Population and Environment

, Volume 34, Issue 3, pp 370–399

The impacts of climate variability on household welfare in rural Mexico

Authors

    • The World Bank
  • Katja Vinha
    • Consultant at the World Bank
Original Paper

DOI: 10.1007/s11111-012-0167-3

Cite this article as:
Skoufias, E. & Vinha, K. Popul Environ (2013) 34: 370. doi:10.1007/s11111-012-0167-3

Abstract

In light of the expected increase in weather variability from climate change, we examine the impact of weather shocks, defined as rainfall or growing degree days more than a standard deviation from their respective long-run means, on household consumption per capita. The analyses suggest that both rainfall and temperature shocks affect both food and non-food consumption. Furthermore, the results show that a household’s ability to protect its consumption from weather shocks depends on the climate region and when in the agricultural year the shock occurs. Especially, households in arid climates are not fully protected from weather shocks occurring during the beginning of the wet season (April, May, June). The results highlight the necessity to account for the underlying climatic variation as well as to carefully define the shocks.

Keywords

Climate changeWeather shocksHousehold welfareConsumptionFood securityRural MexicoLivelihoodsEnvironment

JEL Classification

D13I31Q54Q12

Introduction

While there is uncertainty over the exact magnitudes of the global changes in temperature and precipitation, it is widely accepted that significant deviations in the variability of climate from its historical patterns are likely to occur (IPCC 2007).1 Considering that millions of poor households in rural areas all over the world depend on agriculture, there are increasing concerns that the change in the patterns of climatic variability is likely to add to the already high vulnerability of households in rural areas of developing countries, thus posing a serious challenge to development efforts all over the world. In view of this imminent threat of climate change upon the poor, it is critical to have a deeper understanding of the effectiveness of household adaptation strategies and targeted measures that could mitigate the poverty impacts of erratic weather. With these considerations in mind, in this paper, we carry out an analysis of the welfare impact of climatic variability in the rural areas of Mexico. We use the first two waves of the nationally representative Mexican Family Life Survey (MxFLS), carried out in 2002 and 2005/2006, to examine whether the incidence of rainfall and growing degree days (GDD), a cumulative temperature measure, more than one standard deviation from their respective long-run means, have a significant impact on rural households’ ability to smooth consumption.2 The inability to smooth consumption potentially affects the health and wellbeing of rural populations as well as influences households’ migration decisions.

Erratic weather may affect agricultural productivity that, depending on how effective was the portfolio of ex-ante and ex-post risk management strategies employed, may translate into lower income.3 Based on historical experience and the multiplicity of economic and institutional constraints faced, rural households in Mexico, as most rural households all over the world, have managed to develop traditional strategies for managing climatic risk. For instance, households may undertake ex-ante income-smoothing strategies and adopt low-return, low-risk crop and asset portfolios (Rosenzweig and Binswanger 1993). Specific to Mexico, Eakin (2000) documents how smallholder farmers have adapted to climatic risk in the Tlaxcala. For example, farmers plant both fast maturing but low-yield corn as well as slow maturing but high-yield varieties, or they may switch from the more profitable corn to wheat depending on the prevailing weather. They may also alter fertilizer and pesticide use depending on the climatic conditions, and diversify geographically by having plots of land in different locations. Furthermore, households have been shown to use their savings (Paxson 1992), take loans from the formal financial sector to carry them through the difficult times (Udry 1994), sell assets (Deaton 1992), or send their children to work instead of school in order to supplement income (Jacoby and Skoufias 1997). Additional strategies include the management of income risk through ex-post adjustments in labor supply such as multiple job holding and engaging in other informal economic activities (Morduch 1995; Kochar 1999). These actions enable households to spread the effects of income shocks from unanticipated negative events through time. Certain individual characteristics, such as lower educational attainment, may increase the vulnerability of households to risk (Skoufias 2007).

Quantitative evidence on how successful such risk management strategies are at protecting household welfare from weather shocks in Mexico is quite scarce. Other studies relying on the perceptions of respondents about the incidence of different types of shocks, such as floods, droughts, freeze, fires, and hurricanes, include Garcia-Verdu (2002), Skoufias (2007) and De la Fuente (2010). None of these earlier studies, however, make use of actual meteorological data. To the extent that the current risk-coping mechanisms employed are not effective in protecting welfare from erratic weather patterns, one can be quite certain that the change in the patterns of climatic variability associated with climate change is likely to reduce the effectiveness of the current coping mechanisms even further and thus household vulnerability.

To gain a deeper understanding of who are most affected by weather shocks and where such effects are more pronounced, we first quantify the effect of weather shocks on consumption nationally and subsequently for different climatic regions based on average precipitation. By separating the sample along climate criteria, we group together households who face similar challenges from similar shocks. Since food consumption is sometimes better protected than non-food consumption (Skoufias et al. 2011a; Skoufias and Quisumbing 2005), we analyze the impacts of weather on per capita consumption in food and non-food items separately. Furthermore, it is quite possible that the resilience and the ability to adapt to changes in weather and environmental conditions differs significantly depending on access to different risk-coping mechanisms. Therefore, we investigate the extent to which some such mechanisms discussed above—namely, assets, land titling and education, as well as access to transportation infrastructure—alter the ability of households to smooth consumption.

One distinguishing feature of our study is that we investigate the extent to which the timing of the climatic shock within the agricultural cycle matters. We match each household to the weather shocks experienced in the prior agricultural cycle (encompassing a dry season from October to March and a wet season from April to September), in the prior the wet season, and in the first 3 months of the wet season prior to the survey. April, May, and June—the pre-canícula period—are critical months for many corn growers (Eakin 2000). In addition, although measures based on rainfall have widely been used in determining the effect of weather shocks on consumption (for example, Dercon and Krishnan 2000; Jacoby and Skoufias 1998; Paxson 1992; Rosenzweig and Binswanger 1993; Skoufias et al. 2011a), measures based on temperature have not received the same attention. Temperature measures have been used in assessing the economic impacts of climate change through crop yields (Deschênes and Greenstone 2007; Schlenker and Roberts 2008), but have not been included in models of the impact of weather shocks on consumption. To capture this other important aspect of weather, we include weather shocks based on the cumulative temperature during the three time periods considered.

The rest of the paper is organized as follows: The next section gives some background on Mexico’s climate and agriculture. “Data sources” describes the data sources. “Empirical analysis” presents the empirical analysis. “Concluding remarks” concludes.

Mexico’s climate and agriculture

Both rainfall and temperature are important factors affecting crop yields and exhibit a concave relationship with agricultural productivity. Both extremes of rainfall (drought or flood) and temperature (extremely cold or extremely hot) will negatively impact yields and thus, potentially, income and consumption as well. Within a normal range of rainfall and temperature, additional rainfall or warmer days may increase yields in one climate but they may reduce yields in another. Specific to Mexico, Galindo (2009) identifies both Mexican states where higher temperatures lead to higher yields and where they lead to lower yields, suggesting possible heterogeneity from weather shocks in Mexico. For example, corn production is found to benefit from additional temperature in Hidalgo, Estado de México, Puebla and Querétaro and decrease with additional temperature in Baja California de Sur, Campeche, Chiapas and Guerrero (Galindo 2009). Similarly, he finds the optimal levels of rainfall below and above which yields fall depend on the class of crops considered. Alternatively, Conde et al. (1997) find that in the long run, a climatic change with an increase of 2°C and a 20% decrease in rainfall would increase the amount of unsuitable land for corn production by 8% in a sample of seven corn producing municipalities (from the states of Mexico, Puebla, Veracruz and Jalisco). Likewise, a 2°C increase in temperature but a 20% increase in rainfall would increase the amount of land unsuitable for corn production by 18%. Simulating a temperature increase of 4°C over the mean temperature, the amount of land unsuitable for production, with a 20% increase and a 20% decrease in rainfall, increased by 20 and 37%, respectively. Based on actual production, Appendini and Liverman (1994) estimate that in Mexico, droughts are responsible for more than 90% of all crop losses.

The agricultural year in Mexico runs from October to September. It is composed of a dry season, from October to the end of March, and a wet season, from April to the end of September. About 82% of cultivated land is rainfed (INEGI 2007), and thus very susceptible to weather fluctuations. In the wet season, corn is produced in 59% of cultivated land in seasonal crops and in the dry season, 31% of the land in seasonal crops is in corn. The total area cultivated is more than six times greater in the wet season than in the dry season (INEGI 2007). More importantly, corn is used by many small-scale farmers not only as a source of income but also directly as a subsistence crop. Switching to other crops, such as wheat or barley, which have a shorter growth cycle but are not as useful for household consumption, is considered a last resort (Eakin 2000).

The growing cycle for corn can be divided into three phases (Neild and Newman).4 The first phase (vegetative phase) lasts between 60 and 40 days. The longer it takes for the seed to germinate (i.e., the colder it is after planting) the higher the probability that the seed is weak and subject to disease producing a lower yielding crop. For the first half of this time, the growing point is usually below ground and the plant can withstand to some degree cold temperatures. After the growing point is above ground level, then frost can cause significant damage to the plant. With the ear formation begins, the reproductive phase with the ear-forming stage lasting for about 20 days and an additional 20–30 days are required for the grain fill stage. Inadequate water availability during this phase greatly affects yields with the impacts being the greatest during the ear-forming stage. Also, extremely warm temperature (above 32°C) during the second half of the vegetative phase and the reproductive phase reduces yields. The last phase (maturation phase) lasts between 20 and 35 days.

Planting later in the season ensures that the seed germinates quicker; however, waiting too long does not allow the crop to complete the maturation stage before the growing season ends. Furthermore, specific to Mexico, in July and August, there is a period of mid-summer drought called canícula (Fig. 1) affecting farmer’s planting decisions. In general, farmers want the corn to flower (for the ear formation stage to be complete) before the onset of the canícula in order to better the odds of the crop survival in case it is a drier than normal canícula (Eakin 2000). This implies that the months leading up to the canícula are of special importance in Mexico.
https://static-content.springer.com/image/art%3A10.1007%2Fs11111-012-0167-3/MediaObjects/11111_2012_167_Fig1_HTML.gif
Fig. 1

Timing of MxFLS I and the agricultural cycle in Mexico

Data sources

For the household data, we use the first two waves of surveys from the Mexico Family Life Survey (MxFLS; Rubalcava and Teruel 2006). The first wave of the survey interviewed 3,353 rural households in 75 different localities located in all regions of the country and was conducted between March 2002 and August 2002, with the majority of the information collected in April, May, and June (see Fig. 1 for timing).5 The second wave of the survey was collected between 2005 and 2007 with the majority of the data collection occurring from May 2005 to September 2005. The follow-up survey interviewed 3,271 households. Both waves collected detailed information on each household member including information on the basic characteristics, educational attainment, and migration. Furthermore, the survey collected detailed information on household expenditures.6 Separate surveys were administered to the leaders of each locality on infrastructure and programs accessible in the locality.7

The climate data for this paper come from the Mexican Water Technology Institute (Instituto Mexicano de Tecnología del Agua—IMTA). The IMTA has compiled daily weather data from more than 5,000 meteorological stations scattered throughout the country. The data span a long period of time—from as far back as the 1920 s to 2007—and contain information on precipitation, and maximum and minimum temperature. The meteorological stations registered these variables on a daily basis, and we use this information to interpolate daily values of these variables for a geographical centroid in each municipality in Mexico.8 The centroid was determined as the simple average of the latitude and longitude coordinates of all the localities listed in INEGI’s 2005 catalogue corresponding to each municipality, which resulted in a locality-based centroid. We chose this method over a population-weighted average because that alternative would bias the interpolation toward urban rather than rural areas. The interpolation method used is taken from Shepard (1968), a commonly used method that accounts for relative distance and direction between the meteorological stations and the centroids [see Skoufias et al. (2011b) for a more detailed description].

We carry out an independent interpolation for every day between 1950 and 2007, for each municipality. Since not all meteorological stations existed throughout the entire period and they sometimes failed to report their records, each interpolation is based on a different number of data points—and indeed different weather stations. These problems as well as the accuracy of the data get worse as one looks at earlier years, which has a corresponding effect on our interpolations. Thus, interpolations for the year 1950 are less reliable than those for 2007.

From these weather data, we calculate the total rainfall and growing degree days (GDD) for each agricultural year (October to September), for each wet season (April to September), and for each pre-canícula period (April, May, June), or the months leading to the canícula, from 1951 to 2007.9 Instead of maximum of minimum temperatures we use GDD, a cumulative measure of temperature based on the minimum and maximum daily temperatures. GDD measures the contribution of each day to the maturation of the crop. Each crop, depending on the specific seed type and other environmental factors, has its own heat requirements for maturity. For example, some corn varieties require 2,450 GDDs, whereas others require 3,000 GDDs to mature; some wheat varieties only require 1,800 GDDs, whereas others require 2,000 GDDs.10

Each crop has specific base and ceiling temperatures, Tbase and Tceiling, respectively, which contribute to growth. The base bound sets the minimum temperature required for growth, and the ceiling temperature sets the temperature above which the growth rate does not increase any further, and in fact, temperatures above the ceiling may be detrimental to growth. Thus, the contribution of each day, j, to the cumulative GDD is given by
$$ \left( {T_{j,\overline{\min } } + T_{j,\overline{\max } } } \right)/2 - T_{\text{base}} = {\text{GDD}}_{j} $$
(1)
where \( T_{{j,\overline{\min } }} \) and \( T_{{j,\overline{\max } }} \) are the minimum and maximum daily temperature truncated at the base and ceiling values. In other words, any daily temperature (minimum or maximum) below the base temperature is assigned the base temperature value and any daily temperature above the ceiling temperature is assigned the ceiling temperature value.11 To determine the cumulative GDD at any point in time for a specific cultivation, the daily GDDs since planting are summed.

Given the mixture of different crops grown in the survey areas, we use the generalized bounds of 8 and 32°C (for example, Schlenker and Roberts 2008). In our specific case, any daily minimum or maximum temperature below 8°C is treated as being 8°C and any daily minimum or maximum temperature above 32°C is treated as being 32°C. Thus, a day with a minimum and maximum temperature of 8°C or below will yield no GDDs, whereas a day with a maximum and a minimum temperature of 32°C or above will yield 24 GDDs.

For our measures of weather shocks, we first construct the municipal historical mean rainfall and GDD between 1951 and 1985 for the agricultural year, for the wet season and for the pre-canícula period as well as their standard deviations. We choose this date range to balance the need to calculate the historical means with as many years of information as possible but excluding recent years, which may have been affected by changing climate. Furthermore, we use a 35-year span for the baseline since there is incomplete information for some months for some of the municipalities.12 In our sample of rural municipalities, the average climate is based on 15–35 years of information. Seventy five percentage of the rural households in our sample live in localities located in municipalities with at least 30 years of complete weather information from 1951 to 1985.

Our chosen measures of weather shocks, W, are based on the degree of deviation from the 1951 to 1985 average weather. A shock is defined by an indicator variable identifying those observations where the weather variable is more than one standard deviation from its long-run mean. A municipality is defined to have experienced a negative rainfall shock if the prior period’s rainfall was at least one standard deviation less than the average 1951–1985 rainfall, and a municipality is defined to have experienced a positive rainfall shock if the prior period’s rainfall was at least one standard deviation more than the average 1951–1985 rainfall. Thus, there are in total four measures describing the prior year’s (or wet season’s or pre-canícula period’s) weather. We also use two aggregate shock measures, one for rainfall and the other for GDD, such that the indicator is equal to one if the municipality experienced either a positive or negative shock. One standard deviation rainfall shock translates to an average of about 30% higher or lower rainfall. One standard deviation of GDDs is, on average, about 8% from the mean. Table 1 presents the distribution of shocks from 1986 to 2002. During this time period, there were more temperature shocks (both negative and positive) than rainfall shocks, suggesting that temperature was more a more variable aspect of weather than rainfall when compared to pre-1986 weather.13
Table 1

Prevalence of weather shocks in Mexican municipalities between 1986 and 2002 from mean 1951 to 1985 weather

Shock type

SDs from mean

Annual rainfall

Wet season rainfall

Pre-canícula rainfall

Annual GDD

Wet season GDD

Pre-canícula GDD

Freq.

%

Freq.

%

Freq.

%

Freq.

%

Freq.

%

Freq.

%

Negative

−2

3,487

4.27

3,122

3.69

1,476

1.74

6,102

7.47

7,312

8.63

7,598

8.93

−1

13,925

17.05

13,230

15.62

13,142

15.45

10,961

13.42

12,174

14.37

12,654

14.88

 

0

53,475

65.48

58,014

68.48

59,242

69.64

45,515

55.73

48,346

57.07

48,870

57.45

Positive

1

6,827

8.36

6,804

8.03

8,106

9.53

12,045

14.75

11,198

13.22

10,970

12.9

2

3,951

4.84

3,542

4.18

3,102

3.65

7,042

8.62

5,682

6.71

4,976

5.85

Authors’ calculations from ERIC III (IMTA) weather data from 1951 to 1985 to construct the average weather and the shocks are based on annual, wet season, and pre-canícula period weather from 1986 to 2002

The survey date is used to match each household to the weather information. Each household is assigned the wet season and dry season prior to the survey date. That is, if a household was surveyed in dry season of the agricultural year t, the weather shocks would based on the weather in the dry season t − 1 and the wet season t − 1. However, if the household was surveyed in the wet season of year t, the weather shocks are based on weather in dry season t and wet season t − 1. As an illustration, for the households in the 2002 wave of the MxFLS, the weather variables of interest are rainfall and GDD based on April 2001 to March 2002 weather (see Fig. 1). Thus, we are assuming that the households’ income and production would be based on the harvests of the 2001 wet season and the 2002 dry season, and not on the harvest from the 2002 wet season that is roughly contemporaneous to the survey. Given the long time span for data collection in the second wave of MxFLS, not all households are matched to weather shocks from the same two seasons (as is the case in first wave) but households are matched with the previous completed dry and wet seasons prior to being surveyed. The longer survey period implies that there are more than 75 possible distinct weather pairs in the original 2002 MxFLS sample of municipalities.

Table 2 shows the distribution of rainfall and GDD shocks for the rural municipalities in the final sample from MxFLS. Although the number of municipalities from which the household surveys are drawn is relatively small, we do still have some variability in the weather variables. There are municipalities that experienced positive and negative rainfall as well as GDD events. As Table 2 shows, there are more GDD shocks than rainfall shocks in the sample, which is in line with the national trend from pooling all shocks from 1986 to 2002.
Table 2

Weather shocks in MxFLS sample

Shock type

SDs from mean

Rainfall

GDD

Annual

Wet season

Pre-canícula

Annual

Wet season

Pre-canícula

Freq.

%

Freq.

%

Freq.

%

Freq.

%

Freq.

%

Freq.

%

Negative

−2

10

4.63

6

2.67

8

3.54

13

6.02

13

5.78

14

6.19

−1

36

16.67

39

17.33

26

11.5

35

16.2

35

15.56

35

15.49

 

0

142

65.74

147

65.33

167

73.89

126

58.33

135

60.00

142

62.83

Positive

1

20

9.26

22

9.78

23

10.18

33

15.28

35

15.56

29

12.83

2

8

3.70

11

4.89

2

0.88

9

4.17

7

3.11

6

2.65

Authors’ calculations from ERIC III (IMTA) weather data from 1951 to 1985 to construct the average weather and the shocks are based on annual, wet season, and pre-canícula weather during the agricultural year prior to household’s survey. Sample of municipalities is based on the MxFLS rural municipalities

In our sample, on average information from thirteen weather stations are used to calculate the weather shocks and the average weather station is 13 km from the municipal centroid.14 Since the interpolated weather may more accurately reflect actual weather of a municipality the closer the weather stations are, we also use a sample of municipalities with a mean distance to a weather station less than 20 km. Another possible concern is that since the location of weather stations is not random if less productive areas where households may be more vulnerable to weather shocks have worse coverage, then the interpolation algorithm may be introducing systematic measurement error. To test this, we correlate the mean distance of a weather station used to estimate the weather with average rainfall. We find no statistically significant correlation in the mean distance from municipal centroid to the average weather station and average rainfall or average temperature for the MxFLS survey years. The correlations are higher for earlier years.

The original MxFLS municipalities come from 16 different Mexican states and from all the different regions of the country. Although these states vary in the percentage of land cultivated under rainfed technologies, in most, at least 75% of the land is rainfed (Table 3) and thus with production highly susceptible to weather conditions. Also, in most corn is cultivated on at least 50% of the land cultivated with seasonal crops in the wet season. In all states, the cultivated area in the wet season is greater than the area cultivated in the dry season. These figures suggest that we can expect for an average rural household in our sample the income, as well as production for self consumption, to be relatively highly dependent on the weather and especially on the weather during the wet season. Also, given the relative importance of corn, the pre-canícula period is of interest.
Table 3

Agricultural land cultivated during the 2007 agricultural year

Region

State

Hectares cultivated

% Of land rainfed

Wet season

Dry season

% Of land in seasonal crops

% Of land in seasonal crops in corn

% Of land in seasonal crops

% Of land in seasonal crops in corn

National production

29,902,091

82

46

59

7

31

North

Baja California Sur

129,337

27

22

41

13

12

Coahuila

898,673

66

31

27

5

3

Durango

934,823

80

74

42

6

12

Nuevo Leon

594,937

78

35

49

6

12

Sinaloa

1,335,592

54

44

39

31

63

Sonora

1,259,606

41

27

12

22

4

Center

Guanajuato

1,030,730

67

80

56

13

13

Jalisco

1,694,487

89

43

87

3

26

Estado de Mexico

710,422

85

77

81

6

25

Michoacán

1,422,771

78

47

79

6

18

Morelos

150,219

72

55

37

34

3

Puebla

1,011,643

88

70

73

4

41

Mexico City

Distrito Federal

18,814

94

65

42

16

4

South Pacific

Oaxaca

1,653,708

96

37

84

3

64

Gulf and Caribbean

Veracruz

2,644,988

97

20

86

4

83

Yucatan

568,739

92

39

79

1

46

Data compiled from INEGI (2007). Regional assignations are taken from Conroy (2009), p. 39

Empirical analysis

We adapt a commonly used equation (e.g., Cochrane 1991; Mace 1991; Townsend 1994, 1995) to estimate the degree of consumption smoothing. Instead of using income, we use weather shocks as proxies for it. Specifically, we estimate
$$ \Updelta \ln {\text{PCE}}_{h,m,t} = \alpha + \beta \Updelta W_{m,t} + \gamma X_{h,m,t} + \Updelta \varepsilon_{h,m,t} $$
(2)
where \( \Updelta \ln {\text{PCE}}_{h,m,t} \) is first difference in the logarithm of expenditures per capita of household, h, located in municipality m, between years t0 and t1. Xh,m,t is a vector of other factors explaining consumption levels, such as assets, and household characteristics. \( \Updelta \varepsilon_{h,m,t} \) captures changes in time variant unobserved preferences of the household. ΔWm,t is a vector describing the change in the set of weather shocks in municipality m, experienced at time t0 and t1. We employ both aggregate and disaggregate shocks. For the aggregate measure, if the rainfall in period t is not within one standard deviation of the long-run mean, then \( W_{l,t}^{{{\text{rainfall}}, {\text{agg}}}} = 1 \), otherwise it is equal to zero. Similarly, if the GDD in period t is not within one standard deviation of the long-run mean, then \( W_{l,t}^{{{\text{GDD}}, {\text{agg}}}} = 1 \), and otherwise it is equal to zero. For the disaggregated shocks, we differentiate between negative and positive shocks since the effects of weather shocks on income may differ depending on the direction of the shock. Specifically, a locality has a negative or positive weather shock, \( W_{l,t}^{\text{neg}} = 1 \) or \( W_{l,t}^{\text{pos}} = 1 \), respectively, when the weather variable (rainfall or GDD) in period t is at least one standard deviation less than, or more than, the long-run average climate in the locality. There are thus three possible values for ΔWm,t. It will be −1 if in t0 a particular type of shock (for example and aggregate rainfall shock or a positive GDD shock) occurred, but in t1, such a shock did not occur. It will be one if such a shock occurred in t1 but not in t0. And it will be zero if in both time periods the municipality experienced such a shock or if in neither time periods the municipality experienced such a shock. Thus, β measures the impact of weather shocks on consumption. As long as the weather shock is exogenous, that is E(Wεi,l,t) = 0, the coefficient estimate β is unbiased. Given our definition of weather shocks, E(Wεi,l,t) = 0 should hold. The probability of experiencing a weather shock is the same in all the municipalities. In the absence of full insurance against income shocks, any weather shock that reduces income also reduces consumption and we expect β < 0. If the household’s ex-ante and ex-post strategies successfully insure it from income shocks, then β = 0. If the shock benefits agricultural production, we expect consumption to either increase, β > 0, or remain constant β = 0. Paxson (1992) finds that a large portion (potentially all) of transitory income is saved suggesting β = 0.

We use two distinct measures of consumption, non-food, and food since different types of consumption may be affected differently (Skoufias et al. 2011a; Skoufias and Quisumbing 2005). First, we use the logarithm of per capita annual expenditures on all non-health- and non-food-related items. The expenditures are based on the household’s reported spending in the past week on tobacco and public transportation, in the prior month on personal items, cleaning products, general services, recreation, gambling, and communications, in the prior 3 months on clothing, toys and baby items, household items, healthcare, and vehicle maintenance, annual spending on appliances, furniture, house repairs, vehicles, vacation and taxes, and spending in the current school period on education. Following Thomas et al. (2010), we subtract annual health spending from the total expenditures, which on average are about 11% of total expenditures, since most health spending follows illness and thus is not welfare improving. Second, we use the logarithm of per capita annual expenditures on food. The average share of food expenditures in our sample is 41% of total expenditures (without considering health expenditures). Included in the expenditures are the estimated value of goods consumed from own production and the value of goods received as gifts in the week prior to the survey.15 That is, the expenditure measure we use reflects expenditures after including the monetary value of self-production or resources from any coping mechanisms used by households to smooth consumption (such as selling assets, help from friends and relatives, or benefits from government programs). The extent to which these impacts have implications on the future long-run poverty status of the household is not explored in this paper.

Higher observed expenditure may be a consequence of higher local prices faced by households rather than due to a greater quantity of goods consumed. In order to account for covariate price effects, all expenditures are adjusted by monthly price variation at the regional level. Expenditures of households in each of the seven regions for which a poverty line exists are deflated with respect to the May 2001 poverty line in rural areas of Mexico City (DF), \( {\text{PL}}_{{{\text{DF}},5/2001}}^{\text{rur}} \).16 That is, for a household in region i surveyed at time t, prices are multiplied by the ratio \( {\text{PR}}_{i,t} = {\text{PL}}_{i,t}^{\text{rur}} /{\text{PL}}_{{{\text{DF}},5/2001}}^{\text{rur}} \), where \( {\text{PL}}_{i,t}^{\text{rur}} \) is the poverty line for the rural areas of region i at time t, where t is the month of the survey.

Besides the weather shock variables, we include variables that capture household composition (number of children in the household, number of adult males in the household, number of adult females in the household), characteristics of the household head (years of schooling of the household head, gender of the household head, and the age of the household head), asset index,17 and the characteristics of the housing unit (presence of a kitchen, access to tapped water indoors, presence of a toilet, access to piped sewage or septic tank, electricity, and flooring material). The household composition and asset index variables enter as changes between the two surveys. The rest of the independent variables reflect the household’s situation in the second survey period. Furthermore, to account for the potentially different amount of resources available, or any seasonal consumption patterns depending on the season in which the household responded to the expenditure survey, we introduce a season indicator variable.18 Since all households were surveyed during the wet season in 2002, only one seasonal variable (for second survey season) is introduced.

Table 4 gives the descriptive statistics of the variables used in the analyses. To ensure that the weather shocks reflect the experience of the household, only those households where the head did not migrate in the 2 years prior to each of the surveys are included. If those who migrated are more likely to migrate from an area after a weather shock that negatively affected their agricultural production and income, then any estimate is a lower bounder. Furthermore, we exclude from our analyses households that report extremely large (greater than sixteen standard deviations from the sample mean) per capita food expenditures or per capita non-health/food expenditures. This excludes five households from the study.
Table 4

Sample characteristics of rural households

Variable

All households

Low precipitation

High precipitation

Mean

SD

Mean

SD

Mean

SD

Food expenditures per capita, 2002

310

778

340

840

288

730

Food expenditures per capita, 2005/2006

277

612

314

883

250

278

Non-health/food expenditures per capita, 2002

677

1,103

767

1,202

611

1,019

Non-health/food expenditures per capita, 2005/2006

980

18,049

1,624

27,763

510

1,048

Change in the number of children in the household

−0.15

0.86

−0.18

0.90

−0.13

0.84

Change in the number of adult males (over 16) in the household

0.23

0.58

0.26

0.58

0.21

0.58

Change in the number of adult females (over 16) in the household

0.25

0.58

0.27

0.57

0.24

0.58

Household head has not completed primary school, 2005

0.59

 

0.54

 

0.63

 

Gender of household head (1 = female), 2005

0.21

 

0.21

 

0.20

 

Age of household head, 2005

53.30

15.64

52.86

15.21

53.63

15.94

No separate kitchen, 2005

0.07

 

0.05

 

0.08

 

No tap water, 2005/2006

0.16

 

0.10

 

0.19

 

No toilet, 2005/2006

0.36

 

0.37

 

0.36

 

No sewage, 2005/2006

0.38

 

0.35

 

0.41

 

No electricity, 2005/2006

0.02

 

0.02

 

0.03

 

Dirt floor, 2005/2006

0.15

 

0.15

 

0.15

 

Change in total number of assets (Max. is 13)

−0.46

2.04

−0.55

2.05

−0.39

2.04

Surveyed in the wet season in 2005/2006

0.91

 

0.84

 

0.96

 

Number of observations

1,992

840

1,152

Author calculations from MxFLS 1 and 2. Excluded are households with per capita expenditures on food greater than 50,000 pesos and per capita expenditures on non-health and food greater than 150,000 pesos. The condition excludes four households

On average, the households reported slightly lower per capita food consumption in the second round than in the first round. The non-health/food consumption is higher in the second round than in the first, but the average is influenced by a few households with large expenses. In the second round, there are fewer children per household and more adults per household as expected given that the same set of households is interviewed 3 or 4 years later. In 2005, more than half of the household heads had not completed primary school and there are less household heads without primary education in the arid municipalities than in the humid ones. About one-fifth of the households were headed by a female. About one-third of the households did not have access to a sewage system or a toilet in their dwelling unit.

Consumption and weather shocks

To begin with, we estimate Eq. 2 with aggregate measures and pooling all households together regardless of the climatic region in which they live. We use two different samples—those households that did not experience any type of weather shock in 2002, and all household. By limiting our households to those that did not experience a shock in 2002, we simplify the weather shock variables. The shock variable will be zero if a particular shock did not occur in 2005/2006 and one if such a shock did occur. By including also those household that experienced some weather shock in 2002, we increase the number of observations but also assume that if the shock occurred in 2002 and not in 2005/2006 then its effect would be the negative of what would have occurred had the household experienced the shock in 2005/2006 and not in 2002. We then differentiate the shocks by their direction, that is, negative or positive shocks, to determine whether or not the direction of the shock matters. Furthermore, we assign each household to a climate region based on the average annual rainfall to determine how households in different climates are affected by different types of shocks.

The OLS estimates of Eq. 2 for non-health/food consumption of all rural Mexican households are presented in Table 5 and for food consumption in Table 6. The coefficient estimates of β suggest that an average household’s annual consumption is protected against any negative income shocks from unusual weather. That is, there are no statistically significant negative coefficient estimates. If the shocks do have a negative impact on agricultural production (and income), then the results suggest that households are either able to protect themselves ex-ante by changing their agricultural practices in response to the weather shocks, or in the case of reduced agricultural revenue, that households are able ex-post to keep consumption (and welfare) from deteriorating by, for example, drawing down on their assets, or receiving help from formal and informal safety networks, such as relatives or social programs, or accessing credit. When we exclude household that experienced a weather shock in 2002, none of the aggregate shock coefficient estimates are statistically significant. After including those households, we observe 22% higher non-health/food consumption after an annual rainfall shock and 18% higher food consumption after a wet season rainfall shock. The results suggest that the shocks augment income. Such increases are possible if the climatic conditions brought about by the shocks improve the growing conditions for the crops cultivated.
Table 5

Weather shocks and non-health/food expenditures per capita

Shocks

Annual

Wet season

Pre-canícula

Excluding households with a 2002 weather shock

All rural households

Excluding households with a 2002 weather shock

All rural households

Excluding households with a 2002 weather shock

All rural households

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Rainfall shock

0.140

(0.243)

 

0.226**

(0.095)

 

−0.253

(0.213)

 

−0.001

(0.094)

 

0.193

(0.194)

 

0.009

(0.091)

 

Negative rainfall shock

 

0.310

(0.267)

 

0.453***

(0.133)

 

−0.199

(0.327)

 

0.119

(0.150)

 

0.597*

(0.310)

 

0.146

(0.259)

Positive rainfall shock

 

0.096

(0.265)

 

0.094

(0.107)

 

0.239

(0.231)

 

−0.018

(0.105)

 

0.169

(0.206)

 

−0.011

(0.110)

GDD shock

−0.083

(0.241)

 

0.045

(0.142)

 

0.621

(0.433)

 

0.087

(0.131)

 

0.302

(0.350)

 

0.008

(0.138)

 

Negative GDD shock

 

0.129

(0.160)

 

0.191

(0.115)

 

−0.370

(0.255)

 

−0.089

(0.192)

 

−0.267

(0.189)

 

−0.260

(0.202)

Positive GDD shock

 

−0.328

(0.553)

 

0.024

(0.215)

 

1.493***

(0.210)

 

0.199

(0.173)

 

1.135***

(0.266)

 

0.130

(0.174)

Observations

616

1,992

621

2,000

969

2,000

Robust standard errors in parentheses, clustered by locality, and *** p < 0.01, ** p < 0.05, * p < 0.1. Statistically significant coefficient estimates are in bold. Calculated using MxFLS rounds 1 and 2. Other independent variables included are changes in household composition (number of children in the household, number of adult males in the household, number of adult females in the household), characteristics of the household head (sex, age and education), changes in assets (sum of land owned by household, whether or not the household owns their residence, another house, bicycle, motor vehicle, an electric device, a washing machine or a stove, a domestic appliance, machinery or a tractor, bulls or cows, horses or mules, pigs or goats, or poultry), and characteristics of the housing unit (presence of a kitchen, access to tapped water indoors, toilet, access to piped sewage or septic tank, electricity, floor type). The variables not measured as changes from 2001 to 2005/2006 reflect the household in 2005/2006. Excluded are households with per capita expenditures on food greater than 50,000 pesos and per capita expenditures on non-health and food greater than 150,000 pesos. The condition excludes four households

Table 6

Weather shocks and food expenditures per capita

Shocks

Annual

Wet season

Pre-canícula

Excluding households with a 2002 weather shock

All rural households

Excluding households with a 2002 weather shock

All rural households

Excluding households with a 2002 weather shock

All rural households

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Aggregate

Disaggregate

Rainfall shock

−0.001

(0.195)

 

0.105

(0.110)

 

0.029

(0.256)

 

0.177*

(0.098)

 

−0.038

(0.237)

 

−0.050

(0.123)

 

Negative rainfall shock

 

−0.066

(0.159)

 

0.099

(0.161)

 

−0.195

(0.221)

 

0.208

(0.126)

 

−0.183

(0.191)

 

0.032

(0.202)

Positive rainfall shock

 

−0.038

(0.224)

 

0.124

(0.122)

 

0.355

(0.366)

 

0.172

(0.134)

 

−0.027

(0.253)

 

−0.072

(0.138)

GDD shock

0.181

(0.244)

 

0.167

(0.122)

 

0.181

(0.384)

 

−0.056

(0.166)

 

−0.081

(0.281)

 

−0.110

(0.157)

 

Negative GDD shock

 

−0.150

(0.265)

 

−0.145

(0.202)

 

−0.308

(0.593)

 

−0.101

(0.219)

 

0.096

(0.334)

 

0.112

(0.211)

Positive GDD shock

 

0.602**

(0.223)

 

0.358***

(0.112)

 

0.449

(0.495)

 

−0.028

(0.204)

 

−0.341

(0.341)

 

−0.253

(0.191)

Observations

616

1,992

621

2,000

969

2,000

Robust standard errors in parentheses, clustered by locality, and *** p < 0.01, ** p < 0.05, * p < 0.1. Statistically significant coefficient estimates are in bold. Calculated using MxFLS rounds 1 and 2. Other independent variables included are changes in household composition (number of children in the household, number of adult males in the household, number of adult females in the household), characteristics of the household head (sex, age, and education), changes in assets (sum of land owned by household, whether or not the household owns their residence, another house, bicycle, motor vehicle, an electric device, a washing machine or a stove, a domestic appliance, machinery or a tractor, bulls or cows, horses or mules, pigs or goats, or poultry), and characteristics of the housing unit (presence of a kitchen, access to tapped water indoors, toilet, access to piped sewage or septic tank, electricity, floor type). The variables not measured as changes from 2001 to 2005/2006 reflect the household in 2005/2006. Excluded are households with per capita expenditures on food greater than 50,000 pesos and per capita expenditures on non-health and food greater than 150,000 pesos. The condition excludes four households

By expanding the set of shocks analyzed into negative and positive shocks, we observe that the aggregate shocks mask some of the variation in the effects of shocks. Again, the statistically significant coefficient estimates are all positive. In the sample where households with a shock in 2002 are excluded, there are large effects from positive GDD shocks in the wet season and pre-canícula period. These effects, however, disappear once the excluded households are included in the analyses, suggesting that such effects are particular to some subset of households. Once households that experienced a weather shock in 2002 are included, annual negative rainfall shocks and annual positive GDD shocks are associated with 45% greater non-health/food consumption and 36% greater food consumption, respectively. That is, after either a drier than normal or a warmer than normal prior agricultural year households spend more in real terms, suggesting that if the shocks increase productivity then at least some of the transitory income is spent.

In order to check the robustness of our results, we exclude from the sample municipalities with an average distance of the closest 20 weather stations greater than 20 km. The further away the stations are, the greater the potential for measurement error. The statistically significant effects remain statistically significant, and the point estimates are of the same magnitude (Table 10).

The average results above do not, however, capture any variability across different regions. Mexico spans many different climatic regions and certain shocks that increase yields in one climate, may decrease yields in another climate. Using INEGI (2009) climate classifications, we classify each municipality into either low- or high-precipitation municipality. Low-precipitation municipalities are those that are classified as very dry, dry, or semi-dry. High-precipitation municipalities are those that are classified as sub-humid or humid.19 In total, there are 27 low-precipitation municipalities and 48 high-precipitation ones.

Table 7 presents the results for households in low-precipitation and high-precipitation municipalities for non-health/food consumption and Table 8 for food consumption. In contrast with the average results, grouping households by the average precipitation of their municipality suggests that not all household are able to smooth their consumption from weather shocks. Households living in municipalities with a dry climate have lower consumption after three types of weather shocks. Non-food and non-health consumption is lower after a negative GDD shock in the pre-canícula period; food consumption is lower after a negative rainfall shock in the pre-canícula period and after a negative annual GDD shock. Households have higher per capita consumption after a negative annual GDD shock (of non-food/health) and after a positive annual GDD shock (of food). The results from a negative annual GDD shock for the low-precipitation municipalities are contradictory. On the one hand, food consumption decreases suggesting that income decreases and consumption is not fully protected, but on the other hand, non-food/health consumption actually increases suggesting higher income. Together the results suggest that there is a change in the spending composition after a cooler than normal year in the more arid municipalities. In the arid regions, households are not protected from shocks experienced during the pre-canícula period with drier or colder periods affecting annual food and non-food/health consumption.
Table 7

Impact of weather shocks on non-health/food expenditures per capita (ln) by average precipitation level

Shocks

Annual

Wet season

Pre-canícula

All

Low

High

All

Low

High

All

Low

High

Negative rainfall shock

0.453***

(0.133)

0.152

(0.306)

0.571***

(0.127)

0.119

(0.150)

−0.177

(0.283)

0.266**

(0.112)

0.146

(0.259)

−0.067

(0.456)

0.030

(0.281)

Positive rainfall shock

0.094

(0.107)

0.062

(0.175)

0.286***

(0.100)

−0.018

(0.105)

−0.123

(0.156)

0.135

(0.131)

−0.011

(0.110)

0.270

(0.225)

−0.087

(0.111)

Negative GDD shock

0.191

(0.115)

0.580***

(0.145)

−0.098

(0.127)

−0.089

(0.192)

0.218

(0.504)

−0.261*

(0.141)

−0.260

(0.202)

−0.593***

(0.188)

−0.085

(0.204)

Positive GDD shock

0.024

(0.215)

0.006

(0.210)

−0.132

(0.357)

0.199

(0.173)

0.293

(0.195)

0.194

(0.239)

0.130

(0.174)

0.350

(0.253)

0.001

(0.249)

Observations

1,992

840

1,152

2,000

840

1,160

2,000

840

1,160

Robust standard errors in parentheses, clustered by locality, and *** p < 0.01, ** p < 0.05, * p < 0.1. Statistically significant coefficient estimates are in bold. Calculated using MxFLS rounds 1 and 2. Other independent variables included are changes in household composition (number of children in the household, number of adult males in the household, number of adult females in the household), characteristics of the household head (sex, age, and education), changes in assets (sum of land owned by household, whether or not the household owns their residence, another house, bicycle, motor vehicle, an electric device, a washing machine or a stove, a domestic appliance, machinery or a tractor, bulls or cows, horses or mules, pigs or goats, or poultry), and characteristics of the housing unit (presence of a kitchen, access to tapped water indoors, toilet, access to piped sewage or septic tank, electricity, floor type). The variables not measured as changes from 2001 to 2005/2006 reflect the household in 2005/2006. Excluded are households with per capita expenditures on food greater than 50,000 pesos and per capita expenditures on non-health and food greater than 150,000 pesos. The condition excludes four households

Table 8

Impact of weather shocks on food expenditures per capita (ln), by average precipitation level

Shocks

Annual

Wet season

Pre-canícula

All

Low

High

All

Low

High

All

Low

High

Negative rainfall shock

0.099

(0.161)

−0.100

(0.274)

0.191

(0.224)

0.208

(0.126)

−0.080

(0.189)

0.404***

(0.138)

0.032

(0.202)

−1.101**

(0.524)

−0.055

(0.216)

Positive rainfall shock

0.124

(0.122)

0.055

(0.132)

0.278

(0.175)

0.172

(0.134)

0.136

(0.178)

0.256

(0.165)

−0.072

(0.138)

−0.059

(0.208)

−0.136

(0.179)

Negative GDD shock

−0.145

(0.202)

−0.296**

(0.117)

−0.162

(0.296)

−0.101

(0.219)

−0.225

(0.231)

0.019

(0.207)

0.112

(0.211)

0.061

(0.206)

0.204

(0.280)

Positive GDD shock

0.358***

(0.112)

0.473***

(0.148)

0.173

(0.169)

−0.028

(0.204)

0.136

(0.141)

−0.139

(0.315)

−0.253

(0.191)

−0.282

(0.205)

−0.219

(0.284)

Observations

1,992

840

1,152

2,000

840

1,160

2,000

840

1,160

Robust standard errors in parentheses, clustered by locality, and *** p < 0.01, ** p < 0.05, * p < 0.1. Statistically significant coefficient estimates are in bold. Statistically significant coefficient estimates are in bold. Calculated using MxFLS rounds 1 and 2. Other independent variables included are: changes in household composition (number of children in the household, number of adult males in the household, number of adult females in the household), characteristics of the household head (sex, age, and education), changes in assets (sum of land owned by household, whether or not the household owns their residence, another house, bicycle, motor vehicle, an electric device, a washing machine or a stove, a domestic appliance, machinery or a tractor, bulls or cows, horses or mules, pigs or goats, or poultry), and characteristics of the housing unit (presence of a kitchen, access to tapped water indoors, toilet, access to piped sewage or septic tank, electricity, floor type). The variables not measured as changes from 2001 to 2005/6 reflect the household in 2005/2006. Excluded are households with per capita expenditures on food greater than 50,000 pesos and per capita expenditures on non-health and food greater than 150,000 pesos. The condition excludes four households

Households living in sub-humid and humid climates are better able to protect their annual consumption. Only negative wet season GDD shocks are associated with a decrease in non-food/health consumption; however, the effect is no longer statistically significant when we exclude municipalities further than 20 km from the average weather station. In contrast with the results for the low-precipitation municipalities, shocks during the prior pre-canícula period do not have a statistically significant impact on consumption. Both negative and positive annual rainfall shocks lead to higher non-food/health consumption. Also, negative wet season rainfall shocks lead to higher consumption of both food and non-food/health, suggesting that less than average rain is income improving.

Differences in household consumption by observable characteristics

In order to determine if the impact of a weather shock differs for different types of households, we estimate Eq. 2 separately for different sub-populations. Ideally, we would analyze the sub-populations by climatic regions; however, the limited number of distinct municipalities (and sets of weather shocks experienced) do not allow for such detailed analyses. Instead, we use all the rural households in the sample and use real food expenditures as the measure of consumption. These analyses only reveal the average national effect and not any differing effects of shocks in the various climatic regions. However, as was the case above with the average effects for different regions, any negative coefficient estimates at the national level suggest that some portion of the population may not be fully protected. The populations of interest are low-/high-asset households, households with less/more educated heads, households without/with land title, and households living in a locality without/with a bus station. In order to ensure we are capturing effects for a particular sub-population, we only include those households did not change status between the two surveys. Table 9 presents the results.
Table 9

Estimated β for various subpopulations using all rural households, dependant variable food expenditures per capita

 

Annual

Wet season

Pre-canícula

Obs.

annual/wet & pre-canícula

Rainfall

GDD

Rainfall

GDD

Rainfall

GDD

Neg

Pos

Neg

Pos

Neg

Pos

Neg

Pos

Neg

Pos

Neg

Pos

Household with low assets in 2002 (<5)

0.597**

(0.270)

0.133

(0.191)

0.080

(0.472)

0.253

(0.175)

0.570***

(0.208)

0.196

(0.182)

0.414

(0.627)

−0.143

(0.370)

−0.205

(0.256)

0.152

(0.309)

0.599

(0.529)

−0.189

(0.347)

616/622

Household with high assets in 2002 (>5)

0.056

(0.206)

0.125

(0.161)

−0.205

(0.167)

0.309*

(0.167)

−0.111

(0.206)

0.227

(0.202)

−0.301

(0.216)

0.054

(0.127)

0.056

(0.258)

−0.147

(0.163)

−0.203

(0.151)

−0.352

(0.234)

955/957

Household head hasn’t completed primarya

−0.164

(0.307)

0.056

(0.163)

−0.005

(0.226)

0.607***

(0.216)

0.158

(0.245)

0.140

(0.160)

0.020

(0.230)

0.073

(0.108)

−0.125

(0.201)

−0.218

(0.186)

0.413

(0.318)

−0.149

(0.211)

716/720

Household head has completed primarya

0.133

(0.216)

0.123

(0.154)

−0.210

(0.230)

0.191

(0.147)

0.254

(0.172)

0.225

(0.153)

−0.121

(0.300)

−0.137

(0.325)

0.126

(0.314)

−0.046

(0.172)

0.056

(0.285)

−0.352

(0.272)

1093/1096

Household does not hold a land titlea

0.031

(0.194)

0.123

(0.140)

−0.238

(0.231)

0.358***

(0.110)

0.208

(0.137)

0.060

(0.142)

0.101

(0.256)

−0.009

(0.247)

−0.371

(0.224)

−0.078

(0.145)

0.285

(0.243)

−0.172

(0.206)

1,272/1,278

Household holds a land titlea

0.410

(0.255)

0.168

(0.218)

0.026

(0.318)

0.368

(0.581)

0.442*

(0.259)

0.602**

(0.287)

−0.223

(0.413)

0.045

(0.249)

1.163*

(0.656)

0.053

(0.203)

−0.055

(0.351)

−0.292

(0.409)

315/316

No bus station in localitya

0.275

(0.222)

0.662**

(0.247)

−0.275

(0.225)

−0.898

(0.811)

0.153

(0.215)

0.572**

(0.262)

−1.405*

(0.751)

−1.308***

(0.337)

0.241

(0.329)

0.308

(0.446)

−0.209

(0.374)

−1.242***

(0.352)

348/356

Bus station in localitya

−0.092

(0.169)

−0.111

(0.246)

−0.189

(0.283)

0.387

(0.260)

−0.074

(0.188)

0.024

(0.242)

−0.699*

(0.401)

−0.488

(0.288)

−0.043

(0.433)

−0.379*

(0.214)

−0.349

(0.250)

−0.442

(0.370)

807/807

Robust standard errors in parentheses, clustered by locality, and *** p < 0.01, ** p < 0.05, * p < 0.1. Statistically significant coefficient estimates are in bold. Calculated using MxFLS rounds 1 and 2. Other independent variables included are changes in household composition (number of children in the household, number of adult males in the household, number of adult females in the household), characteristics of the household head (sex, age, and education), changes in assets (sum of land owned by household, whether or not the household owns their residence, another house, bicycle, motor vehicle, an electric device, a washing machine or a stove, a domestic appliance, machinery or a tractor, bulls or cows, horses or mules, pigs or goats, or poultry), and characteristics of the housing unit (presence of a kitchen, access to tapped water indoors, toilet, access to piped sewage or septic tank, electricity, floor type). The variables not measured as changes from 2001 to 2005/2006 reflect the household in 2005/2006. Excluded are households with per capita expenditures on food greater than 50,000 pesos and per capita expenditures on non-health and food greater than 150,000 pesos. The condition excludes four households

aOnly those households where there was no change in the interaction variable are included

One ex-post risk management strategy is selling assets to smooth consumption (Deaton 1992). Households with a greater number of assets may be in a better position to do so. Therefore, households are divided into two asset groups, those that in the first round had less than five assets, and those that had six or more assets. The median number of assets in the sample is five. We find that in our sample of rural households asset, scarcity is not associated with inability to smooth consumption following a weather shock. We do not find inability to smooth consumption even with lower cut-off values for the poor asset sub-population. Focusing on specific asset, whether the household owns title to land, again we do not observe those without a title being less able to smooth consumption following a weather shock. Households with less-educated heads may be more prone to the effects from negative income shocks (Skoufias 2007). As with asset poorness, we do not find that on average in rural Mexico, following a weather shock, those with less education are worse at smoothing consumption than those with higher levels of education.

The last risk-sharing mechanisms that we explore is the locality’s accessibility. Greater integration of the locality to regional economy and access to opportunities outside of the community gives households more opportunities for managing risks. To this end, we separate the sample by those households who live in communities without a bus stop and those that live in communities with a bus stop. Communities with a bus stop have at least some public transportation to other localities and most likely also have better infrastructure and are better integrated in general. The results from the analysis show that for our sample of municipalities those households in communities without a bus stop are unable to smooth consumption after any type of a GDD shock during the wet season or after a positive GDD shock in the pre-canícula period. Households in municipalities with a bus stop are unable to smooth their consumption after a negative GDD shock during the wet season or after a positive rainfall shock in the pre-canícula period. However, the results must be interpreted with caution since only 37 municipalities reported information on the presence of a bus stop and did not change their status between the two survey rounds. Furthermore, since the presence of a bus station is not exogenous to the characteristics of the community, the coefficient estimates may be capturing effects of other covariant characteristics.

Concluding remarks

In this paper, we address the question of whether households are able to protect their consumption after weather shocks. Given that climate change is expected to increase weather variability, if effective coping mechanisms to protect consumption are not available to households, then they may experience significant impacts on health and well-being as well as on migration decisions. We examine the impacts of weather shocks, defined as rainfall or growing degree days more than a standard deviation from their respective long-run means, on household expenditures per capita. Our results suggest that households are not always able to protect their consumption from weather shocks such that the ex-ante and ex-post coping mechanisms available do not provide sufficient protection. However, some weather shocks increase consumption, potentially from a transitory increase in income. The effects on consumption vary according to the timing of the shock and the climatic region. Contrary to other research (Skoufias et al. 2011a; Skoufias and Quisumbing 2005), we do not find evidence of food consumption being more protected than non-food consumption.

Although the average rural household in our sample is able to smooth consumption such that no weather shock produces a negative effect on real expenditures, when the households are grouped by the average precipitation of their municipality, we observe some households unable to smooth consumption. Especially, households in arid climates are prone to lower consumption after weather shocks. In arid regions, colder or drier than average weather during the pre-canícula period negatively affect household consumption.

We do not find conclusive evidence on the effects of access to various risk management strategies in aiding an average household in the sample to smooth consumption. However, given the heterogeneity in household responses to different shocks by climate, ideally such analyses should be carried out separately for each climatic region.

Further research using more fine-tuned climate categories, and a greater number of distinct municipality-year pairs would shed light on the robustness of the results. A greater number of municipalities would also lead to better estimates on the effects of various municipal level ex-ante and ex-post risk management strategies.

Footnotes
1

According to the Intergovernmental Panel on Climate Change (IPCC) a narrow definition of climate refers to the statistical description in terms of the mean and variability of quantities such as temperature, precipitation, and wind over a period of time ranging from months to thousands of years. The norm is 30 years as defined by the World Meteorological Organization (WMO). Climate is different from weather that refers to atmospheric conditions in a given place at a specific time. The term “climate change” is used to indicate a significant variation (in a statistical sense) in either the mean state of the climate or in its variability for an extended period of time, usually decades or longer (Wilkinson 2006).

 
2

There are other channels through which weather may affect the well-being of individuals. For example, the changes in climate may increase (or decrease) the prevalence of certain diseases and thus have impacts on health outcomes. In Skoufias and Vinha (2011), we explore the impacts of weather shocks on children’s health as measured by their height-for-age in rural Mexico.

 
3

In general, households are better able to insure their consumption against idiosyncratic shocks, which are shocks that affect only a particular household, such as the death of a household member, than they are able to insure against covariant shocks, shocks that affect a large number of households in the same locality, such as weather-related shocks (Harrower and Hoddinott 2005).

 
4

The description of corn’s growth cycle is adapted from Neild and Newman.

 
5

Rural households are considered to be those that live in localities with less than 2,500 inhabitants.

 
6

MxFLS collects information on the value spent purchasing various categories of goods—food, dining out, healthcare, transportation, personal items, education, recreation, cleaning services, communications, toys/baby articles/childcare, kitchen items and bedding, clothing, tobacco, gambling, appliances and furniture, and other expenses—as well as the value of goods consumed from own production or received as gifts. It is not possible to estimate the value of goods consumed from own production since this value and the value of goods received from others are reported jointly.

 
7

There are several localities in each municipality. In MxFLS 1, only two municipalities had more than one locality sampled.

 
8

We use INEGI’s 2005 geographic definitions, which contains 2,451 municipalities.

 
9

Given that the agricultural year starts runs from October to September, the first agricultural year that we use is 1951, and we only use the last 3 months of the 1950 calendar year.

 
10

For other important crops in Mexico, the required GDDs are 2,400 for beans and 2,200–2,370 for sorghum. The GDD values are taken from The Institute of Agriculture and Natural Resources Cooperative Extension, University of Nebraska-Lincoln. Growing Degree Days & Crop Water Use. http://www.ianr.unl.edu/cropwatch/weather/gdd-et.html, Accessed July 22, 2010.

 
11

We use the Modified Growing Degree Days formula where the minimum and maximum temperatures are adjusted prior to taking the average. See for example Fraisse et al. (ND).

 
12

A particular month is coded as missing if none of the 20 closest weather stations reported data for five or more consecutive days.

 
13

The correlation of the six different weather shock variables for the MxFLS sample of municipalities is given in Table 10. The rainfall deviations from mean for the various periods (annual, wet season, and pre-canícula period) are positively correlated with annual rainfall and the wet season rainfall being very highly correlated. The GDD deviations from mean are all very highly correlated. Given the high correlations among the different time periods, we only include weather variables from one time period in each regression.

 
14

These averages imply that the closest station is less than 13 km from the municipal centroid and since stations closest to the centroid are given more weight in determining the weather the assigned values should be similar to the actual weather experienced at the municipal centroid.

 
15

Given the way in which the expenditure survey was administered, we are unable to separate the value of consumption from own production from the value of goods received as gifts. For about 7% of the rural households, more than 50% of their food comes from non-purchased sources. On average for a rural household, about 7% of all food comes from non-purchased sources.

 
16

The poverty line data were obtained from CONEVAL. Poverty lines, for both the urban and rural portions separately, are available for the following seven regions: Mexico City metropolitan area, Center-North, Center-South, North-Border, Northeast, Northwest, and South.

 
17

The asset index of the sum of whether the household owns land, a residence, another house, bicycle, motor vehicle, an electric device, a washing machine or a stove, a domestic appliance, machinery or a tractor, bulls or cows, horses or mules, pigs or goats, or poultry.

 
18

For example, Paxon (1992) finds seasonal consumption patterns.

 
19

The average minimum and maximum annual precipitations are 200 and 600 mm for the arid regions and 900 and 1,400 mm for the humid region.

 

Acknowledgments

We wish to thank the four anonymous referees, Mariano Rabassa and Abla Safir, for useful comments in an earlier version of this paper, and Hector V. Conroy for interpolating the weather data.

Copyright information

© Springer Science+Business Media, LLC 2012