Seasonal Variation in Food Insecurity Is Associated with Heating and Cooling Costs among Low-Income Elderly Americans 1

Transcription

1 The Journal of Nutrition Community and International Nutrition Seasonal Variation in Food Insecurity Is Associated with Heating and Cooling Costs among Low-Income Elderly Americans 1 Mark Nord* and Linda S. Kantor U.S. Department of Agriculture/Economic Research Service, Washington DC Abstract In this study we examine the association between household food insecurity and seasonally high heating and cooling costs. Logistic regression models, controlling for socioeconomic and demographic characteristics, were estimated using data on household food security and economic and demographic data from the Current Population Survey Food Security Supplements and state-level data on heating and cooling degree days from the National Oceanic and Atmospheric Administration. Low-income households, especially those consisting entirely of elderly persons, experienced substantial seasonal differences in the incidence of very low food security (the more severe range of food insecurity) in areas with high winter heating costs and high summer cooling costs. In high-cooling states, the odds of very low food security for poor, elderly only households were 27% higher in the summer than in the winter. In high-heating states, the pattern was reversed for such households; the odds of very low food security were 43% lower in the summer. In light of recent sharp increases in home heating and cooling costs in many parts of the U.S., it is important to understand the extent to which households make tradeoffs between heating and cooling costs and other basic needs that affect their food security. J. Nutr. 136: , Introduction Food security, access by all people at all times to enough food for an active, healthy life, is one of several necessary conditions for a population to be healthy and well nourished. Yet, in 2004, 13.5 million U.S. households were food insecure at times during the year, meaning that their access to enough food was limited by a lack of money or other resources (1). Of these households, 4.4 million, or 3.9% were food insecure to the extent that food intake was reduced and normal eating patterns were disrupted for 1 or more members, at least some time during the year, because they could not afford enough food. Individuals from food insecure households are at increased risk for poor nutritional status and negative health outcomes. Food insecurity and food insufficiency (a closely related condition) have been shown to be associated with poorer diets in adults (2,3), lower intakes of several nutrients for adults (3), health status of adults with diabetes (4), poor self-rated general health status and lower scores on physical and mental health scales for adults (5), poorer cognitive, academic, and psychosocial development of children (6), several adverse health outcomes for infants and toddlers (7), meeting diagnostic screening criteria for major depression in women (8), and obesity and weight gain among women and (less clearly) among men (9). Some of these risks may be especially high for elderly persons, particularly if they have existing health problems that may make 1 The views expressed in this article are those of the authors, and may not be attributed to the Economic Research Service or the USDA. * To whom correspondence should be addressed. marknord@ers.usda.gov. it difficult to purchase, prepare, and eat nutritious foods. Lee and Frongillo found that elderly persons from food insecure households had lower skinfold thickness and significantly lower intakes of energy and other key nutrients than food secure elderly (10). Food insecurity, by definition, is closely linked to income. Poor households were 3 times as likely to be food insecure than higher income households in 2004 (1). And food insecure households typically spend less money on food than other households. This reflects in part the often difficult tradeoffs poor households must make between spending for food and other goods and services that are essential to health and well-being. Such tradeoffs may be particularly difficult for low-income households facing seasonally high home heating or cooling costs. Previous research on the association between nutritional status and home fuel expenditures consists almost entirely of the work by Bhattacharya et al. (11). Combining monthly data from the Bureau of Labor Statistics Consumer Expenditure Survey ( ) with monthly ground temperature data from the National Oceanic and Atmospheric Administration (NOAA) they found that poor households increased fuel spending and decreased food expenditures during cold months in the northern United States. Bhattacharya et al. also observed reduced levels of energy intake in poor households during winter months (and to a lesser extent in the fall) in northern regions, using data from the National Health and Nutrition Examination Survey. However, they did not find evidence of reduced food expenditures or reduced energy intake during periods of high temperatures in the south /06 $8.00 ª 2006 American Society for Nutrition Manuscript received 28 April Initial review completed 30 June Revision accepted 1 September 2006.

2 This analysis extends and complements that research by examining the relation between seasonal differences in temperature, measured as heating degree days and cooling degree days, and households food insecurity. Food insecurity is hypothesized to be a mediating condition that links constrained household resources with reduced food spending and food intake. As a proximate outcome of constrained household resources, food insecurity may be more consistently related than food expenditures to seasonal temperature differences and may, therefore, also reveal a link to seasonally high home-cooling costs as well as heating costs. Using nationally representative data on food security from Current Population Survey Food Security Supplements (CPS- FSS) 2 from (12 14) and data on heating and cooling degree days from the National Oceanic and Atmospheric Administration (NOAA) (15), we examined the extent to which greater proportions of poor households, especially poor elderly households, experienced very low food security (the more severe range of food insecurity) during times of the year when home heating and cooling costs were high, controlling for important covariates. The analysis takes advantage of the fact that during this time period, the CPS-FSS data collection alternated between winter and early spring (April) when costs for home heating during the previous 30 d are high in northern regions and low in the South, and summer (August/September) when the opposite condition prevails for cooling costs. Data and Methods Data. Data on households food security, household composition, income, employment, and other characteristics were from the Current Population Survey Food Security Supplements (CPS-FSS) of April 1995, September 1996, April 1997, August 1998, April 1999, September 2000, and April The CPS-FSS is conducted once each year as a supplement to the Census Bureau s monthly Current Population Survey (CPS). The basic CPS collects information on household income and composition and on labor force and employment status of each member of the household $15 y of age from ;50,000 households representative of the U.S. civilian noninstitutional population. The CPS-FSS, sponsored annually by USDA, obtains information on food security, food spending, and food program participation, providing the national data on which USDA s annual reports on household food security in the U.S. are based (1,16 21). The food security status of each household is assessed by interviewing 1 member of the household using a standardized survey instrument. Households are classified as food secure or food insecure based on their responses to a series of questions about food-related behaviors, experiences, and conditions that are known to characterize households having difficulty meeting their food needs. Food-insecure households are further classified as having low food security or very low food security. The questions cover a wide range of severity of food deprivation from worrying about running out of food to not eating for a whole day. Each question specifies a lack of money or other resources to obtain food as the reason for the condition or behavior, so the measure is not affected by behaviors such as voluntary dieting or fasting. Household expenditures for heating and cooling were estimated using state-level monthly data on heating degree days and cooling degree days from the National Climatic Data Center of the National Oceanic and Atmospheric Administration. Data were mean heating degree days in February and mean cooling degree days in July from 1970 to Analysis sample. The analysis sample consisted of CPS-FSS households during 7 y, , with valid 30-d food security data and valid 2 Abbreviations used: CPS, Current Population Survey, CPS-FSS, Current Population Survey Food Security Supplement; OR, odds ratio Nord and Kantor income data. In some years, 1 or more of the 8 CPS rotation groups lacked 30-d food security data because households in these rotation groups were asked experimental food security questions in place of the standard questions. These households were excluded from the sample and appropriate weighting adjustments were made so that their exclusion did not compromise the sample s representative character. Our sample was further restricted to households with reported incomes below the federal poverty line and with no school-age children (i.e., none.4 y of age), for a total of 20,058 households. The main outcome of interest, reduced food security due to seasonally high home heating or cooling costs, was expected to affect primarily low-income households. Related research suggests that seasonal patterns of food insecurity in households with school-age children are quite distinct from those with no school-age children and may be affected by distinct factors, such as summer child-care costs, receipt of free or reduced-price meals through the National School Lunch Program and meals received through the Summer Food Service Program (22). These factors would complicate and possibly bias the analysis. Two sample subsets were also analyzed separately: households consisting entirely of elderly persons (age $65, n ¼ 5768), and households with no elderly members (n ¼ 12,775). These analyses explored whether the heat or eat phenomenon was predominantly an experience of the elderly. Households composed of mixed elderly and nonelderly were also analyzed. Results (not reported) for those households were intermediate between those of households with no elderly and those with only elderly. The elderly were of particular interest because many of them have fixed incomes and might be particularly vulnerable to seasonal differences in expenses for home heating and cooling. Furthermore, anecdotal evidence suggests that tradeoffs in spending for various basic needs are especially problematic for poor elderly households. Very low food security: main outcome. The dependent variable for all the analyses was very low food security during the 30 d prior to the CPS-FSS. Very low food security is a severe range of food insecurity that the USDA described as food insecurity with hunger in reports prior to It refers to households that have reported multiple indications of reduced food intake and disrupted eating patterns due to inadequate resources for food. USDA made this change in response to a recommendation by the Committee on National Statistics (23). Household food security was assessed for the period 30 d prior to each survey using methods described by Nord (24). The 30-d scale is based on the same concepts and statistical methods as the standard 12- mo U.S. Food Security Scale. The 30-d referenced scale was essential for this study because it measures conditions in the household during specific periods of the year when home heating or cooling costs were high. To minimize the measurement effects associated with the presence of infants and young children (0 4 y) in some households, we assessed household food security using only the 7 adult-referenced items in the standard 30-d scale (25,26). That is, we used the same scale for all households that would normally be applied to households without children. Predictor variables. The season in which the CPS-FSS households were surveyed (August/September or April) was represented by a single dichotomous variable, summer, with a value of 1 for surveys in August/ September (summer) and 0 for April (winter) surveys. The difference between summer cooling and winter heating costs in each state was represented by the variable cool 2 heat, the mean difference between cooling degree days in July and heating degree days in February during Cooling and heating degree days were assessed for periods of 1 to 2 mo prior to the food security surveys so that the billing and payment period for the associated cooling and heating expenses would coincide with the 30-d period prior to the survey, i.e., the period for which food security was assessed. Cool 2 heat was normalized (to mean 0 and SD 1) across states. It ranged from ;22 in Alaska to 12 in Florida. Weighted by households rather than by states, the mean was for households in the primary analysis sample, reflecting the preponderance of residents in the warmer regions of the U.S. An interaction term, summer 3 cool 2 heat, was created to estimate the extent to which seasonal differences in food security vary between high-heating and high-cooling states.

3 Control variables. Particular attention was given to controlling income and employment that may vary seasonally or from year to year and could, therefore, confound the seasonal association between very low food security and home heating and cooling costs. Annual household income was entered as a ratio of household income to the federal poverty line for the household (income/poverty). Two additional variables based on this ratio were also included to control for income effects on food insecurity; the square of income/poverty and a dichotomous variable identifying households with incomes,50% of the federal poverty line. In combination, these variables assess associations with income as a nonlinear (quadratic) function, but adjust for deviations from that overall relation in the lowest income ranges. Labor-force participation and employment of all adult members in the household were described by 13 dichotomous variables that summarize the standard labor force classifications of all members. Control variables were also included in the analysis for socioeconomic and demographic factors that are known to be associated with food insecurity, including gender of household head, race and ethnicity (non-hispanic black, non-hispanic white, or Hispanic) and citizenship status of household reference person, educational attainment of the most highly educated adult, home ownership, and residential mobility (indicating that the household had moved to its current address since the beginning of its participation in the CPS). Statistical analysis. A series of multivariate logistic regression models were estimated to assess the association between very low food security and seasonally high home heating and cooling costs. Descriptive data were generated in SAS Proc Means. The logistic regression analyses were implemented using SAS Proc Logistic. For the entire low-income sample and each of the 2 subset samples, logistic regression models were first estimated with the single independent variable, summer, to assess the extent to which very low food security differed between seasons. Then a second model was estimated for each sample with the additional independent variables, cool 2 heat and summer 3 cool 2 heat. The size of the regression coefficient on the interaction variable, summer 3 cool 2 heat, indicated the extent to which the seasonal difference in very low food security was associated with household residence in states with higher summer cooling costs relative to winter heating costs. The final models included socioeconomic and demographic control variables to confirm that the associations of home heating and cooling costs with seasonal differences in very low food security did not result from alternative causal factors. Results Sample characteristics. Descriptive statistics for the entire sample and the 2 subsamples are shown in Table 1. For all households with incomes below the poverty line and no schoolage children (n ¼ 20,058) the prevalence of very low food security during the 30 d prior to the survey was 8.6%. Very low food security was less prevalent among households consisting entirely of elderly persons (3.7%) than among households with no elderly (11.0%). Season of data collection. For all U.S. households as well as for our study sample, the prevalence of very low food security during the 30 d preceding each survey followed the same seasonal pattern observed in the 12-mo food security measure reported annually by USDA (19,20,26). Over the 7-y period, , prevalence of very low food security among all poor households with no school-age children (n ¼ 20,058), not accounting for differences in heating and cooling costs, was 0.5 percentage points higher in the summer than in the winter (Fig. 1). The size of the seasonal difference is masked somewhat by an upward trend from 1999 to Over the period 1995 to 1999, the seasonal difference between the 2 surveys was 0.9 percentage points. TABLE 1 Descriptive statistics for households with no school-age children and incomes less than the poverty line 1 Characteristic All Elderly only No elderly n 20,058 5,768 12,775 Summer, 2 % Cool 2 heat, 3 mean Income/poverty, 4 mean Very low food security, % Household composition, % Child 0 4 y 16.2 na 23.8 Elderly, $65y na Female head Reference person, % Alien Black Hispanic Educational attainment (reference person), %,High school High school Some college Bachelor degree Advanced degree Residence, % Home owner Mover Income, % Income,50% poverty Employment (primary earner), 6 % Full-time Retired, NILF Part-time, economic Part-time, noneconomic Unemployed Disabled, NILF Other, NILF Employment (secondary earner), 11 % Full-time Retired, NILF Part-time, non economic 1.6, Part-time, economic Unemployed Disabled, NILF Other, NILF Values are means or %. Descriptive statistics were calculated using household supplement weights adjusted so that the weighted number of households was equal to the unweighted number of cases in each year. 2 CPS-FSS data collection was in August or September rather than in April. 3 Difference between cooling degree days in July and heating degree days in February, adjusted to mean 0 and SD 1 across states. 4 Annual household income as a ratio to the household s poverty line. 5 Dichotomous variable indicating that household had moved to its current address since the beginning of its participation in the CPS. 6 Dichotomous variables indicating whether the primary wage earner in the household was in the respective labor force category. 7 NILF, not in labor force. 8 Primary wage earner wanting to work full time but only finding part-time work. 9 Primary earner working part-time by choice, i.e., earner not seeking full-time employment. 10 Primary wage earner not in the labor force for reasons other than disability or retirement. 11 Dichotomous variables indicating whether any adult other than the primary earner was in the respective labor force category. Seasonal variation in food insecurity 2941

4 Figure 1 Prevalence of very low food security during previous 12 mo and previous 30 d, Data are for sample households with incomes below the poverty line and no school-age children (n ¼ 20,048) and for all U.S. households based on Current Population Survey Food Security Supplement data from April 1995, September 1996, April 1997, August 1998, April 1999, September 2000, and April 2001 (26,19,20). TABLE 2 Odds ratios (OR) of 30-d very low food security associated with season of data collection and state-level variation in home cooling and heating costs for poor households with no school-age children 1 Variable All Elderly only No elderly Model 1 Summer 1.11 (1.00, 1.22)* 1.08 (0.82, 1.43) 1.08 (0.97, 1.20) Model 2 Summer 1.04 (0.94, 1.16) 0.85 (0.61, 1.19) 1.05 (0.93, 1.17) Cool 2 heat 0.97 (0.90, 1.04) 1.11 (0.91, 1.35) 0.95 (0.85, 1.03) Summer 3 cool 2 heat 1.15* (1.04, 1.28) 1.50* (1.11, 2.01) 1.09 (0.98, 1.22) 1 Values are OR and 95% CI not adjusted for socioeconomic and demographic control variables. Models were estimated using household supplement weights adjusted so that the weighted number of households is equal to the unweighted number of cases in each year. *P, Nord and Kantor Figure 2 Prevalence of very low food security during previous 30 d in sample households with incomes below the poverty line and no school-age child (n ¼ 20,048), based on Current Population Survey Food Security Supplement data from April 1995, September 1996, April 1997, August 1998, April 1999, September 2000, and April 2001 (26,19,20). States were categorized as highheating ( ), moderate-climate ( ), or high-cooling ( ) based on the difference between the number of cooling degree days in July and heating degree days in February (average ), using data from the U.S. Department of Commerce, National Oceanic and Atmospheric Administration (15). Home heating and cooling costs. A large share of the seasonal difference in very low food security in poor households with no school-age children (n ¼ 20,058) was associated with seasonal variations in home heating and cooling costs. The addition of variables cool 2 heat and summer 3 cool 2 heat to the model reduced the coefficient on summer from 1.11 to 1.04 and rendered it insignificant, indicating that the seasonal variation in very low food security was primarily associated with seasonal variations in home heating and cooling costs rather than with other factors differing between seasons of data collection (Table 2). Further evidence of the association between very low food security and seasonally high heating and cooling costs is provided in the positive and significant odds ratio (OR) on the interaction term summer 3 cool 2 heat (1.15, 95% CI 1.04, 1.28) which shows that the seasonal pattern of lower food security during the summer was noticeably stronger in households in high-cooling states and weaker (or reversed) in highheating states. The difference in seasonal patterns of very low food security between high-cooling and high-heating states was consistent across the 7 y studied (Fig. 2). For this part of the analysis only, states were grouped into 3 categories: high-heating, moderateclimate, and high-cooling, based on the cool 2 heat variable, so that approximately one-third of the sampled households were in each category. This analysis was conducted to verify that the association observed in the regression models was consistent across the study period and did not result from 1 or 2 idiosyncratic years. In all years in which data were collected during April, the prevalence of very low food security was higher in high-heating states than in high-cooling states (although the differences were not statistically significant in all years. The opposite was true in years when data were collected in August or September. Elderly only households. The association of home heating and cooling costs with seasonal differences in very low food security was substantially more prominent among elderly only households than in households with no elderly members (Table 2). The OR on the interaction term, summer 3 cool 2 heat was 1.50, 95% CI 1.11, 2.01 for elderly only households and 1.09, 95% CI 0.98, 1.22 for households with no elderly members. To demonstrate the size of the interaction for poor, elderly only households, we calculated the summer-to-winter OR of very low food security in high-cooling states (11 SD) and high-heating states (21 SD). Because the mean (across states) of cool 2 heat is 0 and SD is 1, the summer-to-winter OR of very low food security in a high-cooling state is the product of the 2 OR: summer and summer 3 cool 2 heat, or That is, the odds of very low food security were 27% higher in the summer than in the winter in a high-cooling state. In a high-heating state, the odds of very low food security were 43% lower in the summer than in the winter (the summer-to-winter OR for a high-heating state is the product of the OR for summer and the reciprocal of the OR for the interaction, summer 3 cool 2 heat). Control factors. The addition of control variables for socioeconomic and demographic factors did not reduce the strength of the association of seasonal differences in very low food security with seasonal variations in home heating and cooling costs (Table 3). In fact, for elderly only households, the coefficient on the interaction term, summer 3 cool 2 heat, was somewhat

5 TABLE 3 larger with the controls added to the model (adjusted OR 1.58, 95% CI 1.16, 2.15) than without them. The association of interest appears, therefore, to represent a causal effect of home heating and cooling costs and not to be a spurious artifact caused by other seasonally variable economic factors. If anything, the effects of seasonally high home heating and cooling costs on food insecurity may be somewhat ameliorated by seasonal differences in economic factors. Cooling effects. Assessing the effects of seasonally high home cooling costs separately from heating costs is complicated by the strong inverse relation between the 2 variables. For the analyses described to this point, we combined information on the 2 characteristics in a single variable, cool 2 heat, to avoid problems of colinearity of both the main effects and the interaction terms. To assess the independent effects of seasonally high home cooling and heating costs, we replaced the cool 2 heat variable with 2 normalized variables; cool (the number of cooling degree days in July and heat (the number of heating degree days in February). We also replaced the interaction variable, summer 3 cool 2 heat with separate interaction terms, summer 3 cool and summer 3 heat. For all poor households without school-age children, neither of the coefficients on the interaction terms was significant, but they were jointly significant (analysis not shown) and of about the same magnitude. For poor, elderly only households, only the summer 3 cool interaction was significant (OR 1.75, 95% CI 1.11, 2.77) whereas the summer 3 heat interaction was near 0 and insignificant (OR 1.02, 95% CI 0.65, 1.58). These findings suggest that the cool or eat phenomenon is at least as strong as the heat or eat tradeoff found by Bhattacharya et al. (11). Discussion Adjusted OR for 30-d very low food security associated with season of data collection and statelevel variation in home heating and cooling costs for poor households with no school-aged children 1 Variable 2 Elderly only No elderly OR (CI) Summer 0.85 (0.66, 1.21) 1.09 (0.97, 1.23) Cool 2 heat 0.96 (0.78, 1.18) 1.01 (0.94, 1.10) Summer 3 cool heat 1.58* (1.16, 2.15) 1.09 (0.97, 1.22) 1 Models were estimated using household supplement weights adjusted so that the weighted number of households was equal to the unweighted number of cases in each year. *P, OR and 95% CI adjusted for household composition, gender of household head, race and citizenship status of household reference person, educational attainment of most highly educated adult, employment status of adult household members, household income relative to poverty line, home ownership, and residential mobility. Our analysis shows that in high-heating states, households with incomes below the poverty line were substantially more vulnerable to very low food security during the winter than during the summer, whereas the opposite was true in high-cooling states. These findings were especially prominent for poor elderly households and remained when controls were added for employment, income, and other household-level factors that could vary from season to season or year to year. For poor households in which all members were $65 y of age, the odds of very low food security in high-heating states were 43% lower in the summer than in the winter; in high-cooling states, the odds were 27% higher in the summer than in the winter. The observed pattern for households with no elderly members was similar, although smaller in magnitude and not statistically significant. This research builds on the earlier work by Bhattacharya et al. (11) by examining the seasonal effects of home heating and cooling costs on households economic access to food. These effects are presumed to underlie the changes in food spending and energy intake observed by Bhattacharya et al. Our findings support the heat or eat phenomenon identified by Bhattacharya et al. that low-income households reduce food spending and caloric intake during cold periods in northern states. Our findings also suggest that the cool or eat effect, i.e., the effect of high home cooling costs on food insecurity, is nearly as strong as the heat or eat effect. Bhattacharya et al. did not find strong evidence for the cool or eat effect on food spending and caloric intake, possibly because those outcomes are more distal and more difficult to measure than food insecurity. The associations between food insecurity, season of data collection, and state-level heating and cooling costs provide evidence that, for many poor households, the tradeoffs between food spending and seasonally high heating and cooling costs are not made easily, that is, without human cost or within a zone of comfort. The difficulty of these tradeoffs may be exacerbated if home energy costs become unusually high due to supply disruptions or unusually high demand. Our findings also suggest that public assistance programs that support spending for home energy needs may provide a measure of protection against severe levels of food insecurity. Future research might usefully examine whether factors such as home ownership, energy assistance, and participation in food assistance programs moderate seasonal effects of home heating and cooling costs. Literature Cited 1. Nord M, Andrews M, Carlson S. Household food security in the United States, Economic Research Service, Economic Research Report No. (ERR-11), Washington (DC): U.S. Department of Agriculture; Tarasuk V. Household food insecurity with hunger is associated with women s food intakes, health, and household circumstances. J Nutr. 2001;131: Dixon L, Winkleby M, Radimer K. Dietary intakes and serum nutrients differ between adults from food-insufficient and food-sufficient families: Third National Health and Nutrition Examination Survey, J Nutr. 2001;131: Nelson K, Cunningham W, Andersen R, Harrison G, Gelberg L. Is food insufficiency associated with health status and health care utilization among adults with diabetes? J Gen Intern Med. 2001;16: Stuff J, Casey P, Szeto K, Gossett J, Robbins J, Simpson P, Connell C, Bogle M. Household food insecurity is associated with adult health status. J Nutr. 2004;134: Alaimo K, Olsen C, Frongillo E. Food insufficiency and American school-aged children s cognitive, academic, and psychosocial development. Pediatrics. 2001;108: Cook J, Frank D, Berkowitz C, Black M, Casey P, Cutts D, Meyers A, Zaldivar N, Skalicky A, et al. Food insecurity is associated with adverse health outcomes among human infants and toddlers. J Nutr. 2004; 134: Heflin C, Siefert K, Williams D. Food insufficiency and women s mental health: findings from a 3-year panel of welfare recipients. Soc Sci Med. 2005;61: Wilde P, Peterman J. Individual weight change is associated with household food security status. J Nutr. 2006;136: Lee JS, Frongillo EA, Jr. Nutrition and health consequences are associated with food insecurity among U.S. elderly persons. J Nutr. 2001; 131: Bhattacharya J, DeLaire T, Haider S, Currie J. Heat or eat? Cold weather shocks and nutrition in poor American families. Am J Public Health. 2003;93: Seasonal variation in food insecurity 2943

6 12. Bickel GS, Nord M, Price C, Hamilton WL, Cook JT. Guide to measuring household food security, revised Reports of the Federal Interagency Food Security Measurement Project No. (6), Alexandria (VA): U.S. Department of Agriculture, Food and Nutrition Service; Hamilton WL, Cook JT, Thompson WW, Buron LF, Frongillo EA, Jr., Olson CM, Wehler CA. Household food security in the United States in 1995: Summary report of the food security measurement project, Alexandria (VA): U.S. Department of Agriculture, Food and Nutrition Service; Hamilton WL, Cook JT, Thompson WW, Buron LF, Frongillo EA, Jr., Olson CM, Wehler CA. Household food security in the United States in 1995: Technical report, Alexandria (VA): U.S. Department of Agriculture, Food and Nutrition Service; National Oceanic and Atmospheric Administration. State monthly temperature, precipitation, and degree days (and previous normals periods). National Environmental Satellite, Data, and Information Service Historical Climatography Series No. (4 and 5), Asheville (NC): U.S. Department of Commerce, National Oceanic and Atmospheric Administration, National Climatic Data Center; Bickel GS, Carlson S, Nord M. Household food security in the United States, : Advance report, Washington (DC): U.S. Department of Agriculture, Food and Nutrition Service; Andrews M, Nord M, Bickel G, Carlson S. Household food security in the United States, Economic Research Service, Food and Nutrition Research Report No. (FANRR8), Washington (DC): U.S. Department of Agriculture; Nord M, Kabbani N, Tiehen L, Andrews M, Bickel G, Carlson S. Household food security in the United States, Economic Research Service, Food Assistance and Nutrition Research Report No. (FANRR21), Washington (DC): U.S. Department of Agriculture; Nord M, Andrews M, Carlson S. Household food security in the United States, Economic Research Service, Food Assistance and Nutrition Research Report No. (FANRR29), Washington (DC): U.S. Department of Agriculture; Nord M, Andrews M, Carlson S. Household food security in the United States, Economic Research Service, Food Assistance and Nutrition Research Report No. (FANRR35), Washington (DC): U.S. Department of Agriculture; Nord M, Andrews M, Carlson S. Household food security in the United States, Economic Research Service, Food Assistance and Nutrition Research Report No. (FANRR42), Washington (DC): U.S. Department of Agriculture; Nord M, Romig K. Hunger in the summer: seasonal food insecurity and the National School Lunch and Summer Food Service Programs. Journal of Children and Poverty. 2006;12: Wunderlich G, Norwood J, editors. Food insecurity and hunger in the United States: An assessment of the measure. National Research Council, Committee on National Statistics, Washington (DC): The National Academy Press; Nord M. A 30-day food security scale for Current Population Survey Food Security Supplement data. Economic Research Service, Electronic Publication from the Food and Nutrition Research Program No. (E-FAN02 015), Washington (DC): U.S. Department of Agriculture; [cited 2005 July 1]. Available from: publications/efan02015fm.pdf. 25. Nord M, Bickel G. Measuring children s food security in U.S. households, Economic Research Service, Food Assistance and Nutrition Research Report No. (25), Washington (DC): U.S. Department of Agriculture; Cohen B, Nord M, Lerner R, Parry J, Yang K. Household food security in the United States, 1998 and 1999: Technical report. Economic Research Service, Electronic Publication from the Food and Nutrition Research Program No. (E-FAN02 010), Washington (DC): U.S. Department of Agriculture; [cited 2005 July 1]. Available from: Nord and Kantor