recession. Finally, the relative share of the food insecure who lived in an owner-

occupied home increased by about 2.5 percentage points during the Great Recession.

Figure 7 illustrates these demographic changes graphically. The circle represents

the relative change in the characteristic among food insecure households during the great

recession, surrounded by the 95 percent confidence interval on the change. The figure

16

shows that the food insecure have become statistically significantly relatively less likely to

be headed by a female (with or without children), by someone with less than a high school

diploma or college or more, or by an African American. The food insecure have become

relatively more likely to have a high school diploma or some college, be a homeowner, or be

white or Hispanic.

In sum, Table 5 and Figure 7 indicate that during the Great Recession food

insecurity became relatively more prevalent among groups that are historically better off

(such as those with intermediate levels of education, or who live in owner occupied

housing), and became relatively less prevalent among groups that are historically worse off

(such as female-headed households and African Americans). This finding that those

traditionally thought of as among the poor are relatively less represented among the food

insecure during the Great Recession seems consistent with the previous finding that the

national poverty rate did not do a very good job of predicting food insecurity during the

Great Recession.

Table 6 turns to examining reported labor force status among the food secure and

insecure. Although the probability of being employed has fallen during the Great Recession,

regardless of food security status, there has not been a significant relative change in the

probability of being employed. There has, however, been a relative shift in other

employment categories. For example, the relative probability of being unemployed has

risen by about 2.5 percentage points among the food insecure.7 The duration of

unemployment has also relatively risen for the food insecure. The probability of being “not

7 Note that this is from the individual labor force status questions in the month in which the FSS supplement was administered in that year. Since 2001, that has been in December – thus this number may differ substantially from the average annual unemployment rate.

17

in the labor force” has relatively declined among the food insecure, and the probability of

being self-employed has relatively increased. In sum, during the Great Recession, the food

insecure are more likely to be unemployed, unemployed for longer, and self-employed than

are the food secure, and they are less likely to be out of the labor force. This finding of a

relative increase in the number of the unemployed and the long-term unemployed within

the food insecure group is not inconsistent with the previous findings that the national

unemployment rate is an important predictor of food insecurity.

Finally, Table 7 summarizes the change in participation in food assistance programs

before and during the Great Recession. It is worth reiterating here that this is a purely

descriptive exercise. There is likely to be selection into who participates in these nutrition

programs so we cannot speak to, for example, counterfactual food security status among

those who do and do not participate in nutrition programs. We can see, however, that

participation in food stamps, and the monthly amounts received, increased relatively for

the food insecure during the Great Recession. Participation in WIC, for which eligibility is

more categorical, however, was unchanged.

Figure 8 expands this to investigate changes in participation among a wider range of

social programs. Food insecure households are relatively more likely to participate in SNAP

and the unemployment insurance program during the Great Recession, and less likely to

participate in Social Security, public assistance, or workers’ compensation. Relative

participation across a wide range of other programs – such as the free lunch program,

energy assistance, and supplemental security income – was unchanged.

Most striking here is the relative increase in SNAP participation among the food

insecure during the Great Recession, but interpreting these changes is difficult. Clearly, if

18

there had been no increase in SNAP receipt among those who qualify as food insecure, that

might have indicated either a lack of eligibility or a lack of knowledge of programs – both of

which are possible given the implication from the earlier tables that the food insecure

during the Great Recession appear to be coming from households that are better off (at

least in terms of education levels and home ownership). What one would really like to

know is whether SNAP participation and payments went up “enough” during the Great

Recession to compensate for the increase in need. We attempt to address this question in

more detail in the conclusion, below.

c. Factors that Predict the Increase in Food Insecurity: Oaxaca

Decompositions

Next we try to understand how much of the increase in food insecurity can be

explained by the demographic and economic factors we can measure. The Oaxaca

decomposition is a framework for understanding the determinants of an outcome change.

An outcome can vary both because on average the factors determining that outcome have

changed and also because the impact of those determinants on the outcome has changed.

The literature typically refers to this first component as being the change that is explained,

and the latter component as the unexplained change. Performing a Oaxaca decomposition

is straightforward. One regresses the outcome variable on its determinants separately for

the two groups – in this case, before and after the recession – and also calculates the means

of the explanatory variables separately. In this case, we estimate one regression predicting

the probability that a household is food insecure on the pre-Great Recession sample, and a

19

parallel regression on the post-Great Recession sample.8 Under the assumption that

nothing changed over time other than the means of the explanatory variables, we can

create a counterfactual, predicted food insecurity measure that is pre-GR*Xpost-GR and

subtract it from the actual post-Great Recession outcome to determine how much of the

observed change was predicted by changes in the covariates. Similarly, we can calculate the

“unexplained” portion as how much of the observed change was due to the changing impact

of the determining factors by subtracting the estimated coefficients (GR - pre-GR) and

multiplying the difference by the post-recession levels. Adding these two differences

together yields the overall difference in food insecurity post-Great Recession minus pre-

Great Recession.

Table 8 presents the results of this exercise. Using a sample of households with

income below 300 percent of the poverty line, we use a linear probability model to predict

the probability of being food insecure using a set of income/poverty bins, a set of

household labor force variables, household head dummies for being a homeowner, a set for

race/ethnicity, for being foreign born, and a set for completed education, household

composition dummies, and state-year variables on housing: the vacancy rate and a price

index. Overall for this lower income sample (with non-missing covariates) the rate of food

insecurity increased 4.9 percentage points, from 20.5 percent pre-Great Recession to 25.4

percent post-Great Recession, as can be seen in the top row of the table. Also shown in the

top row is that overall, only 1.2 percentage points of the total can be explained by changes

8 While it is not strictly necessary to maintain balanced group sizes, we drop the earliest years of our sample, as well as 2007 to use similar four-year periods for our groups. Thus, from 2003 to 2006 is the pre-GR period and from 2008-2011 is the post-GR period.

20

over time in the means of the covariates, with the other 3.7 remaining unexplained (i.e. due

to changes in the relationships between the covariates and food insecurity).

Looking more closely at the individual covariates, we see that almost half of the total

explanatory power comes from the labor force variables, which explain 0.53 percentage

points. The main source of this explanatory power is the increase in the annual weeks

spent looking for work by adults in the household, which accounts for 0.43 percentage

points. The declines in weeks and hours worked account for the remainder. The next

largest source of explanatory power comes from the increase in the state housing vacancy

rate, which explains 0.46 percentage points of the overall change. Finally, the drop in the

likelihood that the household owns its home explains another 0.31 percentage points.

These three determinants alone account for a bit over the 1.2 percentage points that can be

explained, meaning changes in all of the other covariates combined would have predicted a

decrease in food insecurity. This is most notable for education, where the increase in

college educated household heads would have predicted a 0.2 percentage point drop in

food insecurity. Interestingly, increases in the fraction of the population in lower

income/poverty bins does not help explain the increase in food insecurity very much,

contributing 0.2 percentage points to the total 4.9 point increase.9

Turning to the unexplained column, we see that the vast majority of the increase is

truly unexplained. That is, it is due to the constant increasing by just over 2 percentage

points. The largest contribution by an actual covariate is the 1.55 percentage point increase

in the relationship between housing prices and food insecurity, due to higher housing

prices becoming less protective of food insecurity, followed by another 1.4 points due to

9 Note that the sample composition (households below 300 percent of the poverty line) before and after the Great Recession will also change, and this change is not captured by our decomposition.

21

changes in the correlation to race. Being white or Hispanic both become less protective

against food insecurity. Finally, the changing role of income contributes another 1.1

percentage points. Here, it is mainly the middle-income bins becoming much less

protective, with the change for the 180 to 200% of the poverty line group implying almost a

3 percentage-point increase in food insecurity. At the same time, many covariates became

more protective (or less deleterious) over time, implying a prediction of lower food

insecurity rates. Overall, while the Oaxaca decomposition can only explain about a quarter

of the increase in food insecurity over the Great Recession (i.e. 0.012/0.049=0.245), it does

reinforce the idea that changes in the labor and housing markets were important aspects of

the changes engendered by the Great Recession.

d. Changing Transitions in Food Insecurity Status

We are able to observe a household’s food insecurity status across two consecutive

years, and in some cases we can also link this to detailed information about income and

program participation, and we can use this information to see how transitions into and out

of food insecurity have changed during the Great Recession. In particular, respondents to

the Current Population Survey are interviewed for four consecutive months, off for eight

months, and then interviewed again for four months. Food security status is collected in the

December interview, and information on income and program participation is collected in

March, so for individuals for whom the first interview is conducted in December can be

matched to their fourth interview in March, and followed across two years. We use this

subset of data to investigate transitions across food security status between years, both

before the Great Recession and after. With these two years of data, we can assign each

22

household to one of four mutually exclusive food security transition categories: never food

insecure, always food insecure, became food insecure, and left food insecurity. Figure 9

illustrates the differences in this categorization before and after the Great Recession. Note

that this figure is restricted to households with incomes less than 300 percent of the

poverty line.10

As Figure 9 shows, prior to the recession about half of households that experienced

food security in the prior year were no longer food insecure the following year. Prior to the

Great Recession, 73 percent of households remained food secure across both years, while

post-Great Recession less than 68 percent remained so, a drop of over five percentage

points. This change is driven in almost equal measure by an increase in households that

become food insecure and those that remain food insecure. There is almost no change in

the rate of leaving food insecurity – in both periods close to 10 percent of households

become food secure.

A potentially misleading aspect of this figure is that we do not take into

consideration household income, and we know that incomes fell with the Great Recession.

Thus, Table 9 explores the impact of the Great Recession on the transition probabilities,

controlling for the income/poverty ratio using dummies for each 20-point bin.

Turning first to column (1), we still see a 5 percentage-point drop in the probability of

remaining food secure across both years. Similarly, columns (2) and (3) show the similar

sized increases in the probability of remaining food insecure and becoming food insecure,

at 2 and 2.5 percentage points, respectively. Finally, column (4) confirms that there is no

significant change in the probability that a household leaves food insecurity. Recall that

10 The full sample is qualitatively similar, but the large percentage of households who are never food insecure (85.3% pre- and 81.6% post-recession) somewhat overwhelm the other transitions in the figure.

23

Figure 2 showed an increase in food insecurity, conditional on the income/poverty ratio.

From these transition data, we conclude that this increase was driven by an increased flow

into food insecurity that was not matched with an increased outflow. Thus, our main focus

should be on aspects of the Great Recession that might cause a household to become food

insecure, even if their income was unchanged.

IV. Behavior and Consumption Changes Among Food Insecure Households

We next turn to whether behavior of individuals changed between food insecure

and secure households over the Great Recession. These behaviors include, for example,

consumption of fast food, and time use.

a. Changes in Food Consumption

Data from the National Health and Nutrition Examination Surveys (NHANES) can

provide insight on how food consumption and nutrition intake differ between the food

secure and insecure and whether these differences changed over the Great Recession. The

NHANES uses dietary recall diaries to collect information on food consumed, and provides

an analysis of the nutritional content of that food. It also collects information on where the

food has been consumed: in a restaurant, at home, in a feeding center, etc. The NHANES

data cover different periods than the CPS-FSS. In particular, the NHANES survey is in the

field for two-year cycles with data representative of each two-year period. We use data

from 1999-2000 through 2009-2010.

Turning to Table 10, on average, the food secure ate 3.9 meals per day, while the

food insecure ate 3.7, in the years prior to the Great Recession. While meals per day grew

for both groups during the Great Recession, the differences expanded, with the food

24

insecure consuming relatively fewer daily meals during the Great Recession. Despite this,

differences in total calories consumed and in Body-Mass-Index (BMI) show no statistically

significant relative change over the Great Recession years. Interestingly, there is a relative

increase in percent of meals consumed at home as well as an increase in those consumed at

“fast food” restaurants; both of these are statistically significantly different from zero.

Even though total caloric intake did not change, the composition of calories may

have. Results for nutrition intake are shown in Figure 10, Panel A. The circles indicate the

estimate of the change in (log) nutrition intake across a variety of macro- and

micronutrients for the food insecure during the Great Recession. The lines surrounding the

circles display the 95 percent confidence intervals. The results indicate a statistically

significant decline in the consumption of fat and protein among the food insecure during

the recession, with a magnitude of approximately 4 percent. There was no significant

change in consumption of carbohydrates, though the point estimate is positive. Among

micronutrients, we estimate a decline in consumption of both “healthier” nutrients (fiber

and iron), and less-healthy ones (cholesterol, saturated fat, and sugar).

Panel B illustrates the potential role of food price increases on the shift in

nutritional intake. We use data from the USDA’s Center for Nutrition Policy and Promotion,

which provided 2003-04 average national prices by USDA food code for approximately

4600 foods. We merge these prices to the dietary data in the NHANES, and use the CPI for

food overall, to create inflation-adjusted prices for all years. Note that this does not allow

us to separate particular foods for which the price increases were highest during the Great

Recession, but instead because of data limitations we hold the relative prices across foods

constant. The results suggest that the food insecure are spending less per calorie during the

25

recession. They are also spending less per gram of saturated fats and per milligram of

cholesterol, and more per gram of fiber. Results for other nutrients are imprecise and not

statistically different from zero.

b. Changes in Time Use

Changes in how Americans spent their time during the Great Recession, and

whether this has changed differentially for the food insecure, may provide additional useful

insight. To investigate this, we turn to the American Time Use Survey (ATUS), which is a

nationally representative dataset that collects information on how, where, and with whom

respondents spend their time. Respondents are a randomly chosen subset drawn from

households that have completed their final CPS monthly survey response. In order to match

a household’s time use to its food insecurity status, a household must have participated in

the December Food Security Supplement. Since the ATUS is asked between 2 and 5 months

after a household completes its final CPS survey, the households that participated in a

December CPS were surveyed for the ATUS between the months of February and August.

The adult respondent is surveyed about his or her activities sequentially, walking through

the 24-hour period that began at 4 a.m. on the designated day and continued through 3:59

a.m. on the following day. Respondents describe in their own words the primary activity in

which they were engaged at each point in the day, and these activities are coded into

categories. While we primarily show results across the major groupings (e.g. eating and

drinking; working; household services), we also break out some activities such as food

preparation and food shopping in more detail. When the data are pooled across 2002-2010

December CPS data that can be linked to the ATUS, we have a sample of 36,544

respondents. Note that the results are an average across weekdays and weekends, and are

26

only for one adult in the household.

Table 11 presents simple regression results on time use controlling for food security

status, whether the interview was conducted during the Great Recession, and the

interaction between the two. The constant in each column represents the average amount

of time spent on the category among the food secure prior to the Great Recession, and the

coefficient on the indicator for food insecurity estimates the difference in mean time use for

food insecure households prior to the Great Recession. The coefficient on the Great

Recession indicates the change in time use for all households during that time period, and

the interaction term measures whether time use changed differentially during the Great

Recession for food insecure households.

Food insecure households spend more time on personal care, and the difference is

primarily driven by more time spent sleeping. Furthermore, the mean time spent sleeping

increased statistically significantly by an average of 8 minutes during the Great Recession.11

During the recession the mean time spent working declined by 7 minutes per day

(averaged across weekdays and weekends), and the likelihood of reporting any doing any

work on the interview day declined by 1.4 percentage points. Food insecure households

report less time and a lower likelihood of work overall, but there is no significant

differential impact of the recession. During the recession, Americans spent slightly less

time on average engaged in educational activities, shopping, doing household activities

(such as cleaning), or traveling, and spent more time in leisure. Again, there are differences

on average for food insecure households – they typically spend more time on education,

11 Note that increased time reported sleeping can be correlated with depression (Tsuno et al. 2005).

27

leisure, and caring for household members, and less time shopping or traveling – but no

differential changes during the Great Recession.

The Great Recession had relatively modest impacts on food preparation and

consumption patterns among households overall (columns 11-13), with no change in time

spent eating and drinking, nor in time spent preparing food, but a 3 percentage point

increase in the likelihood that the respondent reported any time spent in food preparation.

Food insecure households generally spend less time eating and drinking, and more time in

food preparation, but their time use was not differentially changed during the Great

Recession.

Overall, given the pattern in the earlier findings that the food insecure looked

different across a variety of characteristics during the Great Recession, it is surprising that

there are no statistically significant differences along those lines in the time use data. Part

of this could be driven by the relatively small sample size and lack of statistical precision.

V. The Potential Roles of Macroeconomic Conditions in Explaining Food

Insecurity

As described in Section II above, the Great Recession differed from other

recessionary periods and was characterized by high and long-duration unemployment, and

collapse of the housing and credit markets. Below we test to what extent these factors may

have directly contributed to the increase in food insecurity.

a. Potential Role of Leverage

A household’s financial leverage is measured as its debt to income ratio, and there

was a sharp drop in leverage during the Great Recession. As can be seen in Figure 11,

28

financial obligations relative to disposable income (which is closely related to leverage)

was growing throughout 2006, before dropping sharply in 2007, rising a bit in 2008, and

then dropping continuously. Recent work by Mian and Sufi (2010) has shown that

geographical patterns in household leverage in 2006 is a good predictor of the severity of

the Great Recession, explaining such things as house prices, unemployment, and durable

goods consumption patterns. While they base their findings on county-level leverage and

outcomes, we are only able to explore this idea at the state level, since that is the lowest

level of geography consistently defined in the CPS. As can be seen in Figure 12, though,

there is a good deal of variation in household leverage across states.12 The map delineates

quintiles of state leverage in 2006, with the most leveraged states appearing in the darkest

colors. On the assumption that household credit was financing consumption, the

deleveraging that occurred would reduce consumption. To the extent that consumption of

food is similarly affected, pre-recession leverage could help explain the increase in food

insecurity over the Great Recession.

To that end, Figure 13 plots food security over time by the highest and lowest state

leverage quintiles, indexed to 2006. While we see an increase in food insecurity for both

groups in 2008 at the start of the Great Recession, the jump up is larger and more

persistent for households in states with the highest levels of leverage. While this graph is

suggestive of an important role for leverage in worsening the impact of the Great Recession

on household’s food security status, it ignores the role of income. Thus, to further explore

this issue, we use a probit model to estimate whether a household is food insecure as a

12 We follow Mian and Sufi in defining leverage as the sum of per capita auto, mortgage and credit card debt for a state (from the New York Fed’s Equifax data), multiplied by state population and divided by state adjusted gross income (from IRS tax files).

29

function of income/poverty bins, a Great Recession dummy, a highest state leverage

quintile dummy, and the interaction. We estimate this model both for the entire sample (i.e.

the sample represented in the figure) and for the sample limited to households below 300

percent of the poverty line.

Table 12 shows the results of this exercise, presenting the marginal effects from the

probit model, along with standard errors that are robust to both heteroskedasticity and

correlation in the error term within states. Looking first at model (1), with results for the

full sample, we see that even controlling for the income/poverty ratio, there is a

significantly positive effect of the Great Recession on the probability that a household is

food insecure, raising it by almost 3 percentage points. While the interaction with being in

the highest leverage quintile state is positive, it is not significantly different from zero when

performing a two-sided test. Our null hypothesis, however is that the interaction term

should be positive, making the effect significant at the 10 percent level for this one-sided

test. Note that while this point estimate is relatively small, implying an additional increase

in the probability of being food insecure over the Great Recession of 1 percentage point,

relative to the main effect of the Great Recession is an increase of over 35 percent.

Turning to model (2), which restricts the sample to less well-off households with

incomes below 300 percent of the poverty line, we see a larger impact of the Great

Recession. For this group, the probability of being food insecure increased by almost

5 percentage points. While the point estimate of the interaction effect is similar to that from

model (1), the smaller sample size increases the standard error such that we cannot reject

a zero impact with even a one-sided test. Note that while the point estimate of the

additional effect of leverage in the Great Recession is similar across the columns, the larger

30

main effect in model (2) implies a smaller percentage impact. In this case, a 1 percentage-

point increase is under a 25 percent increase relative to the effect of the Great Recession in

less leveraged states.

b. Potential Role of the Housing Crisis

The Great Recession was marked by a steep decline in housing prices, although not

all areas of the country were equally affected. Figure 14 shows this variation by creating a

ratio of the average state housing price index (HPI) post-2007 to the 2006 HPI and

mapping the state quintiles. Lighter colors indicate sharper declines in housing prices. The

map makes clear that the housing price crash was especially severe in states like California,

Nevada, Florida and New Jersey. To the extent that homeowners may have been consuming

out of their housing value, the decline in housing prices during the Great Recession may

have affected homeowners, and homeowners in these states, especially hard.13 Figure 15

indexes food insecurity rates to 2006 for both homeowners and non-homeowners. The

effect of the Great Recession is clearly seen for both groups, but food insecurity jumps by

much more (an almost 50 percent increase) for homeowners than for non-homeowners.

An alternative way to think about the impact of the housing crisis is to focus on the

impact of the HPI ratio on these homeowners. Thus, Figure 16 repeats the exercise of

indexing food insecurity to 2006, but separately for the quintile with the biggest housing

price decline (quintile 1) and the smallest (quintile 5). Here we see that homeowners in

states most affected by the housing crisis have about a 70 percent increase in food

insecurity in 2008 relative to 2006, while homeowners in the less impacted states increase

less than 50 percent. While this figure is restricted to homeowners, a similar figure for the

13 Note that this consumption could be driven either by actual home equity loans, or simply from reduced savings rates based on the high value of housing assets.

31

full sample is qualitatively similar but somewhat muted, with only about a 30 percent

increase in food insecurity for the most affected group, and a 25 percent increase for the

least affected. Taken together, Figures 15 and 16 make clear that the housing crisis may

have had an important role to play in the increase in food insecurity over the Great

Recession. However, these figures ignore the role of income.

To investigate whether the housing crisis remains an important predictor of

changing food insecurity during the Great Recession, conditional on the income/poverty

ratio, we estimate probit models similar to those presented for leverage. Table 13 presents

the marginal effects from these probit models, along with standard errors that are robust

to both heteroskedasticity and within state correlation in the error term. Columns (1) and

(2) use the full sample, while column (3) is restricted to homeowners. Similarly, columns

(4) and (5) use all households with income below 300 percent of the poverty line, while

column (6) is restricted to homeowners from this lower income group. Columns (1) and (4)

investigate whether being a homeowner has a changing impact over the Great Recession on

the probability of being food insecure, while the other columns focus on the role of being

from a state in the quintile with the most severe housing price drop.

Looking first at columns (1) and (4), we see that conditional on income there is no

significant change during the Great Recession in the impact of being a homeowner.

However, being a homeowner is always very protective against being food insecure,

reducing the probability by 8.3 percentage points overall, and by 12.8 percentage points for

the lower income group. Additionally, the recession itself always increases food insecurity

– by 2.8 percentage points overall and 4.9 percentage points for the lower income. While

the impact of being a homeowner on food insecurity is unchanged over the Great

32

Recession, the effect of being in a state from the biggest housing price drop quintile is

significantly positive. This additional positive effect of the drop in house prices on the

probability of being food insecure is apparent for both the full sample and the lower

income sample and homeowners and non-homeowners alike.

Turning next to column (2), we again see the significant positive effect of the Great

Recession on the probability of being food insecure, conditional on the household’s

income/poverty ratio. In this case, though, the main 2.5 percentage point increase during

the Great Recession is augmented by another 1.6 percentage point increase for households

from states in the biggest housing price drop quintile. Note that overall, though, being from

such a state does not have a significant impact. Nonetheless, it implies a 64 percent

increase in the probability of being food insecure during the Great Recession (i.e.

0.016/0.025 = 0.64). Restricting the sample to homeowners adds an additional wrinkle.

Now, not only do we find a significantly positive main effect of the Great Recession and its

interaction with the housing quintile, we also find a significantly negative main effect of the

housing quintile. On the assumption that many of the states with the largest price drops

also saw large increases pre-Great Recession, this result is completely consistent with the

idea of households financing consumption with housing equity.14 The point estimates of

0.015 for the main effect of the Great Recession and 0.014 for its interaction with the

housing quintile implies the probability that a household is food insecure during the Great

Recession increases by almost 100 percent if they are from a state in the biggest housing

price drop quintile.

14 Recall that it is not necessary to actually take out a home-equity loan, one may simply reduce savings and increase consumption given the rise in assets (i.e. housing value).

33

Home ownership is perhaps higher in the lower income group than might be

expected, with over 50 percent of households with an income below 300 percent of the

poverty line reporting home ownership. The significance pattern of estimates for this

group, in columns (5) and (6) are identical to those for the overall group, while the point

estimate are larger (in absolute value) than those for the group overall. Note, however, that

the lower income group has a higher rate of food insecurity. In the pre-period it is 20.5

percent for the lower income group compared to 11.4 for the overall group, rising to 15.4

and 25.9 for each group respectively in the Great Recession. While homeowners have a

lower overall rate of food insecurity, the pattern is similar. It is 6.9 percent for the overall

group and 13.9 for the lower income group in the pre-period and rises to 9.4 percent and

17.8 percent for each group in the Great Recession.

The findings on the potential role of the housing crisis can be summarized as

follows. While in the aggregate it appeared that the impact of the Great Recession on food

insecurity was larger for homeowners than non-homeowners, this result does not hold

when conditioning upon income. However, it is true that even conditional upon income,

being from a state where the housing price drop was most precipitous implied a much

bigger increase in the probability of being food insecure during the Great Recession. For

homeowners, though, living in such a state was protective against food insecurity pre-Great

Recession, a finding consistent with consumption in that period being at least partially

financed via home equity. These results hold for both the overall sample and the sample

restricted to households with income below 300 percent of the poverty line.

34

c. Potential Role of Increased Unemployment Duration

As seen previously, unemployment duration reached lengths during the Great

Recession that were well above those seen in any previous recession, including during the

early 1980s, when the unemployment rate exceeded those of the Great Recession. As with

the credit and housing crisis, there was a large amount of geographic variation in

unemployment durations. Figure 17 maps the state quintiles of the ratio of 2009 average

unemployment duration to 2007 average unemployment duration, with the darker shading

reflecting larger increases in duration. Separating unemployment duration by quintile and

indexing it to its 2006 level, Figure 18 shows that, at the aggregate level, food insecurity

jumped in 2008 in both the highest and lowest unemployment duration quintiles, but fell

back in 2009 in the lowest quintile.15 As before, these graphs do not take into account

income – Figure 18 may simply reflect lower resource availability for households in the

highest quintile states. Thus, we again use a probit model to estimate whether a household

is food insecure as a function of income/poverty bins, a Great Recession dummy, a highest

state unemployment duration change quintile dummy, and the interaction, estimating the

model both for the entire sample (i.e. the sample represented in the figure) and for the

sample limited to households below 300 percent of the poverty line.

Table 14 presents the results from these models. For neither group does being in the

top quintile have a significant effect on food insecurity. This result confirms that there was

nothing different about these states prior to the Great Recession. After the Great Recession,

food insecurity increased in all states by 2.5 percentage points overall, and by 4.5 for

households below 300 percent of the poverty line. For those states where unemployment

15 Recall that quintiles are defined based on average 2009/2007 duration – it is possible that there was less difference in duration across these states in 2008.

35

duration increased the most (i.e. the highest quintile of 2009/2007 average unemployment

duration), the increase in food insecurity was significantly greater. The probability of being

food insecure increased another 1.6 percentage points overall for households from the

highest quintile states, and 1.9 percentage points for the lower income households from

these states. For the overall sample, that is a 64 percent (i.e. 0.016/0.025 = 0.64) increase

in the probability of being food insecure during the Great Recession from being in a high

duration state.

As can be seen in the maps above, many of the same states were high leveraged in

2006, and saw large drops in housing prices and increases in unemployment duration.

Thus, it is worthwhile to estimate our base model controlling for income bins and allowing

all three to have a differing impact on the probability of being food insecure over the Great

Recession. Table 15 presents the results of this exercise. As before, column (1) presents

marginal effects from a probit model estimating the probability of being food insecure

using the full sample, while column (2) is restricted to households with incomes below 300

percent of the poverty line. While none of the interactions with the Great Recession are

individually significant, they are highly significant in a joint test. Thus, a household in a

state in the highest quintile of 2006 leverage, the quintile with the largest drop in housing

prices, and the highest quintile of 2009/2007 average unemployment duration is predicted

to be 2 percentage points more likely to be food insecure over the Great Recession than a

household from a state that is in none of these quintiles. That household would simply be

predicted to have a 2.3 percentage point higher probability of being food insecure over the

Great Recession. Thus, being from a state that is most impacted by increasing leverage pre-

Great Recession, dropping home values and increasing unemployment durations means

36

being almost twice as likely to be food insecure during the Great Recession than being from

a less-impacted state.

d. Potential Role of Price Changes

While inflation was relatively low during the Great Recession (and even negative

during 2009), as we saw earlier, the price of food relative to non-food items was increasing.

To the extent that lower income (and hence potentially food insecure) households spend a

larger fraction of their budget on food, price changes could have a larger impact on these

families. Because we control for income using the income/poverty ratio, the base level of

inflation is already incorporated. However, it may be the case that a household at a given

income/poverty ratio has less purchasing power in the post-Great Recession period than it

did before. Our first attempt to explore the role of price changes uses the Chicago Fed

Income-Based Index (IBEX).16

The IBEX are designed to better capture the inflation experience of different

population groups that may have different consumption baskets. The IBEX group uses the

Consumer Expenditure Survey (CEX) to define consumption shares for different groups,

and prices these using data from the Bureau of Labor Statistics, coming up with a 12-month

inflation rate for each group. We create an index using the IBEX inflation for those with

income under 200 percent of the poverty line relative to IBEX inflation for all households,

and adjust household income in our sample using this measure. Thus, to the extent that the

IBEX determines that poor households are facing a higher inflation rate, we will consider

our households to have lower income. We then compare the increase in food insecurity

over the Great Recession conditional on the true income/poverty ratio with that

16 See https://www.chicagofed.org/webpages/research/data/ibex/index.cfm for more information.

37

conditional on this adjusted income/poverty ratio. Adjusting the income/poverty ratio in

this manner only reduces the estimated increase in food insecurity over the Great

Recession by 0.1 percentage points. The Great Recession leads to an overall increase in the

probability that a low-income household is insecure of 3.3 percentage points using true

income, and 3.2 percentage points using income adjusted for being in a low-income

population.17 Thus, we conclude that a higher inflation rate faced by lower income

households is not a major contributor to the impact the Great Recession had on food

insecurity.

VI. Discussion and Conclusions

This project presents evidence on how food insecurity has changed over the last

several years. Food insecurity is higher during the Great Recession than during any other

period for which we have national food insecurity information. There are many reasons

why the Great Recession might be expected to have particularly high food insecurity. In

particular, unemployment rates are high and unemployment duration is double what it was

even in the deep recession of the early 1980s. At the same time, housing prices have fallen

and many families, even those considered middle class, lost their ability to use their

housing equity to smooth periods of unemployment. In addition, food and energy prices

have been relatively high during this period. We investigated the role played by these

factors in the increase in food insecurity during the Great Recession.

17 The low income group used for the model estimation is households under 300 percent of the poverty line, controlling for 20 percentage point bins of income/poverty, as in our other models. The IBEX definition of low income group is households under 200 percent of the poverty line.

38

Much of the work in this report has been motivated by the puzzle presented in

Figure 2, which showed that food insecurity increased markedly over the Great Recession,

even within narrow 20-percentage point income to poverty bins. Having discovered an

important role played by deleveraging, the housing bust and increased unemployment

durations during the Great Recession, it is important to explore an alternate approach to

thinking about the shift in food insecurity. Rather than considering it to be a pure shift up,

conditional on income/poverty ratio, it could instead be thought of as a shift over. In the

pre-Great Recession time period when credit was flowing, house prices were rising and

unemployment was low, personal saving as a percentage of disposable income was quite

low, but it rose markedly during the Great Recession, as can be seen in Figure 19. As a

result, it is quite possible that consumption for a household in a given income/poverty bin

pre-Great Recession was inflated due to this low personal saving rate, returning to

“normal” during the Great Recession. Thus, if a household with, for example, income at 100

percent of the poverty line was consuming more like a household with income well above

the poverty line, the implication would be that the pre-Great Recession relationship

between income and food insecurity from Figure 2 really should be thought of as being

shifted further to the right, and thus closer to the post-Great Recession line.

By making some simplifying assumptions, we can estimate how much additional

income a household would need to have for the pre-Great Recession food insecurity line to

shift over completely to the post-Great Recession line. First, we limit our analysis to the

downward sloping portion of Figure 2, which is fairly linear, and thus estimate a linear

probability model predicting food insecurity as a function of income/poverty for the Great

39

Recession Period.18 We can now use this regression to shift the pre-Great Recession line to

the right – we simply need to see what value of income/poverty would predict the

observed rate of food insecurity pre-Great Recession, using the estimated regression. To

translate the income/poverty bins from Figure 2 into incomes, we assume the 2014

poverty line for a family of four of $23,850. The result of this exercise is shown in Table 16,

where we present the weekly increase in 2014 dollars that a family of four would have

needed to shift the Pre-Great Recession food security line in Figure 2 to the right far

enough to be equivalent to the Post-Great Recession line. The amount of income required

to accomplish the shift varies by income-to-poverty, but around the poverty line it is about

$170 per week.

As a comparison, we also calculated a simple 3 percent income increase, again

presented in 2014 dollars for a family of four. We chose 3 percent because in Figure 19

saving as a percent of disposable income was 3.2 percent in October 2006, but had risen to

6.2 percent by June 2011 – an increase of 3 percentage points. Thus, focusing purely on a

change in consumption due to an increase in saving would imply the income increase

shown. The final column presents this 3 percent increase as a percentage of the total

weekly increase needed to account for the entire horizontal shift in the relationship.

Focusing more closely on Table 16, it is clear that, for lower-income families,

thinking of the rightward shift as being purely due to a change in the saving rate is

generally not sufficient to move the pre-Great Recession line completely over to the post-

18 The regression estimated for households between 40 and 300 percent of the poverty line is Probability of food insecurity = 0.45 – 0.0012(income/poverty).

40

Great Recession line.19 For those below about 180 percent of the poverty line, though, a

3 percent change in income will only account for about a quarter of the difference. It is

important to realize that this calculation ignores the possibility of home equity loans

and/or increasing credit card debt, since it focuses purely on the change in savings. Note

additionally that per-capita spending on food stamp benefits increased by $18 per week

between 2006 and 2010, suggesting that the relationship would have been shifted even

further without these payments.

In Section III, we find that the factors that had previously been protective against

food insecurity have lost some of their predictive power. We also find that while the

increase in the unemployment rate can explain the increase in food insecurity, the

relationship works through more than just the decline in income. It appears that the

unemployment rate may be serving as a proxy for other factors in the macro economy that

were particularly harmed during the Great Recession, and as we find in Section VI, declines

in rates of leverage and housing prices appear to play an important role in the explanation.

Overall, we find that food insecurity has been aberrantly high during the Great

Recession. Across a wide range of evidence, the best explanation we can find for this

increase is that the overall rate of consumption relative to income dropped during this

period. The reasons for this decline are not entirely clear, and without individual-level data

linking food insecurity, consumption patterns and more detailed financial assets, it will be

hard to give more precise explanations with existing data. Nonetheless, the pattern is

consistent with explanatory factors that include psychological factors, an increase in

saving, and a loss of credit.

19 Interestingly, though, for those at the higher end of the income/poverty scale, the implied shift in savings is actually more than enough to explain the change in food insecurity.

41

REFERENCES Alaimo, K., C.M. Olson, E.A. Frongillo Jr., and R.R. Briefel, “Food Insufficiency, family income, and health in U.S. preschool and school-aged children,” American Journal of Public Health, vol. 91, iss. 5, 2001. Alaimo, Katherine, Christine M. Olson, and Edward A. Frongillo Jr., “Food Insufficiency and American School-Aged Children’s Cognitive, Academic, and Psychosocial Development,” Journal of American Academy of Pediatrics, vol. 108, no. 1, July 2001. Almond, Douglas, Hilary Hoynes and Diane Whitmore Schanzenbach, “Inside the War on Poverty: The Impact of the Food Stamp Program on Birth Outcomes,” Review of Economics and Statistics, May 2011. Andrews, Margaret, Mark Nord, Gary Bickel, and Steven Carlson, “Household Food Security in the United States, 1999,” FANRR-8, USDA, Economic Research Service, 2000. Available at: www.ers.usda.gov/publications/fanr8 Anderson, Patricia M., Kristin F. Butcher, and Phillip B. Levine, “Maternal Employment and Overweight Children,” Journal of Health Economics, vol. 22, no. 3, May 2003, pp.477-504. Anderson, Patricia M., Kristin F. Butcher, and Diane Whitmore Schanzenbach, “Childhood Disadvantage and Obesity: Is Nurture Trumping Nature?,” in Jonathan Gruber (ed.), An Economic Perspective on the Problems of Disadvantaged Youth, Chicago University of Chicago Press, 2009. Bhattacharya, Jayanta, Janet Currie, and Steven Haider, “Breakfast of Champions? The School Breakfast Program and the Nutrition of Children and Families,” Journal of Human Resources, vol. 41, no. 3, pp. 445-466, 2006. Bhattacharyaa, Jayanta, Janet Currie, and Steve Haider, “Poverty, food, insecurity, and nutritional outcomes in children and adults,” Journal of Health Economics, vol. 23, iss. 4, pp. 839-862, July 2004. Bickel, Gary, Mark Nord, Cristofer Price, William Hamilton, and John Cook (2000), “Guide to Measuring Household Food Security, Revised 2000,” U.S. Department of Agriculture, Food, and Nutrition Service, Alexandria, VA. Bitler, Marianne and Hilary Hoynes (2010). “The State of the Safety Net in the Welfare Reform Era,” Brookings Papers on Economic Activity, Fall 2010, pp. 71-127. Borjas, George J. “Food Insecurity and Public Assistance,” Journal of Public Economics, vol. 88, no. 7-8, pp. 1421-1443, July 2004. Burman, Leonard E., and Deborah I. Kobes, “EITC Reaches More Eligible Families than TANF, Food Stamps,” Tax Notes, Tax Policy Center, March 17, 2003.

42

Casey, Patrick H., Kitty Szeto, Shelly Lensing, Margaret Bogle, and Judy Weber, “Children in Food-Insufficient, Low-Income Families,” Archives of Pediatrics & Adolescent Medicine, vol. 155, no. 4, Apr 2001. Cristofar, S, and P Basiotis, “Dietary Intakes and Selected Characteristics of Women Ages 19-50 Years and their Children Ages 1-5 Years by Reported Perception of Food Sufficiency,” Journal of Nutrition Education 24:53-58, 1992. DeLeire, Thomas, and Ariel Kalil, “How do Cohabiting Couples with Children Spend their Money?” Journal of Marriage and Family, vol. 67, no. 2, pp. 286-295, May 2005. Dietz, William, “Does Hunger Cause Obesity?” Pediatrics 95(5): 766-767, May 1995. Dinour, Laura M., Dara Bergen, Ming-Chin Yeh, “The Food Insecurity-Obesity Paradox: A Review of the Literature and the Role Food Stamps May Play,” Journal of the American Dietetic Association, vol. 107, iss. 11, pp. 1952-1961. November 2007. Gluckman, Peter, and Mark Hanson, The Fetal Matrix: Evolution, Development and Disease. Cambridge: Cambridge University Press, 2005. Gundersen, Craig, “Measuring the Extent, Depth, and Severity of Food Insecurity: An Application to American Indians in the USA,” Journal of Population Economics, vol. 32, no. 2, pp. 292-325, Jan 2008. Gundersen, Craig, and Victor Oliveira, “The Food Stamp Program and Food Insufficiency,” American Journal of Agricultural Economics, vol. 83, no. 4, pp. 875-887, Nov 2001. Gundersen, Craig, and James Ziliak, “The Role of Food Stamps in Consumption Stabilization,” Journal of Human Resources, vol. 38, pp. 1051-1079, 2003. Guryan, Jonathan, Erik Hurst, and Melissa Kearney, “Parental Education and Parental Time with Children,” Journal of Economic Perspectives, vol. 22, no. 3, Summer 2008. Hoynes, Hilary, Diane Whitmore Schanzenbach and Douglas Almond, “Long Run Impacts of Childhood Access to the Safety Net,” Mimeo, UC Davis, 2012. Kaiser, Lucia L., and Marilyn Townsend, “Food Insecurity Among U.S. Children: Implications for Nutrition and Health,” Food Insecurity and Special Populations, vol. 20, iss. 4, pp. 313-320, 2005. Krueger, Alan B., and Andreas Mueller, “The Lot of the Unemployed: A Time Use Perspective,” Princeton University Industrial Relations Section Working Paper, no. 524, May 2008.

43

Krueger, Alan B. & Andreas Mueller, “Job Search and Unemployment Insurance: New Evidence from Time Use Data,” Journal of Public Economics, vol. 94, no. 3-4, pp. 298-307, 2010. Levy, Helen and Thomas DeLeire, “What Do People Buy When They Don’t Buy Health Insurance and What Does that Say about Why They Are Uninsured?” Inquiry 45(4): 365-379, Winter 2008/09. Martin, Katie S. and Ann M. Ferris, “Food Insecurity and Gender are Risk Factors for Obesity,” Journal of Nutrition Education and Behavior, 39(1): 31-36, Jan-Feb 2007. Mian, Atif and Amir Sufi (2010). “Household Leverage and the Recession of 2007-2009,” IMF Economic Review 58: 74-117. Nord, Mark, Margaret Andrews, and Steven Carlson, “Household Food Security in the United States, 2008,” Economic Research Report number 83, Economic Research Service, USDA, November 2009. Nord, Mark, Alisha Coleman-Jensen, Margaret Andrews, and Steven Carlson, “Household Food Security in the United States, 2009,” ERR-108, Economic Research Service, USDA, 2010. Available at: http://www.ers.usda.gov/Publications/ERR108/ERR108.pdf Ohls, James, Larry Radbill, and Allen Schirm, “Household Food Security in the United States, 1995-1997: Technical Issues and Statistical Report, prepared by Mathematica Policy Research, Inc., for USDA, Food and Nutrition Service, 2001. Available at: www.fns.usda.gov/oane/MENU/Published/FoodSecurity/FoodSecurityTech.pdf/ Rose, Donald, and J. Nicholas Bodor, “Household Food Insecurity and Overweight Status in Young School Children; Results From the Early Childhood Longitudinal Study,” American Academy of Pediatrics, vol. 117, no. 2, pp. 464-473, Feb 2006. Rose, D., and V. Oliveira, “Nutrient intakes of individuals from food-insufficient households in the United States,” American Journal of Public Health, vol. 87, iss. 12, 1997. Siefert, Kristine, Colleen M. Heflin, Mary E. Corcoran, and David R. Williams, “Food Insufficiency and Physical and Mental Health in a Longitudinal Survey of Welfare Recipients,” Journal of Health and Social Behavior, vol. 45, no. 2, pp. 171-186, Jun 2004. Wehler, Cheryl, Linda F. Weinreb, Nicholas Huntington, Richard Scott, David Hosmer, Kenneth Fletcher, Robert Goldbert, and Craig Gundersen, “Risk and Protective Factors for Adult and Child Hunger Among Low-Income Housed and Homeless Female-Headed Families,” American Journal of Public Health, Jan 2004. Winicki, Joshua, and Kyle Jemison, “Food Insecurity and hunger in the Kindergarten Classroom: Its Effect on Learning and Growth,” Contemporary Economic Policy, vol. 21, iss. 2, pp. 145-157, Apr. 2003.

44

Figure 1: Rates of Food Insecurity and Poverty, 1998-2012

Figure 2: Rates of Food Insecurity by Income-to-Poverty Group, Before and After the Recession

8.0

9.0

10.0

11.0

12.0

13.0

14.0

15.0

16.0

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012

poverty food insecurity

45

Figure 3: Percent of Households Answering Affirmatively to Adult Food Insecurity Items Panel A: High-Prevalence Adult Items

Panel B: Low-Prevalence Adult Items

10

15

20

Perc

ent of H

ouse

hold

s

1995 2000 2005 2010Adult Food Insecurity Items by Year

(mean) worried (mean) notlast

(mean) notbalanced

02

46

81

0P

erc

ent

of H

ousehold

s

1995 2000 2005 2010Adult Food Insecurity Items by Year

(mean) skip_adults (mean) often_skip_adults

(mean) eatless (mean) hungry

(mean) loseweight (mean) day_no_eat

(mean) often_day_no_eat

46

Figure 4: Percent of Households with Children Answering Affirmatively to Child Food Insecurity Items Panel A: High-Prevalence Child Items

Panel B: Low-Prevalence Child Items

12

34

56

Perc

ent of H

ouse

hold

s

1995 2000 2005 2010Child Food Insecurity Items by Year

(mean) lowcost_kids (mean) notbalanced_kids

(mean) notenough_kids

0.2

.4.6

.8P

erc

en

t of

Househo

lds

1995 2000 2005 2010Child Food Insecurity Items by Year

(mean) skip_kid (mean) often_skip_kid

(mean) cutmeal_kid (mean) hungry_kid

(mean) day_no_eat_kid

47

Figure 5: Food Insecurity Rates by State Panel A: Food Insecurity Rates Prior to the Great Recession

Panel B: Food Insecurity Rates After the Great Recession

48

Panel C: Change in Food Insecurity Rates, After vs. Before the Great Recession

Figure 6: Macroeconomic Indicators Panel A: Unemployment Rate

49

Panel B: Median Unemployment Duration

Panel C: Housing Prices

50

Panel D: Relative Consumer Price Index for Food

Panel E: Consumer Sentiment

51

Figure 7: Demographic Changes Among the Food Insecure During the Great Recession

Figure 8: Change in Program Participation Rates Among the Food Insecure During the Great Recession

52

Figure 9: Changes in Household Food Security Status Across 2-Year Periods, Before and After the Great Recession

Figure 10: Changes in Nutrient Intake Among the Food Insecure During the Great Recession Panel A: Log of Nutrient Intake

53

Panel B: Log of Price per Unit of Nutrient Intake

Figure 11: Average Leverage (Financial Obligations to Income) Over Time

54

Figure 12: Household Financial Leverage in 2006 by State

Figure 13: Food Insecurity Rates by State Leverage Quintiles

55

Figure 14: Changes in Housing Prices by State during the Great Recession

Figure 15: Food Insecurity Rates by Homeowner Status

56

Figure 16: Food Insecurity Rates by Quintile of Housing Price Decline

Figure 17: Increases in Unemployment Duration by State During the Great Recession

57

Figure 18: Food Insecurity Rates by Quintile of Unemployment Duration Increase

Figure 19: Personal Saving as a Percentage of Disposable Income

58

Table 1: Questions Included in the Food Security Scale

Guide to Measuring Household Food Security – 2000

23

Exhibit 2-2

QUESTIONS INCLUDED IN THE FOOD SECURITY SCALE

Question

Number* Q uestion

Optional Preliminary Screen

Stage 1:

Q2 Now I’m going to read you several statements that people have made about their food

situation. Please tell me whether the statement was often, sometimes, or never true in

the last 12 months.

“I worried whether our food would run out before we got money to buy more.”

Was that often, sometimes, or never true for you in the last 12 months?

Q3 “The food that we bought just didn't last, and we didn’t have money to get more.”

Was that often, sometimes, or never true for you in the last 12 months?

Q4 “We couldn’t afford to eat balanced meals.” Was that often, sometimes, or never

true for you in the last 12 months?

Q5**

“We relied on only a few kinds of low-cost food to feed the children because we

were running out of money to buy food.” Was that often, sometimes, or never true for

you in the last 12 months?

Q6**

“We couldn’t feed the children a balanced meal because we couldn’t afford that.”

Was that often, sometimes, or never true for you in the last 12 months?

1st-Level Internal Screen

Stage 2:

Q7**

“The children were not eating enough because we just couldn’t afford enough

food.” Was that often, sometimes, or never true for you in the last 12 months?

Q8,

Q8a

In the last 12 months, did you or other adults in your household ever cut the size of

your meals or skip meals because there wasn’t enough money for food?

How often did this happen — almost every month, some months but not every month,

or in only one or two months?

Q9 In the last 12 months, did you ever eat less than you felt you should because there

wasn’t enough money to buy food?

Q10 In the last 12 months, were you ever hungry but didn’t eat because you couldn’t

afford enough food?

Q11 Sometimes people lose weight because they don’t have enough to eat. In the last 12

months, did you lose weight because there wasn’t enough food?

59

Table 1 continued:

Source: Exhibit 2-2 in Bickel et al. 2000, “Guide to Measuring Household Food Security”

Guide to Measuring Household Food Security – 2000

24

The four kinds of situation are:

· Anxiety or perception that the household food budget or food supply was inadequate

(Q2, Q3); · Perceptions that the food eaten by adults or children was inadequate in quality (Q4,

Q5, Q6);

· Reported instances of reduced food intake, or consequences of reduced intake, for

adults (Q8, Q8a, Q9, Q10, Q11, Q12, Q12a); and · Reported instances of reduced food intake or its consequences for children (Q7, Q13,

Q14, Q14a, Q15, Q16).

Exhibit 2-2 (continued)

QUESTIONS INCLUDED IN THE FOOD SECURITY SCALE

Question

Number*

Q uestion

2nd-Level Internal Screen

Stage 3:

Q12,

Q12a

In the last 12 months, did you or other adults in your household ever not eat for a

whole day because there wasn’t enough money for food?

How often did this happen — almost every month, some months but not every month,

or in only one or two months?

Q13**

In the last 12 months, did you ever cut the size of any of the children’s meals

because there wasn’t enough money for food?

Q14,**

Q14a**

In the last 12 months, did any of the children ever skip meals because there wasn’t

enough money for food?

How often did this happen — almost every month, some months but not every month,

or in only one or two months?

Q15**

In the last 12 months, were the children ever hungry but you just couldn’t afford

more food?

Q16**

In the last 12 months, did any of the children ever not eat for a whole day because

there wasn’t enough money for food?

* See Appendix D, technical note 1 for comparison of core-module item numbers and CPS Supplement

numbers for the same items.

** Questions asked only of households with children. Children are defined as persons age 0-17--i.e.,

less than 18 years old--(but see note 22 in Endnotes).

60

Table 2: Estimating the Relationship between Food Insecurity and Aggregate Measures of Poverty and Unemployment

(1) (2) (3) (4) (5) (6)

VARIABLES Food

Insecure Residual

from Col. 1 Food

Insecure Residual

from Col. 3 Food

Insecure Residual

from Col. 5

Percent in Poverty, US 0.012 0.003

(0.002) (0.002)

p-value 0.000 0.119

Percent Unemployed, US 0.008 0.006 (0.001) (0.001) p-value 0.000 0.001 Indicator for Great Recession

0.014 0.006 0.005

(0.007) (0.005) (0.006)

p-value 0.056 0.323 0.399

Constant -0.032 -0.003 0.073 -0.001 0.038 -0.001 (0.020) (0.002) (0.004) (0.002) (0.021) (0.002)

Observations 574,009 574,009 574,009 574,009 574,009 574,009

R-squared 0.001 0.000 0.002 0.000 0.002 0.000

(7) (8) (9) (10)

VARIABLES Food

Insecure Residual

from Col. 7 Food

Insecure Residual

from Col. 8

Percent in Poverty, US 0.005 -0.003

(0.002) (0.002)

p-value 0.012 0.139

Percent in Poverty, State 0.007 0.007

(0.000) (0.000)

p-value 0.000 0.000

Percent Unemployed, US 0.006

(0.001)

p-value 0.001

Indicator for Great Recession

0.014 0.005

(0.007) (0.006) p-value 0.0565 0.399 Constant -0.030 -0.003 0.038 -0.001 (0.019) (0.002) (0.021) (0.001) Observations 574,009 574,009 574,009 574,009 R-squared 0.004 0.000 0.005 0.000

61

Table 3: Estimating the Relationship between Adult High and Low Food Insecurity Items and Aggregate Measures of Poverty and Unemployment

Panel 1: Worried that food would not last

(1) (2) (3) (4) (5) (6)

VARIABLES Worried Residual from Col.

1 Worried

Residual from Col.

3 Worried

Residual from Col.

5

Percent in Poverty, US 0.015 0.005 (0.002) (0.002)

p-value 0.000 0.047

Percent Unemployed, US

0.010 0.008

(0.001) (0.002) p-value 0.000 0.001

Indicator for Great Recession 0.016 0.005 0.005 (0.008) (0.007) (0.008)

p-value 0.080 0.481 0.560

Constant -0.030 -0.004 0.107 -0.001 0.055 -0.001 (0.025) (0.003) (0.008) (0.002) (0.023) (0.002)

Observations 592,358 592,358 592,358 592,358 592,358 592,358

R-squared 0.002 0.000 0.002 0.000 0.002 0.000

Panel 2: Often go a day without eating

(1) (2) (3) (4) (5) (6)

VARIABLES Often Day without Eating

Residual from Col.

1

Often Day without Eating

Residual from Col.

3

Often Day without Eating

Residual from Col.

5

Percent in Poverty, US 0.002 0.001 (0.000) (0.000)

p-value 0.000 0.001

Percent Unemployed, US

0.001 0.000

(0.000) (0.000)

p-value 0.000 0.005

Indicator for Great Recession 0.001 0.000 0.000 (0.001) (0.000) (0.001)

p-value 0.232 0.490 0.740

Constant -0.011 0.000 0.004 0.000 -0.006 0.000 (0.003) (0.000) (0.001) (0.000) (0.003) (0.000)

Observations 594,505 594,505 594,505 594,505 594,505 594,505

R-squared 0.000 0.000 0.000 0.000 0.000 0.000

62

Table 4: Estimating the Relationship between Children’s High and Low Food Insecurity Items and Aggregate Measures of Poverty and Unemployment

Panel 1: We relied on a few low-cost foods to feed the children….

(1) (2) (3) (4) (5) (6)

Variables Low Cost Foods for

Kids

Residual from Col. 1

Low Cost Foods for

Kids

Residual from Col. 3

Low Cost Foods for

Kids

Residual from Col. 5

Percent in Poverty, US

0.000 -0.003

(0.001) (0.001)

p-value 0.854 0.017

Percent Unemployed, US 0.001 0.002 (0.000) (0.001) p-value 0.085 0.012 Indicator for Great Recession 0.003 0.000 0.000 (0.002) (0.002) (0.001)

p-value 0.064 0.989 0.777

Constant 0.058 -0.001 0.051 0.000 0.081 0.000 (0.008) (0.001) (0.004) (0.001) (0.010) (0.001)

Observations 593,144 593,144 593,144 593,144 593,144 593,144

R-squared 0.000 0.000 0.000 0.000 0.000 0.000

Panel 2: Children ever not eat for whole day because not enough money….

(1) (2) (3) (4) (5) (6)

Variables Day w/o

eating for Kids

Residual from Col. 1

Day w/o eating for

Kids

Residual from Col. 3

Day w/o eating for

Kids

Residual from Col. 5

Percent in Poverty, US

0.000 0.000

(0.000) (0.000)

p-value 0.422 0.700

Percent Unemployed, US 0.000 0.000 (0.000) (0.000) p-value 0.489 0.778 Indicator for Great Recession 0.000 0.000 0.000 (0.000) (0.000) (0.000)

p-value 0.628 0.573 0.531

Constant 0.001 0.000 0.001 0.000 0.001 0.000 (0.001) (0.000) (0.000) (0.000) (0.001) (0.000)

Obs. 594,381 594,381 594,381 594,381 594,381 594,381

R-squared 0.000 0.000 0.000 0.000 0.000 0.000

63

Table 5: Differences in Characteristics Among the Food Secure and Insecure Before and During the Great Recession

(1) (2) (3) (4) (5) (6)

Food Secure Food Insecure Relative Change for

FI

P-value for Relative Change

Before GR

Great Recession

Before GR Great

Recession

Female 0.510 0.509 0.531 0.520 -0.009 0.002

Female Headed HH 0.198 0.199 0.391 0.372 -0.020 0.002

Female Headed HH w Kids

0.085 0.084 0.284 0.253 -0.030 0.000

Married 0.559 0.550 0.378 0.372 0.004 0.274

Less than High School 0.182 0.156 0.379 0.320 -0.033 0.000

High School Degree 0.301 0.282 0.328 0.336 0.026 0.000

Some College 0.261 0.271 0.223 0.256 0.024 0.000

College Plus 0.256 0.291 0.070 0.089 -0.016 0.000

African American 0.111 0.109 0.245 0.215 -0.029 0.000

Disabled 0.036 0.040 0.131 0.124 -0.010 0.001

Own Home 0.753 0.746 0.429 0.448 0.025 0.006

Notes: Data are from the 1998 to 2010 CPS-FSS. Observations: 1,516,065

64

Table 6: Employment Outcomes Among the Food Secure and Food Insecure Before and During the Great Recession

(1) (2) (3) (4) (5) (6)

Food Secure Food Insecure Relative Change for FI

P-value for

Relative Change

Before GR Great

Recession Before GR

Great Recession

Employed 0.638 0.604 0.520 0.483 -0.003 0.405

Unemployed 0.027 0.046 0.085 0.129 0.025 0.001

Not in Labor Force 0.336 0.351 0.396 0.388 -0.022 0.002

Duration of Unemployment

0.42 1.24 1.52 3.90 1.555 0.007

Discouraged Worker 0.005 0.007 0.016 0.020 0.002 0.017

Total Hours Worked Last Week

24.8 23.0 18.7 16.7 -0.243 0.102

Self Employed 0.073 0.069 0.035 0.044 0.012 0.000

Notes: Data are from the CPS-FSS 1998-2010. Observations: 1,191,368.

65

Table 7: Program Participation Among the Food Secure and Food Insecure Before and During the Great Recession

(1) (2) (3) (4) (5) (6)

Food Secure Food Insecure Relative Change for FI

P-value for

Relative Change

Before GR Great

Recession Before GR

Great Recession

Received SNAP in Last 30 Days

0.030 0.052 0.292 0.439 0.125 0.003

SNAP Amount in last 30 days

7.6 16.0 59.8 95.3 27.1 0.000

Received WIC in Last 30 Days

0.029 0.035 0.161 0.165 -0.002 0.615

Notes: Data are from CPS-FSS 1998-2010. Observations: 1,407,335.

66

Table 8: Factors that Explain the Food Insecurity Increase

(1) (2) (3) (4) (5) (6)

Coefficients Means Amount

explained Amount

unexplained

Pre-Recession

Great Recession

Pre-Recession

Great Recession

Food insecurity Rate/Totals 0.205 0.254 0.012 0.037

Change: 0.049

Explanatory Variables:

Intercept 0.369 0.390 1.000 1.000 0.0000 0.0214

20-40% of poverty line 0.003 0.022 0.027 0.030 0.0000 0.0005

40-60% 0.043 0.055 0.039 0.042 0.0001 0.0005

60-80% 0.038 0.041 0.062 0.064 0.0001 0.0002

80-100% 0.035 0.025 0.068 0.077 0.0003 -0.0007

100-120% -0.010 0.013 0.075 0.077 0.0000 0.0017

120-140% -0.038 -0.026 0.080 0.082 -0.0001 0.0010

140-160% -0.042 -0.016 0.082 0.073 0.0004 0.0019

160-180% -0.051 -0.039 0.078 0.081 -0.0001 0.0010

180-200% -0.075 -0.038 0.080 0.074 0.0005 0.0027

200-220% -0.086 -0.074 0.078 0.073 0.0004 0.0009

220-240% -0.095 -0.072 0.075 0.077 -0.0003 0.0018

240-260% -0.089 -0.088 0.075 0.069 0.0006 0.0000

260-280% -0.095 -0.105 0.070 0.071 -0.0002 -0.0007

280-300% -0.117 -0.119 0.070 0.065 0.0005 -0.0001

Total for Income Variables

0.0022 0.0107

67

Table 8 (continued): Factors that Explain the Food Insecurity Increase

(1) (2) (3) (4) (5) (6)

Coefficients Means Amount

explained Amount

unexplained

Pre-Recession

Great Recession

Pre-Recession

Great Recession

Annual weeks worked 0.000 0.000 23.078 21.563 0.0004 0.0007 Annual weeks looking for work

0.006 0.005 1.310 2.041 0.0043 -0.0023

Average hrs per week worked

0.000 -0.001 19.128 17.840 0.0005 -0.0033

Total for Labor Force Variables 0.0053 -0.0049

African American 0.068 0.061 0.202 0.214 0.0008 -0.0015

Hispanic 0.025 0.072 0.169 0.193 0.0006 0.0090

White -0.005 0.007 0.575 0.535 0.0002 0.0064

Total for Race/Ethnicity Variables 0.0016 0.0140

High School graduate -0.042 -0.040 0.376 0.366 0.0004 0.0010

Some college -0.040 -0.038 0.255 0.274 -0.0008 0.0007

College plus -0.106 -0.095 0.127 0.147 -0.0021 0.0016

Total for Education Variables -0.0025 0.0033

Homeowner -0.101 -0.108 0.565 0.534 0.0031 -0.0038

Foreign born -0.021 -0.031 0.177 0.196 -0.0004 -0.0020

Number of children 0.028 0.025 0.996 0.931 -0.0018 -0.0029

Number over age 60 -0.053 -0.069 0.431 0.442 -0.0006 -0.0074

State housing vacancy rate 0.007 0.004 1.948 2.638 0.0046 -0.0072 State housing price index/100 -0.013 -0.008 3.495 3.453 0.0005 0.0155

Notes: See text for explanation of the Oaxaca decomposition method.

68

Table 9: Food Insecurity Transitions Over the Great Recession

(1) (2) (3) (4)

Never Food Insecure

Always Food Insecure

Became Food Insecure

Left Food Insecurity

Great Recession -0.050*** 0.020*** 0.025*** 0.003

(0.009) (0.005) (0.005) (0.006)

Observations 15,671 15,671 15,671 15,671 Notes: Robust standard errors in parentheses. Sample limited to households with income less than or equal to 300 percent of the poverty line. All models include 20-point income/poverty bins. Marginal effects from probit model shown.

69

Table 10: Nutritional Differences for Food Secure and Insecure, Before and After the Great Recession

(1) (2) (3) (4) (5) (6)

Food Secure Food Insecure Relative Change for FI

P-value for

Relative Change

Before GR Great

Recession Before GR

Great Recession

Number of Meals 3.9 4.3 3.7 4.1 -0.06 0.090

BMI 26.1 26.5 25.7 26.1 -0.04 0.905

Calories 2123 2027 2115 2003 -16.1 0.677

Percent of Meals at Home

71.1 72.0 74.1 76.8 1.90 0.045

Percent of Meals Fast Food

5.9 9.3 6.6 11.0 0.99 0.075

Panel B: Adults Only Protein (gm) 83.4 82.8 81.7 77.8 -3.3 0.113

Potassium (gm) 2753 2675 2577 2405 -93.4 0.129

Calcium (mg) 894 946 837 856 -32.9 0.250

Phosphorus (mg) 1341 1346 1297 1240 -62.4 0.057

Iron (mg) 16.0 15.7 15.3 14.3 -0.76 0.081

Fiber (gm) 15.8 16.3 15.0 14.6 -0.99 0.047

Sources: NHANES 1999 to 2010. Observations : 46,363 overall, and 25,756 for adults.

70

Table 11: Differences in Time Use Among Food Secure and Insecure, before and during the Great Recession

(1) (2) (3) (4) (5) (6)

Personal Care incl

sleep Sleep Work Any Work Education Shopping

Great Recession 7.7*** 7.5*** -7.3*** -0.014** -1.6* -2.7*** (1.6) (1.5) (2.7) (0.006) (0.9) (0.6)

Food insecure 35.5*** 32.5*** -27.1*** -0.064*** 4.3** -5.1*** (3.7) (3.5) (5.1) (0.011) (2.0) (1.1)

Food insecure * GR -7.2 -5.9 -1.8 0.008 0.8 0.8

(5.4) (5.1) (7.5) (0.016) (2.9) (1.5)

Constant 563.7*** 517.7*** 167.6*** 0.399*** 16.6*** 27.0***

(1.0) (0.9) (1.8) (0.004) (0.6) (0.4)

Table 11 (continued): Differences in Time Use Among Food Secure and Insecure, before and during the Great Recession

(8) (9) (10) (11) (12) (13)

HH Activities

Care for HH

members Travel

Eating Drinking

Food Prep

Any Food Prep

Great Recession -5.0*** 0.0 -3.1*** -0.3 0.9 0.030*** (1.6) (0.9) (0.9) (0.6) (0.6) (0.006)

Food insecure -4.9 3.4* -6.2*** -15.5*** 6.7*** 0.024** (3.2) (1.9) (1.8) (1.0) (1.3) (0.011)

Food insecure * GR 2.3 1.0 -3.1 1.6 -1.9 -0.003 (4.8) (2.8) (2.6) (1.6) (1.9) (0.016)

Constant 127.2*** 33.87*** 76.7*** 70.9*** 33.6*** 0.551***

(1.0) (0.6) (0.6) (0.4) (0.4) (0.004)

Notes: Data from the American Time Use Survey 2000-2011 merged to the December Current Population Survey. N=36,544.

71

Table 12: Effect of Leverage on Food Insecurity Over the Great Recession

(1) (2)

Highest leverage quintile state 0.006 0.012

(0.005) (0.009)

Great Recession 0.027*** 0.047***

(0.003) (0.004)

Great Recession * 0.010 0.011

highest leverage quintile state (0.007) (0.011)

Observations 150,736 68,397 Notes: Robust standard errors in parentheses, clustered at the state level. Model (1) includes 33 income/poverty bins – 20-point bins up 400% of the poverty line, 50-point bins up to 1000%, and one dummy for all higher-income households. Model (2) includes 15 income/poverty bins – 20-point bins up to 300% of the poverty line, and then excludes all higher-income households. Marginal effects from probit model shown.

72

Table 13: Effect of Housing Crisis on Food Insecurity Over the Great Recession

(1) (2) (3) (4) (5) (6)

Great Recession 0.028*** 0.025*** 0.015*** 0.049*** 0.044*** 0.031***

(0.003) (0.002) (0.002) (0.005) (0.004) (0.004)

Homeowner -0.083*** -0.128***

(0.005) (0.010) Great Recession* -0.001 -0.003 Homeowner (0.004) (0.007) Biggest Housing Price Drop Quintile State -0.007 -0.011** -0.010 -0.027**

(0.005) (0.004) (0.012) (0.011)

Great Recession* 0.016*** 0.014** 0.020** 0.022* Biggest Housing Price Drop Quintile State

(0.006) (0.006)

(0.009) (0.012)

Homeowners only? YES YES

Observations 150,736 150,736 107,794 68,397 68,397 39,244 Notes: Robust standard errors in parentheses, clustered at the state level. Models (1)-(3) includes 33 income/poverty bins – 20-point bins up to 400% of the poverty line, 50-point bins up to 1000%, and one dummy for all higher-income households. Models (4)-(6) include 15 income/poverty bins – 20-point bins up to 300% of the poverty line, and exclude all higher-income households. Marginal effects from probit models shown.

73

Table 14: Effect of Unemployment Duration on Food Insecurity Over the Great Recession

(1) (2)

Highest Quintile State for 2009/2007 -0.003 -0.002

Average Unemployment Duration (0.005) (0.012)

Great Recession 0.025*** 0.045***

(0.002) (0.004)

Great Recession * 0.016*** 0.019**

highest duration quintile state (0.005) (0.008)

Observations 150,736 68,397 Notes: Robust standard errors in parentheses, clustered at the state level. Models (1)-(3) includes 33 income/poverty bins – 20-point bins up to 400% of the poverty line, 50-point bins up to 1000%, and one dummy for all higher-income households. Models (4)-(6) include 15 income/poverty bins – 20-point bins up to 300% of the poverty line, and exclude all higher-income households. Marginal effects from probit models shown.

74

Table 15: Combined Effect of Leverage, Housing and Unemployment Duration on Food Insecurity Over the Great Recession

(1) (2)

Great Recession 0.023*** 0.043***

(0.002) (0.005)

Highest leverage quintile state 0.011* 0.019*

(0.006) (0.010)

Biggest Housing Price Drop Quintile State -0.010** -0.018**

(0.004) (0.008)

Highest Quintile State for 2009/2007 -0.001 0.001

Average Unemployment Duration (0.004) (0.007)

Great Recession * 0.002 0.001

highest leverage quintile state (0.004) (0.008)

Great Recession* 0.009 0.012

Biggest housing price drop quintile state (0.007) (0.010)

Great Recession * 0.009 0.011

highest duration quintile state (0.006) (0.010)

p-value for test of joint significance 0.0006 0.0218

Observations 150,736 68,397 Notes: Robust standard errors in parentheses, clustered at the state level. Models (1)-(3) includes 33 income/poverty bins – 20-point bins up to 400% of the poverty line, 50-point bins up to 1000%, and one dummy for all higher-income households. Models (4)-(6) include 15 income/poverty bins – 20-point bins up to 300% of the poverty line, and exclude all higher-income households. Marginal effects from probit models shown.

75

Table 16: Simulating Income Increase Necessary to Shift Pre-GR Food Insecurity Rate over to Post-GR Food Insecurity Rate

Weekly 3% Weekly Amount Accounted for

Income/Poverty Bin Increase Needed Increase by 3% Saving Change

40 to 60% $117 $24 20.5%

60 to 80% $94 $33 35.4%

80 to 100% $170 $42 25.0%

100 to 120% $254 $52 20.4%

120 to 140% $301 $61 20.3%

140 to 160% $240 $71 29.4%

160 to 180% $230 $80 34.8%

180 to 200% $219 $90 40.9%

200 to 220% $216 $99 45.8%

220 to 240% $180 $108 60.4%

240 to 260% $89 $118 132.7%

260 to 280% $16 $127 782.0%

280 to 300% $41 $137 331.0%