Temperature Variability and Trust in Vietnamese Rural Households

1 Introduction

This paper analyses the relationship between temperature variations and trust in rural households in Viet Nam. Following Alesina and La Ferrara (2002), trust generally refers to the firm belief in the reliability and honesty of other people. Trust and, more generally, social cohesion are important concepts in economics. The existing literature has argued that social cohesion has a positive impact on several economic outcomes including economic growth and development (Algan and Cahuc 2010), financial development (Guiso et al. 2004), political accountability (Nannicini et al. 2013) and international investments (Massa et al. 2015). A recent study of Viet Nam argues that trust positively impacts on farming investments (Dang et al. 2020). When facing higher barriers from land administration, Vietnamese households with higher levels of trust tend to invest more in agricultural land than low-trust households.

As a result, the literature investigating the factors affecting social cohesion has flourished (Alesina and La Ferrara 2002; Becchetti et al. 2014; Conzo and Salustri 2019; Glaeser et al. 2002). Among these studies, there is a growing body of empirical work linking social cohesion with the environment (Hsiang et al. 2013; Kuper and Kropelin 2006). The idea is that, in harsher environmental conditions, competition over scarce resources may either increase cooperation among people or lead to conflict. Previous studies, however, have mainly focused on the impact of natural disasters on trust (Calo-Blanco et al. 2017; Cassar et al. 2017; Dussaillant and Guzmán 2014; Fleming et al. 2014). There is, however, a lack of evidence on the effect of increased temperatures on trust. Rising temperatures, however, have a severe impact on agriculture and more generally on economic growth, reducing economic opportunities (Dell et al. 2012).

The aim of this paper, therefore, is to understand if temperature variations impact on trust using Viet Nam as a testbed. The main contribution of the paper is to use a measure that is specific for the economy and the social set-up of the country analyzed. In Viet Nam rice is the staple food, accounting for about half of the total intake of calories per person. Needless to say, this crop is also a key element for the Vietnamese economy. The country is the world’s fifth-largest rice-producing country and the fourth-largest rice exporter.^[1] Therefore, following the literature on rice production and climate variability (Baronchelli and Ricciuti 2022; Caruso et al. 2016), I argue that the relevant climate measure to understand the impact of increasing temperatures on trust in Viet Nam is the minimum temperature during the core month of the rice growing season.^[2] In this period rice is more vulnerable to weather shocks which, at this stage of the growing process, impact harshly on the future harvest.

To test this hypothesis, I use the VARHS dataset which collects information on households in different rural communes in Viet Nam over four rounds (2008, 2010, 2012, 2014). I then regress deviations of the minimum temperature in June on two alternative measures of trust using a linear probability model. The results show that there is a positive association between increases in the minimum temperature in June and trust among rural households in Viet Nam. This finding is robust to several robustness checks.

The paper is organized as follows. Section 2 reviews the literature. Section 3 describes the data and the method and Section 4 sets out the results. Section 5 concludes.

2 Literature

This paper is grounded on three strands of the literature. First of all, I rely on a growing body of studies which investigate factors that can potentially affect trust (Alesina and La Ferrara 2002; Becchetti et al. 2014; Conzo and Salustri 2019; Glaeser et al. 2002). Among these factors, social demographic characteristics such as age, income, education and ethnicity are key determinants of trust and, more generally, of social cohesion.^[3] (Ananyev and Guriev 2019; Bjørnskov 2007; Gereke et al. 2018). People with higher incomes and levels of education are more likely to have higher levels of trust. Similarly, older people are more trusting than the young. Conversely, being part of discriminated group such as immigrants or women can reduce trust. Finally, culture, traditions and religion are also important drivers of trust and, more generally, of social cohesion since some cultures are associated with a more trusting vision of the world (La Porta et al. 1997).

Empirical works have extensively investigated the determinants of trust. Ananyev and Guriev (2019) show that the dramatic fall in Russian GDP caused by the 2008 economic crisis caused a dramatic reduction in trust. This reduction was uneven across Russia, with less developed regions undergoing a more severe fall in trust. Gereke et al. (2018) used data from a survey conducted in Germany to point out that income inequality and not ethnic diversity determines trust. The authors find that after accounting for income, German and non-German respondents are similarly trusting. Finally, Andrabi and Das (2017) argue against the idea of religion as a persistent factor in explaining trust. They show that the level of trust toward foreigners in the Muslim population in Pakistan increased as a consequence of the inflow of foreign aid in a region hit by the earthquake in 2007.

Secondly, this paper is also grounded on the literature which analyses the relationship between adversarial environmental conditions and social cohesion. The idea is that the natural environment is a key determinant of social cohesion (Kuper and Kropelin 2006). However, there is no agreement on how the environment influences this variable. On the one hand, tougher environmental conditions may increase social cohesion. In a hostile environment people may be forced to cooperate more in order to survive or, alternatively, stronger evolutionary pressures may select for more cooperative communities (Henrich and Henrich 2007). On the other hand, harsher environmental conditions could bring about more conflict and, consequently, less social cohesion as a result of increased competition for scarce resources (Brancati 2007; Hsiang et al. 2013).

The empirical work studying the link between environmental conditions and trust has mainly focused on the effect of natural disasters on this variable. The empirical evidence on this issue is mixed. Fleming et al. (2014), Papanikolaou et al. (2012) and Ahsan (2014) argue that natural disasters had a negative or no impact on trust. Fleming et al. (2014), for instance, show that there are no differences in the level of trust between Chilean regions affected or unaffected by an earthquake in 2010, while reciprocity was lower in the regions involved. Conversely, several studies find out that natural disasters have a positive impact on trust (Andrabi and Das 2017; Calo-Blanco et al. 2017; Cassar et al. 2017; Dussaillant and Guzmán 2014; Kang and Skidmore 2018). Cassar et al. (2017) find that people living in the villages hit by the 2004 Indian Ocean tsunami in Thailand show higher levels of trust and trustworthiness than those living in non-affected villages. Dussaillant and Guzmán (2014) found evidence of a positive impact of the 2010 earthquake in Chile on levels of trust in the medium run.

Finally, I also draw insights from the natural science literature stating that variations in minimum temperature negatively impact rice production. These variations increase the maintenance respiration requirement of the crops and shorten the time to maturity. As a result, they reduce net growth and productivity. Natural experiments have confirmed this hypothesis. Peng et al. (2004) analyzed weather data from 1979 to 2003 from irrigated field experiments at the International Rice Research Institute Farm. Rice yield declined by 10% for each 1°C increase in growing-season minimum temperature in the dry season, whereas the effect of maximum temperature on crop yield was insignificant. Welch et al. (2010) studied 277 farm-managed rice fields in six major production countries finding that a higher minimum temperature reduced yield.

3 Data and Method

The data used in this paper is a balanced version of the VARHS household dataset^[4] including only the families continuously interviewed over four waves of the survey, i.e. 2008, 2010, 2012, 2014. Therefore, the resulting dataset contains information on 2134 households living in 459 communes. Figure 1 maps the communes involved in our analysis.

Figure 1:

VARHS communes included in our sample. Source: Author processing of VARHS data.

The dependent variable is a binary variable equal to one if in the survey the interviewee answers that she trusts others, zero if she doesn’t. More specifically, I use two different variables to proxy trust. The first variable, trust (a), corresponds to the question “Most people are basically honest and can be trusted”. The second variable, trust (b), corresponds to the question in the survey “In this commune one has to be careful, there are people you cannot trust”. There are significant differences between these two measures. In the first case, about 85% say that they trust other people; in the latter, about 35%. In light of these differences, I included both variables in my analysis.

The independent variable measures variations of the minimum temperature in June for each commune in the dataset. This variable indicates by how much a particular month was warmer or colder than the average monthly temperature. These variations are calculated as deviations from the average minimum temperature in June. Formally,

where tmn level _jt denotes the minimum temperature in June recorded in each commune j in year t, and tmn avg _j denotes the average minimum temperature in June.^[5]

As mentioned before, the key climate variable of interest in this paper are the variations in the minimum temperature in June. I argue that these variations may have a large impact in Viet Nam where rice is the staple food, the agricultural sector is still a prominent part of the economy and rice is the main crop grown in the country. Importantly, the household dataset used in this paper confirms this: 82.33% of households say that they earn an income from agriculture with rice accounting for by far the largest crop. However, to control for the level of water available – also a key factor in determining agricultural productivity – I include deviations in rainfall in June as a control. The deviations are calculated as in equation (1) above.

Climate data are drawn from the CRU TS4.01 dataset of the Climatic Research Unit at the University of East Anglia which includes monthly rainfall figures and the minimum, mean and maximum temperature from 1901 to 2016 on a 0.5 × 0.5-degree grid.^[6] To combine VARHS with climate data, I first imputed the latitude and longitude of 459 VARHS communes using data from GADM (Global Administrative Areas).^[7] I then matched the latitude and longitude of the VARHS communes with the 0.5 × 0.5-degree cells of the CRU data. Thus, I attributed the recorded minimum temperature and rainfall in June to each commune.

Other regressors measure socio-economic characteristics of the households and are in line with the literature on the determinants of trust as shown in the review of the literature in Section 2 (Alesina and La Ferrara 2002; Bjørnskov 2007; Glaeser et al. 2002). Specifically, I include some socio-demographic characteristics such as the household head’s education, age, marital status and ethnicity. Furthermore, I also control for some economic features of the households: their income and the land owned as a proxy of wealth. Finally, I account for the quantity of money received from external sources, such as loans and public and private transfers. The literature in this field shows a possible association between trust and being the recipient of external funds (Accetturo, De Blasio, and Ricci 2014). Definitions and descriptive statistics for the data used in the analysis are provided in Table 1. Correlation matrix is reported in Table A.1 in the appendix.

To investigate the relationship between variations in the minimum temperature and trust in Vietnamese rural households, I first regress trust on deviations of the minimum temperature as shown in equation (2):

where trust _it is a dummy variable indicating with 1 if the respondent trusts others and with zero otherwise (the subscript i indicates the household). Dev min temperature June _jt is the deviation of the minimum temperature in June recorded in the commune where the household is located (the subscript j indicates the commune). Similarly, Dev rainfall June _jt indicates deviations in rainfall in June for each commune. The model also includes a vector of controls at the household level, X _it as well as time and regional fixed effects respectively y _t and a _r . ε _it is the error term.

Time fixed effects account for those factors that could impact on trends. Regional fixed effects capture unobservable time-invariant regional specific factors that could influence trust such as cultural factors. As an alternative specification I use household rather than regional fixed effects.

An OLS estimator is used to estimate the model presented in equation (2). Since the dependent variable is binary, the estimated model is a linear probability model (LPM). The LPM is frequently adopted in economics (Wooldridge 2002) because it gives approximations for marginal effects that are like estimates resulting from a non-linear model (Angrist and Pischke 2009). The use of the model, however, can be subject to criticism and alternative estimators are usually adopted. In the Appendix, I address this issue as a further robustness check (Table A.2).

4 Results

The main results indicate a positive and significant association between increases in the deviation in the minimum temperature in June and trust in rural Vietnamese households. This may suggest that variations of the minimum temperature in the key month of the rice growing season may lead to an increase in social cohesion in Vietnamese communes. Furthermore, the coefficient is positive and significant for both the measures of trust adopted. There are, however, some differences between the two variables in the magnitude of the coefficient. An increase by 0.1 degrees in the deviation of the minimum temperature is associated with an increase of 1% in the probability that the respondent answers yes to the question “Most people are basically honest and can be trusted”, trust (a). On the other hand, an increase of 0.1 degrees in the deviation of the minimum temperature is linked with a 3% rise in the probability that the respondent answers no to the question “In this commune one has to be careful, there are people you cannot trust”, trust (b).

Regarding the covariates, the results show that the ethnicity of the head of the household is negatively related to trust. Being Kinh, decreases the probability of being trusting by about 1%. However, I found no relationship with household income and trust and weak evidence that household wealth is positively associated with trust. The coefficient for the specification in column 4 in Table 2 is positive and significant at 10%. It indicates that an increase of 1% in the land owned by a household, increases the probability that the respondent is trusting by 1%. Furthermore, receipt of a public transfer is positively and significantly related to trust, but this result is valid only for the first measure of trust (trust a). The increase of 1% in the quantity of transfers received is associated with an increase of 0.3% in the probability that the respondent trusts others. Conversely, there is a negative association between obtaining an informal loan and trust, but only for the second measure of trust, trust (b). Raising the quantity of informal loans obtained by 1% decreases the probability that the respondent is trusting by 0.8%.

To check the robustness of our results, two alternative measures of trust are used as indicated in the literature (Alesina and La Ferrara 2002). The first alternative independent variable (past) is a binary variable which is equal to one if in the survey the interviewer answers “yes” to the question “This commune has prospered in the last five years” and zero otherwise. The second alternative variable (future) is a dummy variable where 1 corresponds to “yes” to the question “This commune will prosper in the coming five years”, and zero otherwise.

Results are robust (Table 3). They show that the association between trust and deviations of the minimum temperature in June is positive and significant at 10%. The coefficient is similar for all four specifications. An increase by 0.1 degree in the deviation of the minimum temperature is related to an increase of 0.5% that the respondent is trusting about both the past and the future of the commune where she lives. Among the covariates, I found weak evidence that income, education and household wealth are positively related to trust. When controlling for regional fixed effects (specifications in columns 1 and 3) an increase by 1% in household income is associated with an increase of about 2% in the probability of the respondent being trusting. Similarly, the rise of 1% in the land owned by the household leads to an increase of 1% in trust. Furthermore, there is also a positive and significant relationship between private transfers and trust. When the quantity of transfers received by the household increases by 1%, the probability of the respondent being trusting rises by 0.3%

Second, as a further robustness check, I also run a placebo test using the deviation of maximum temperature in June as independent variable. As explained above, this paper adopts the minimum temperature as the main independent variable. The intuition is that variations in the minimum temperature are more relevant in a country like Viet Nam than variations in other climate variables. Increases in the minimum temperature negatively impact on the production of rice which is the staple food in Viet Nam as well as the main commercial crop. Hence, a significant coefficient for maximum temperature would contradict the main findings of this paper.

The results in Table 4 show no significant relationship between the deviation of maximum temperature and trust in almost all specifications, thus confirming the hypothesis of this paper.

5 Conclusions

This paper investigates the relationship between trust and variations in minimum temperatures using Viet Nam as a testbed. The idea is that in a country where rice is the staple food and an important element of the economy, the relevant climate measure to use is the minimum temperature in June. Therefore, relying on a dataset on Vietnamese rural households from 2008 to 2014, I analyze the impact of deviations in the minimum temperature on two alternative measures of trust. The results indicate a positive and significant association between increases in the deviation in the minimum temperature in June and trust in rural Vietnamese households. An increase by 0.1 degrees in the deviation of the minimum temperature is associated with an increase of 1–3% in the probability of the respondent being trusting.

Further research must delve into the specific casual mechanism linking harsher environmental conditions to social cohesion. There is still a gap in the literature about why, faced by deteriorating climate conditions, some communities choose to cooperate and others to compete over scarce resources. This can explain why the empirical evidence relating natural disasters to trust is mixed. Needless to say, the search for possible explanations linking climate events and trust must be specific to the area analyzed. The impact of the environment changes according to the institutional and economic characteristics of the area studied.