THE EFFECTS OF FISCAL POLICY ON HOUSEHOLDS DURING THE COVID-19 PANDEMIC: EVIDENCE FROM EMERGING ECONOMIES

In response to the economic crisis created by the COVID-19 pandemic, many governments provided financial assistance to households. Using representative consumer surveys conducted during the pandemic in 2020, we examine the effects of this fiscal policy instrument on households in two emerging economies, Vietnam and Thailand. Our paper contributes to the literature by studying consumer sentiment and durable spending responses to government financial support and the underlying transmission channels for these responses. We find that government support improves consumer sentiment and increases the likelihood of durable spending. Possible channels for these effects include more optimistic macroeconomic expectations and higher trust in the government's ability to deal with the pandemic, as well as less concern about the general impact of the crisis. We also find that financial support improves individuals' mental health and life satisfaction. Our results suggest that government financial support not only helps stimulate the economy but also enhances people's well-being more generally.


Introduction
In response to the economic crisis created by the COVID-19 pandemic, many governments provided nancial support to households.In light of the substantial public funds involved, it is important to assess the eectiveness of this scal policy instrument.Indeed, a growing literature studies the eect of government nancial support on consumer spending, including, among others, Baker et al. (2020), Bayer et al. (2020), Christelis et al. (2020), Coibion et al. (2020a), and Karger and Rajan (2020).Our paper contributes to this literature by studying consumer sentiment and durable spending responses to government nancial support and the underlying transmission channels for these responses.In particular, we focus on the transmission channels of macroeconomic expectations, trust in the government in dealing with the pandemic, and households concerns due to the pandemic.
As an important part of the scal policy package created to reduce the economic damage caused by the COVID-19 pandemic, the governments of Vietnam and Thailand provided nancial support to qualifying households for a period of up to three months, typically from April to June 2020.The aid targeted individuals whose jobs were aected by the pandemic as well as poor households more generally (in Vietnam) and farmers (in Thailand).As a consequence of these programs, each eligible individual received nancial assistance ranging from $35 to $240 per month in Vietnam and up to $412 per month in Thailand (U.S. dollar in PPP in 2019).This scal policy response was unprecedented in both countries.
To assess the impact of nancial support on households, our study uses two novel Internet-based consumer surveys conducted in Vietnam and Thailand in May and December 2020.For each country and wave, the surveys include about 1,000 respondents aged 18 or older.Our analysis focuses on the second wave, as it contains information about government cash transfers received by individuals.We employ the rst wave mainly for robustness checks.According to our survey, about 30% and 60% of Vietnamese and Thai respondents, respectively, beneted from pandemic-related nancial support from the government.Our survey combines various measures of macroeconomic expectations, an indicator for trust in the government, various household concerns caused by the pandemic, and subjective well-being, which makes it possible to study not only the eects of nancial support on consumer sentiment, but also identify some of their underlying channels.
Our main ndings suggest that the scal spending programs had statistically significant and economically substantial eects.Respondents who received nancial support from the government because of the pandemic show a 7% and 16% increase in consumer sentiment relative to the average value of consumer sentiment in Vietnam and Thailand, respectively.The likelihood that they bought durable goods in the period from May to December 2020 rises by 22 and 13 percentage points (pp), respectively.Regarding future consumption, nancial assistance recipients indicate an increase in the probability that they will certainly buy durable goods in the next 12 months by 7 pp in Vietnam and 6 pp in Thailand.Moreover, we nd that nancial support increases individuals' mental well-being, for example, by inducing feelings such as calm and being less nervous, and their overall life satisfaction.
Further analysis suggests several possible channels through which nancial support from the government leads to an increase in consumer sentiment and durable spending.
First, nancial assistance recipients express more optimistic macroeconomic expectations, such as lower expected ination and unemployment rates as well as higher expected economic growth.Second, they trust more in the government's ability to mitigate the side eects of social distancing on the economy.They are also more likely to state that the government has been doing a good job in terms of supporting households and rms aected by the pandemic.Finally, government nancial support reduces respondents' pandemicinduced concerns about their health, job security, personal nancial situation, and the economy in general.Using mediation analysis, we discover that all these channels aect consumer sentiment in a signicantly positive way.In both countries, the largest indirect eect on consumer sentiment is due to people's assessment of government policies supportive of rms and households aected by the pandemic.
Our results control for a large number of socio-demographic and other individual characteristics, which ensures that the eects of government nancial support on consumer sentiment and durable spending, as well as the transmission channels mentioned above, are not explained by any of these factors.As the scal programs in both countries target specic groups, we use Heckman selection models to control for a potential selection bias.Since our core results remain unchanged, we conclude that selection bias is not a major concern.Furthermore, we employ the information from the rst survey wave conducted in May 2020 to control for lagged values of our left-hand-side variables in the baseline models.These robustness checks show that our results remain mostly unchanged, implying that government nancial support also has positive dynamic eects on the change in our variables of interest within individuals over a period of seven months.
Our paper makes two main contributions to the current literature on the eect of scal policy on households' consumption during the pandemic (Baker et al., 2020;Bayer et al., 2020;Christelis et al., 2020;Coibion et al., 2020a;Karger and Rajan, 2020).First, we shed light on the mechanisms underlying the consumption response to the government cash transfer, particularly the transmission eect via macroeconomic expectations, trust in the government's ability to deal with the pandemic, and households' concerns due to the pandemic.To the best of our knowledge, this paper is one of the rst attempts to consider such a variety of possible channels.By doing so, we also add to the literature studying the direct eect of the pandemic on aggregate expectations (Binder, 2020;Coibion et al., 2020b), trust in government (Devine et al., 2020;Sibley et al., 2020) and household concerns (Binder, 2020;Christelis et al., 2020).
Second, we provide new empirical evidence on the eect of cash transfers during the COVID-19 pandemic on households' consumption in the emerging economies Vietnam and Thailand, whereas the current literature focuses on industrialized economies.For instance, in the United States, a burgeoning literature studies the eect of the one-time cash transfers from the CARES Act in response to the COVID-19 pandemic.Karger and Rajan (2020) and Baker et al. (2020) report that this policy increases recipients' spending immediately upon receiving the cash payments and Bayer et al. (2020) show positive output multipliers for both unconditional and conditional cash payments.In addition, Baker et al. (2020) and Coibion et al. (2020a) provide evidence that consumers spend more on nondurable goods and less on durables compared to the economic stimulus in 2008.Christelis et al. (2020) survey consumers in the six largest economies of the euro area and nd that household concerns due to the pandemic reduce consumption.
We add to this literature by analyzing two emerging countries from the same region in Asia.Moreover, while the main measures describing households' consumption patterns in the previous literature are total household consumption expenditure and some of its subcategories, we focus on individual consumer sentiment and durable spending.The individual index of consumer sentiment is based on responses to the same questions that are used to calculate the aggregate consumer sentiment index in the University of Michigan Surveys of Consumers (Bui et al., 2021): consumers' current and expected nancial situation, their macroeconomic expectations, and their readiness to purchase durable goods.
Regarding durable spending, we measure not only respondents' actual spending, but also their plans to buy durable goods in the next 12 months.We focus on durable consumption, as nondurable good consumption is dominated by less elastic expenditure categories, such as food and clothing.
Our paper is also more generally related to a large body of literature studying the impact of cash transfers on households in emerging economies before the pandemic, such as the eect on reducing poverty, improving health conditions, and fostering economic autonomy (see Bastagli et al. (2016) for a review).For Kenya, Egger et al. (2019) show large positive eects of cash transfers on household income and consumption in rural areas and Haushofer and Shapiro (2016) nd a strong consumption response to unconditional cash transfers at the village and household levels.Moreover, lump-sum transfers are more likely to be spent on durables, a nding that motivated us to focus our study on durables.
Moving beyond consumption, Haushofer and Shapiro (2016) also report an increase in psychological well-being (happiness, life satisfaction, reduction in stress and depression), which is consistent with research conducted by Lund et al. (2011), who demonstrate that conditional cash transfer and asset promotion programs have positive mental health benets.Finally, Evans et al. (2019) show for Tanzania that cash transfers signicantly enhance trust in elected leaders.
Our research has at least three important policy implications.(i) Government nancial support is eective in terms of stimulating current and planned consumption spending.
(ii) The cash transfers improve people's economic outlook by making consumers more optimistic about their own economic situation as well as the general macroeconomic situation.(iii) These transfers have various eects over and above a direct consumption response.They signicantly bolster households' trust in the government, reduce personal concerns, and raise subjective well-being.Thus, at least during times of crisis, government nancial assistance appears to be a highly eective scal policy instrument.
The rest of this paper is organized as follows.Section 2 presents the data, Section 3 shows the results and robustness checks, and Section 4 concludes.

Data
To study the impact of COVID-19 on households' well-being and economic situation, we implemented two waves of online surveys during May and December 2020 in Vietnam and Thailand.In Vietnam (Thailand), 3,300 respondents (2,200 respondents) were surveyed over the period May 49, 2020(May 410, 2020).We conducted a second wave over the period December 1827, 2020 and re-interviewed 1,016 Vietnamese and 1,189 Thai respondents from the rst wave.Our surveys were conducted by GMO-Z.comRUNSYS-TEM, one of the largest private market research and public opinion survey companies in South-East Asia.The survey company has a large number of registered participants who are familiar with online surveys.All participants who complete the survey receive reward points, which can be exchanged for gifts.
Our analysis mainly relies on the second wave, in which we additionally asked respondents whether they had received any nancial support from the government due to the pandemic.This is a unique dataset because it combines consumer sentiment indicators, actual and planned durable spending, macroeconomic expectations, trust in the government, household concerns, and subjective well-being.To ensure representativeness of our samples, we construct population weights based on the respective national distribution of age, education, and share of people living in an urban area and employ these throughout our empirical analysis.In all estimations, we control for a large number of demographic characteristics, such as age, gender, marital status, living area, income, employment status, and health conditions, as well as pandemic-induced job loss and income loss.

Key variables of interest
Our main explanatory variable is a dummy from the second survey wave conducted in December 2020 indicating whether respondents and/or other household members received nancial support from the government due to the COVID-19 crisis (n_support ).We study the eect of government nancial support on various outcome variables.First, we employ the measure of individual consumer sentiment proposed by Bui et al. (2021), which is a simple average of the following ve qualitative questions: (i) perceptions about the nancial situation of the household in the past 12 months, (ii) expectations about the nancial situation of the household in the next 12 months, (iii) expectations about the national business condition in the next 12 months, (iv) expectations about the national economic situation in the next ve years, and (v) current readiness to spend on durables.Thus, individual consumer sentiment ranges from 1 to 5, with higher values denoting more optimistic sentiment.Note that these ve questions are used in the University of Michigan Surveys of Consumers to calculate an aggregate consumer sentiment index.In addition, we inquire whether respondents purchased durable goods between May and December 2020 (purchased_durable ) and ask them about their plans to buy durable goods in the next 12 months (plans_to_buy_durables ).
Other key variables of interest include subjective well-being (measured by feeling calm or nervous, and life satisfaction), macroeconomic expectations (with respect to ination, unemployment, economic growth), assessment of government in supporting rms (govt_support_rm ) and households (govt_support_household ) aected by the pandemic, trust in the government in mitigating the negative eects of the pandemic on the economy (govt_trust_econ ), and household concerns due to COVID-19 (with respect to health, job security, nancial situation, and the economy in general).In the Appendix, we show summary statistics for all our variables of interest (Table A1), as well as the exact wording of the underlying questions (see Appendix A.4).
In the baseline analysis, we exclude respondents who do not know the answer to or do not voice an opinion on the following topics: macroeconomic expectations, assessment of and trust in government, personal concerns, and consumer sentiment.Our nal samples consist of 847 Vietnamese and 713 Thai respondents who participated in both waves.As a robustness check, we follow the approach taken by the University of Michigan Surveys of Consumers and assume that respondents who state that they do not know the answer or do not form opinions are expressing a neutral position (e.g., expecting no change or viewing policies as neither good nor bad or being not concerned at all).
1 We re-estimate our baseline results with these extended samples, which include 1,016 observations for Vietnam and 1,189 observations for Thailand (see Appendix A.3.3).

Stylized Facts about the Impact of COVID-19 on Households and Assessment of Government Reaction
Our analysis reveals that the COVID-19 pandemic has had severe negative eects on Vietnamese and Thai consumers.Figure 1  ing hours.This and other factors contribute to a situation in which the vast majority of households in both countries (approximately 80%) report income losses.While these numbers are similar to those measured in other surveys conducted in the same countries during the COVID-19 pandemic (Morgan and Trinh, 2020;MDRI and UNDP, 2020), 2 they are considerably higher than those reported from industrialized countries (Parker et al., 2020;Major et al., 2020).
3 Second, the pandemic exacerbates personal concerns and reduces mental health.In both countries, consumers have similar concerns about their health, job security, and personal nances, as well as about the whole economy.Only a minority of respondents have no concerns about these topics, with an underlying range of 8-13% and 3-5% in Vietnam and Thailand, respectively.Consumers worry most about the eects of COVID-19 on their households' nancial situation and the whole economy (Vietnam: 48% somewhat worried and 43-44% very worried; Thailand: 34-40% somewhat worried and 57-62% very worried).These results suggest a high degree of awareness about the seriousness of the pandemic in both countries, which corresponds to the worldwide personal anxiety due to COVID-19 shown in other surveys (Fetzer et al., 2020).
Despite these similarities, Thai and Vietnamese respondents report opposite views on their governments' eorts to mitigate the negative economic eects of the pandemic, as shown in Figure 2 (similar results are also reported in Dölitzsch (2020) and Fetzer et al. ( 4 Although almost 60% of our Thai respondents stated that they or someone in their household already received nancial assistance, they are neither content with their government's support to individuals and households (44% answer that the government does a poor job, 43% answer fair job, and only 13% say good job), nor with its support to rms (48% say poor job, 39% say fair job, and only 13% state good job).

5
In light of this assessment, they put little trust in their government's ability to return the economy to pre-pandemic levels (about 49% have no trust, 29% have a neutral view, and only 21% have at least some trust).In contrast, most Vietnamese people believe that their government does well in terms of support to individuals and households (only 17% say poor job, 31% answer fair job, and 52% state good job) as well as support to rms (only 15% say poor job, 37% answer fair job, and 49% state good job).Moreover, they rmly trust that their government will revive the economy (about 5% have no trust, 27% have a neutral view, and 67% have at least some trust).These results are astonishing in light of the fact that less than one-third (30%) of Vietnamese respondents actually beneted from government nancial assistance.
6 The diverging results between the two countries can be linked to the pre-crisis level of government assessment, which was much more positive in Vietnam (66% respondents say good job) than in Thailand (14% respondents say good job).
This suggests that government trust is to some extent deep-rooted and only partially inuenced by actual government policy.
6 According to survey data from September 2020, about 20% of Vietnamese received nancial support from the government (MDRI and UNDP ( 2020)).As of 25 December 2020, VND 12.8 trillion had been disbursed to roughly 13 million people and 30,569 household businesses (Ngan-Anh, 2021).3 Results We estimate the eect of COVID-19-related nancial support from the government on our dependent variables of interest using the following equation: pp lower likelihood of reporting that they strongly agree with the statement that they are nervous when thinking about their current situation.In the case of Thailand, we nd a 3 pp higher probability that beneciaries strongly agree with the statement that they are calm and relaxed.Vietnamese and Thai respondents who received nancial support have an increased likelihood of 4 pp and 2 pp, respectively, of answering that they are totally satised with their life as a whole.While the magnitude of these eects is small, the eects corroborate our previous results that nancial support makes individuals more optimistic with respect to consumer sentiment and willingness to spend on durables.

Transmission Channels of Financial Support
In this subsection, we investigate three potential channels for explaining how government cash transfers aect economic outcomes at the household level, that is, consumer sentiment and durable spending.Do consumers spend more because they (i) are more optimistic about future macroeconomic development, (ii) believe the government has been doing a good job in terms of mitigating the negative eects of the pandemic on the economy, or (iii) are less concerned about the eect of the pandemic on their health, job security, nancial situation, and the economy in general?
(i) Macroeconomic Expectations: Table 3 shows the eect of government support on qualitative measures of macroeconomic expectations with respect to ination, unemployment, and economic growth (GDP).In both countries, receipt of nancial support leads to expectations of lower ination and unemployment and higher economic growth in the next 12 months.For Vietnamese respondents, the likelihood that beneciaries of nancial support state that ination and unemployment will increase signicantly declines by 9 pp and 5 pp, respectively.In the Thai sample, nancial support reduces the likelihood of stating that unemployment (GDP growth) will increase by 0.8 pp (4 pp).All the estimated eects are statistically signicant at conventional levels, except for ination expectations (GDP growth expectations) in the Thai (Vietnamese) sample.The eects on expected unemployment and economic growth are intuitive.From a macroeconomic perspective, the negative eect of government nancial support on ination expectations is somewhat surprising, as an increase in government spending might be expected to raise ination.However, one explanation for this result is that nancial support recipients have more trust in the government being able to manage the economy, which includes keeping the ination rate under control.This interpretation is consistent with our results from studying government trust.A second explanation arises from the observation that high ination is often interpreted as a negative economic signal.Thus, low ination expectations may mirror overall optimism regarding future macroeconomic development (Binder, 2020).This interpretation is in line with our ndings, shown below, regarding the eect of nancial support on households' concerns about the general economy.
(ii) Assessment of and Trust in the Government Reaction: Table 4 shows signicantly positive eects of nancial support on the assessment of and trust in the government in dealing with negative spillovers to the economy from the pandemic.The likelihood that beneciaries state that the government has been doing a good job to support rms and households aected by the pandemic increases by about 2225 pp in the Vietnamese sample and by 1314 pp in the Thai sample.Moreover, beneciaries in Vietnam and Thailand have a 9 pp and 5 pp, respectively, higher probability of saying that they strongly trust the government to mitigate the negative side-eects on the economy of social distancing.Our results remain generally unchanged when we additionally control for the assessment of the government's macroeconomic policies before the pandemic, implying that nancial support does aect assessment of and trust in the government's responses to the pandemic.Although the marginal eects of receiving nancial support on government trust are relatively larger in Vietnam, they are not statistically dierent from those estimated for Thailand.Note, however, that the average level of assessment of and trust in the government is signicantly lower in Thailand.to answer that because of the pandemic they are very concerned about their health and job security (19 pp), nancial situation (13 pp), and the economy in general (18 pp).In Thailand, government nancial support reduces the probability of respondents reporting that they are very concerned about the economy in general by 11 pp, whereas the negative marginal eects for the other concerns are not statistically signicant.However, the eects of nancial support on concerns about job security and nancial situation become statistically signicant when using the extended sample (see Section 2 and Table A16 in the Appendix).This suggests that these insignicant eects are due to the smaller sample size in the baseline analysis.As the magnitude of the estimated eects is large, we conclude that nancial support plays an important role in mitigating household distress during the pandemic.We conduct a mediation analysis to measure how macroeconomic expectations, trust in the government, and household concerns aect the impact of nancial support on consumer sentiment and durable spending.Following Imai et al. (2010), we estimate the indirect eect of nancial support through each of these factors.To facilitate implementation of the mediation analysis using OLS, we assume that our outcome variables, which proxy the three transmission channels, are continuous.Figure 3 shows the relative inuence (in percent) of the indirect eects on the total eect of nancial support on consumer sentiment and durable spending.For both countries, the results show that all three channels mediate the eect of nancial support on consumer sentiment and plans to buy durables at a 10 percent level of signicance.Regarding actual durable spending in Vietnam, we nd that the positive eect of nancial support is reduced for those with personal concerns about job security and the economy, whereas all other channels increase the eect on spending.Our results also suggest that in both countries, the eect of nancial assistance on sentiment or durable spending is mediated most strongly via consumers' assessment of government support.

Heterogeneity Across Household Characteristics
In this subsection, we examine whether government nancial support has heterogeneous eects on consumer sentiment.Employing our rst survey wave from May 2020, we allow for potential heterogeneity associated with income quartiles, expenditure vs. income of the household, and net asset position.We use these household economic characteristics measured in May because government nancial support in both countries was initiated based on household conditions during the early phase of the pandemic and most of the nancial support in the two countries was provided between April and June 2020.We also examine the heterogeneous eect of nancial support across various demographic characteristics of households, including age, education, and rural/urban residence.We test for the potential importance of heterogeneous eects by regressing consumer sentiment on each of the above characteristics as well as these variables interacted with the dummy capturing receipt of nancial support from the government.
Figure 4 presents heterogeneous marginal eects of nancial support on consumer sentiment and their 90% condence intervals.In the Appendix, we show the heterogeneous eects of government nancial support on durable spending (Figures A1 and A2).In general, we nd that the estimated point eects dier only slightly across categories and are not statistically dierent from each other.However, in a number of cases, we discover that the estimated eects are statistically dierent from zero only for some groups but not for others.Thus, we conclude that the eects of government nancial assistance on consumer sentiment and durable spending in both countries are quite homogeneous across important socio-demographic and economic groups.For both countries, these results suggest that the estimated eects of government nancial support discussed in the previous sections are not inuenced by any specic types of household groups.

Robustness Checks
To this point, our regressions have controlled for household income per capita, employment status, dummies measuring whether any household members experienced job loss or income losses due to the pandemic, subjective health assessment, and various demographics, thus implying that our results are not explained by any of these characteristics.
It is still possible that omitted variables that our controls do not fully capture aect both the probability of receiving government nancial support and consumer sentiment, such as social status.However, in our view, these omitted variables are more likely to cause our estimates to be downward biased.For instance, due to the design of the program, those who have lower social status or are less well-o are more likely to receive nancial support from the government.However, this group of people is typically less optimistic about the future, that is, more likely to express more pessimistic consumer sentiment and have poorer macroeconomic expectations, as shown by Das et al. (2020).This implies that the true eects might be even larger than our results suggest.
As the scal programs in both countries target specic groups, we check our results using Heckman selection models based on the following procedure.In the rst-step, the selection probit regression, we regress n_suport on a set of demographic control variables from our rst survey wave conducted in May 2020 and calculate the inverse Mills ratio (IMR).In the second-step, we use the same models as in Equation 1of the baseline analysis, but additionally control for the IMR.Tables A2-A6 in the Appendix show the estimates from the second-step regressions.Our baseline results remain unchanged, suggesting that our conclusions do not suer from selection bias.
To capture possible autoregressive behavior of our dependent variables, we integrate information from our May survey into our data from December 2020.Equation 2 illustrates that our model now contains dynamic eects in the form of lagged dependent variables: Note that some outcome variables in the baseline models were not elicited in the rst wave, such as the durable spending measures and the assessment of the government's response in terms of supporting households and rms aected by the pandemic.Tables A7-A10 in the Appendix show that estimating Equation 2 barely inuences our previous conclusions.In many regressions, especially those for Vietnam, we nd signicant lags of our respective dependent variables.This enables us to compute the long-term eect of government nancial support on the changes in our variables of interest as β/(1−η).Using Tables A7 to A11, we discover that the long-term eects are up to about 20% larger than the short-term estimates.For instance, the long-term inuence of nancial support on consumer sentiment is about 17% larger than its short-term inuence.Overall, the nding that our results are robust to any persistence in the dependent variables suggests that government nancial support has positive dynamic eects on the change in our variables of interest within households over a period of more than half a year.
Finally, we re-estimate our baseline models using extended samples.As discussed in Section 2, for this purpose we assume that those who state that they do not know the answer or report that they do not form opinions are considered as having a neutral position.Tables A12-A16 in the Appendix set out the results, which are generally unchanged from our baseline results.

Conclusion
In this paper, we study the reaction of consumers in Vietnam and Thailand to their respective government's nancial support programs during the COVID-19 pandemic.We utilize two waves of representative population surveys conducted in May and December 2020 in these two emerging countries of Southeast Asia.We discover that by December 2020, government nancial support had reached about 30% of citizens in Vietnam and 60% in Thailand.In our survey, we nd that nancial support has statistically signicant and economically notable eects on indicators of future economic activity as well as indicators of people's well-being.For instance, Vietnamese and Thai respondents who received COVID-19-related cash transfers show a 7% and 16% increase in consumer sentiment, respectively.The probability that they purchased durable goods in the period from May to December 2020 rises by 22 and 13 pp in Vietnam and Thailand, respectively.Regarding future consumption, for those who beneted from government nancial assistance, we estimate an increase in the likelihood that they will certainly buy durable goods in the next 12 months.At 6 pp in Vietnam and 5 pp in Thailand, the magnitude of the eect is moderate, but similar across the two countries.Furthermore, we nd that beneting from government nancial support programs increases individuals' mental well-being, expressed through feeling calm and less nervous, and increases recipients stated value of life satisfaction.
We identify three channels through which these eects may manifest.First, respondents receiving nancial assistance from the government express more optimism about the macroeconomic outlook, such as lower expected ination and unemployment rates as well as higher expected economic growth.Second, these respondents have a higher degree of trust in the government's ability to deal with the negative side-eects of COVID-19 on the economy, for example, employment and income losses.Moreover, recipients of cash transfers have a greater probability of answering that the government has been doing a good job in terms of supporting households and rms aected by the pandemic.Third, government cash transfers appear to alleviate various concerns arising from the crisis, such as concerns over health, job security, nancial situation, and the general economic situation.
Conducting a mediation analysis, we demonstrate that all these channels play a significantly positive role in shaping the inuence of government nancial support on consumer sentiment.The analysis also reveals that the largest individual indirect eect of nancial support on consumer sentiment is via people's assessment of and trust in the government in supporting rms and households aected by the pandemic.In our study, we control for many socio-demographic and economic variables.Thus, the impact of government nancial support on consumer sentiment, durable spending, and subjective well-being, as well as the transmission channels mentioned above, are not due to these controls.Moreover, we use a Heckman approach to control for non-randomness in the selection of individuals to receive government nancial support.Finally, we use the information from the two survey waves in each of the countries to control for lagged values of our left-hand side variables.This allows us to estimate the long-term eects of government nancial support, and we discover that the short-run results discussed in the baseline models are likely lower bounds of the actual eects.Overall, our conclusions are robust to all these extension.
An important nding from our investigation is that government nancial assistance during a crisis appears to have a number of eects that go beyond a direct consumption response.First, such support makes people more optimistic about their future personal economic situation as well as about the aggregate economic situation.Second, nancial assistance helps sustain trust in the government, which may be important when a country experiences a prolonged lockdown and other severe measures.Third, the psychological pressure due to personal concerns lessens and this coincides with an improvement in subjective well-being.Thus, when designing scal policy in the form of cash transfers, governments are well advised to factor these additional positive spillovers into the process.

A Appendix
A.1 Summary statistics Note: This table shows the summary statistics of our key variables of interest based on population weights.These samples exclude respondents who do not know the answer or who do not have opinions on the survey questions of our key variables.Section A.4 show the exact wording of these questions.

2
Conducting population surveys in eight South-East Asian countries during May and July 2020, Mor- gan and Trinh (2020) show that about 50% of households in Thailand and Vietnam experienced job losses and/or a reduced work load and two-thirds of respondents in Vietnam and three-quarters in Thailand report income losses.Another survey in Vietnam conducted during September 2020 nds that 65% of respondents report income losses due to the pandemic (MDRI and UNDP (2020)).

Figure
Figure 1: The Impact of COVID-19 on Households

Figure 2 :
Figure 2: Financial Support and the Assessment of Government Reaction

Figure 3 :
Figure 3: The Proportion of Indirect Eects in the Total Eect of Financial Support on Consumer Sentiment and Durable Spending

Figure
Figure A1: The Heterogeneous Eects of Government Financial Support on Purchased Durables with 90% Condence Intervals

•
life_satisfaction: All things considered, how satised are you with your life as a whole?[Totally dissatised, Partly dissatised, Neither dissatised nor satised, Partly satised, Totally satised] Macroeconomic expectations • ination_expectations: How do you think prices in general (which are used to measure the ination rate) will develop over the next 12 months compared to the previous 12 months?They will [Decrease a lot, Decrease a little, Stay about the same, Increase a little, Increase a lot, I do not form opinions about future general price level, Don't know.]• unemployment_expectations: How do you think unemployment will develop over the next 12 months compared to the previous 12 months?It will [Decrease a lot, Decrease a little, Stay about the same, Increase a little, Increase a lot, I do not form opinions about future unemployment, Don't know] • gdp_expectations: How do you think national economic growth (GDP growth) will develop over the next 12 months compared to the previous 12 months?It will [Decrease a lot, Decrease a little, Stay about the same, Increase a little, Increase a lot, I do not form opinions about future economic growth, Don't know] Assessment of and trust in the government • govt_assessment_normal_times: As to the macroeconomic policy of the government before the COVID-19 outbreak -we mean steps taken to ght ination or unemploymentwould you say the government was doing a good job, fair job, or a poor job? [Good job, Fair job, Poor job, Don't know] • govt_support_household: Please think about the economic policies initiated by the government to support individuals and households aected by the COVID-19 pandemic.Would you say the government has been doing a good job, fair job, or a poor job? [Poor job, Fair job, Good job , Don't know] • govt_support_rm: Now think about the economic policies initiated by the government to support rms aected by the COVID-19 pandemic.Would you say the government has been doing a good job, fair job, or a poor job? [Poor job, Fair job, Good job , Don't know]

Table 1 :
where Y it is the outcome of interest, that is, households' consumption indicators (consumer sentiment, purchased durables, plans to buy durables), subjective well-being (mental health and life satisfaction), macroeconomic expectations (with respect to ination, unemployment, GDP growth), trust in the government in dealing with the pandemic, and personal concerns due to COVID-19 (health, job security, nancial situation, the general economy); f in_support it is a dummy variable indicating whether household i Marginal Eects of Fiscal Policy on Consumer Sentiment and Durable Spending received nancial support from the government due to COVID-19; X it is a vector of control variables and includes household income per capita, employment status, dummies measuring whether any household members experienced job loss or income losses due to the pandemic, subjective health assessment, as well as various demographics, including urban/rural area, age, age-squared, education, gender, marital status, number of children, and number of old people in the household.βisour coecient of interest.3.1 The Eect of Financial Support on Consumption and Subjective Well-beingTable 1 shows that nancial support has a signicantly positive inuence on consumer sentiment and durable spending.Based on Columns 1 and 2, we compute that receiving nancial support corresponds to a 7% and 16% increase in consumer sentiment compared to the sample averages in Vietnam and Thailand, respectively.These eects amount to a moderate change of about 0.4 standard deviations in the consumer sentiment index in both countries.Columns 3 and 4 show that Vietnamese and Thai beneciaries are 22 pp and 13 pp, respectively, more likely to report that they bought durable goods between May and As the COVID-19-related government programs aim at both stimulating the economy and improving social protection, we study the eect of nancial support on subjective well-being outcomes, such as mental health (feeling calm or nervous) and overall life satisfaction.Table 2 sets out the results.For both countries, we nd that nancial support positively aects mental health and life satisfaction.Vietnamese beneciaries show a 3 (column 3 & 4), and marginal eects for choosing the highest answer category from ordered probit estimations (column 5 & 6).Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table 2 :
Marginal Eects of Government Financial Support on Subjective Well-Being jective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table 3 :
Marginal Eects of Government Financial Support on Macroeconomic Expecta- Note: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table 4 :
Marginal Eects of Government Financial Support on the Assessment of and Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01 Note:(iii) Households' Concerns Due to the Pandemic: The results set out in Table5show that, in both countries, government nancial support signicantly reduces various household concerns due to the pandemic.Vietnamese beneciaries are less likely

Table 5 :
Marginal Eects of Government Financial Support on Household Concerns Due Note: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A1 :
Summary Statistics of Key variables

Table A2 :
Marginal Eects of Government Financial Support on Consumer Sentiment Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report coecients from OLS estimations (column 1 & 2) and marginal eects of probit estimations (column 3 & 4) and marginal eects for choosing the highest answer category from ordered probit estimations (column 5 & 6) based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A4 :
Marginal Eects of Government Financial Support on Macroeconomic Expec-Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A5 :
Marginal Eects of Government Financial Support on Trust in Government in , and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A6 :
Marginal Eects of Govt Financial Support on Concerns Due to COVID-19: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A7 :
Marginal Eects of Government Financial Support on Consumer Sentiment:

Table A8 :
Marginal Eects of Government Financial Support on Subjective Well-Being: Note: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01 status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A10 :
Marginal Eects of Government Financial Support on Trust in Government Note: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A11 :
Marginal Eects of Government Financial Support on Concerns Due to status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A12 :
Marginal Eects of Government Financial Support on Consumer Sentiment Note: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report coecients from OLS estimations (column 1 & 2) and marginal eects of probit estimations (column 3 & 4) and marginal eects for choosing the highest answer category from ordered probit estimations (column 5 & 6) based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01

Table A13 :
Marginal Eects of Government Financial Support on Subjective Well-Being : Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01 Note

Table A14 :
Marginal Eects of Government Financial Support on Macroeconomic Ex-: Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01 Note

Table A15 :
Marginal Eects of Government Financial Support on the Assessment of and : Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.* p < 0.10, * * p < 0.05, * * * p < 0.01 Note

Table A16 :
Marginal Eects of Government Financial Support on Household Concerns : Demographic controls include job loss, income loss, log of household income per capita, employment status, urban/rural area, age, age squared, education, gender, marital status, number of children, number of the old, and subjective health assessment.We report marginal eects for choosing the highest answer category from ordered probit estimations based on population weights.Standard errors are in parentheses.*p < 0.10, * * p < 0.05, * * * p < 0.01• calm: To which extent do the following statement apply to you right now?I am calm and relaxed when I think about the current situation.[Strongly disagree, Moderately disagree, Neither agree nor disagree, Moderately agree, Strongly agree] Note