POVERTY DYNAMICS IN RURAL VIETNAM: WINNERS AND LOSERS DURING REFORM PATRICIA JUSTINO Poverty Research Unit University of Sussex, UK JULIE LITCHFIELD Poverty Research Unit University of Sussex, UK Abstract This paper analyses poverty dynamics in Vietnam during the ‘Doi Moi’ renovation period and tries to identify the winners and losers from the economic and trade reform process implemented in Vietnam in the late 1980s. Our results are based on data available for a panel of 3494 rural households interviewed in 1992-93 and 1997-98. We find that movements in and out of poverty between the two periods vary substantially across population subgroups, suggesting that not everyone benefited equally from the process of reform. We model poverty dynamics using a multinomial logit model that explains movements in and out of poverty between the two periods of time in terms of household characteristics, characteristics directly related to the economic reforms and changes in the returns to those characteristics. The results suggest that changes in household poverty status in Vietnam are correlated with geographic location, access to key institutions and infrastructure, the education level of the head and spouse, as well as changes induced by the economic reform. These results are robust to shifts in the poverty line and changes in model specification. The paper forms part of a wider study funded by the UK Department for International Development that examines the impact of trade reform and trade shocks on household poverty dynamics. JEL codes: C23; I32; O53. Keywords: Poverty dynamics, trade liberalisation, economic reform, panel data, Vietnam. April 2003 Corresponding author: Patricia Justino, Poverty Research Unit, University of Sussex, Afras C, Falmer, Brighton BN1 9SJ, UK. Tel. +44 1273 877402. E-mail: [email protected] Acknowledgements: We are very grateful to Bob Baulch, Rhys Jenkins, Neil McCulloch, Andy McKay, Howard White and, in particular, Alan Winters, for useful discussions and to Yoko Niimi and Puja Vasudeva-Dutta for excellent research assistance. We would also like to thank the participants of seminars and conferences at Sussex (November 2001 and January 2003), Nottingham (April 2002), Hanoi (September 2002) and the 2002 International Association of the Review of Income and Wealth conference (Stockholm) for helpful comments. This paper is part of the project “The Impact of Trade Reforms and Trade Shocks on Household Poverty Dynamics” (ESCOR-R7621) funded by the UK Department for International Development as part of their Globalisation and Poverty Research Programme. Views and opinions expressed in this paper are, however, those of the authors alone. We are grateful to the World Bank for making the trade data available for the DFID-funded Globalisation and Poverty Research Programme’s projects on Vietnam. Poverty Dynamics in Rural Vietnam 1. INTRODUCTION Vietnam has been one of the success stories in the attack on poverty in developing countries. During the 1990s, poverty rates in Vietnam decreased dramatically from 58.1% in 1992-93 to 37.4% in 1997-98 (table 1), whilst more than a quarter of all households that were poor in 1992-93 moved above the official poverty line in 1997-98 (table 2).1 Although poverty amongst urban households, households living in the south of the country,2 households belonging to the ethnic majority Kinh3 and households in which the head had a white collar job decreased more sharply, the reduction in poverty was substantial for most sub-groups of the population (table 1) and was accompanied by large improvements in a range of human development indicators (see Haughton, Haughton and Phong, 2001; Glewwe, Gragnolati and Zaman, 2001 and Litchfield and Justino, 2002). The economic and social changes experienced by Vietnam have been largely ascribed to the “Doi Moi” or renovation policies introduced in 1986.4 These policies aimed at transforming Vietnam into an outward-oriented economy and included the removal of price controls on many goods (including fertilisers), the decollectivisation of land,5 the provision of greater autonomy to the private sector, the reduction or removal of trade barriers and the opening up of the Vietnamese economy to foreign direct investment.6 The largest and most visible economic changes took place in Vietnam’s export sector, particularly in the rice market, where many tariffs were eliminated. Export rice quotas also increased 1 A household is defined as being poor if real consumption expenditure per capita is below an officially defined poverty line. In Vietnam that poverty line is 1160 thousand dongs per year per person in 1992-93 and 1790 thousand dongs per year per person in 1997-98. Those values were calculated based on the cost of a basket of food items that provide the minimum amount of 2100 calories per person per day plus a non-food component. For a detailed discussion of poverty issues in Vietnam and the analysis of the sensitivity of poverty estimates in Vietnam to the choice of poverty line see Litchfield and Justino (2002). 2 For administrative purposes, Vietnam is divided into seven regions: the Northern Uplands, the Red River Delta, the North Central, the Central Coast and the Central Highlands in the north and the South East and the Mekong River Delta in the south. 3 The VLSS specify eight ethnic groups: the Kinh (the Vietnamese majority), the Tay, the Thai, the Chinese, the Moung, the Nung, the H’mong and the Dao. All remaining ethnic groups are aggregated under the heading ‘others’. These groups are usually grouped for convenience into the majority Kinh, the ethnic Chinese (a special group that includes a small minority of quite welloff households) and other ethnic groups that are often located in specific regions only. See van de Walle and Gunewardena (2001) and Baulch et al. (2002) for more detailed analyses of the social and economic position of ethnic groups in Vietnam. 4 Economic reforms, in particular agricultural reforms, started earlier in the late 1970s. However, the ‘Doi Moi’ process that started in 1986 comprised a larger commitment to economic reforms and their extension to the whole economy. 5 Land was owned solely by the state prior to the ‘Doi Moi’ and land transactions were not allowed. After 1993, land tenure was extended to 20 years for annual crop land and 50 years for perennials. Households were also given extended rights to exchange, transfer, lease, inherit and mortgage land (Benjamin and Brandt, 2002). 2 Poverty Dynamics in Rural Vietnam significantly and the private sector was allowed a greater role in the production and export of rice. As a result, rice production and rice exports increased sharply during the 1990s (table 3), turning the country from a net rice importer to the world’s second largest exporter of rice in 1997 after Thailand (in quantity terms).7 Increases in land titling, removal of price controls and other trade reforms have also affected the markets for other agriculture crops (particularly coffee) and industrial products (especially, marine products, textiles and garments, footwear and food processing) (table 3).8 Despite the commitment of the Vietnamese government to the new economic and trade liberalisation programmes, not all reform policies were immediately implemented. Until very recently, the trade regime in Vietnam was still quite restrictive and involved a high level of state intervention (Niimi, Vasudeva-Dutta and Winters, 2002; Nielsen, 2002). Even though the liberalisation programme has not been fully implemented, Vietnam’s gross domestic product per capita grew at an average of almost 7% during the period between 1990 and 1998 (table 3) and real consumption increased significantly in the same period, both in the rural and urban sectors and across all regions (though it was higher in the urban sector and in the southern regions - Mekong River Delta and South East) (table 1). These changes were accompanied by an increase in private sector employment, which led to an overall employment growth rate of almost 2% during the 1990s (table 2). Furthermore, although data on wages in Vietnam is scarce, evidence suggests that real wages started to rise after 1995 (Economist Intelligence Unit, various issues; O’Connor, 1996; World Bank, 1999; CIEM, 2000; IMF, 2000). Increases in per capita incomes and real consumption during the 1990s, combined with a large reduction in poverty, suggest that the economic reforms that took place after 1986 have had a significant impact on the social and economic structures of Vietnamese households. However, although the existence of a link 6 Although flows of foreign direct investment in Vietnam were close to zero during the 1980s, they averaged around 5% of the country’s GDP in the second half of the 1990s (Dollar, 2002). 7 Vietnam’s exports of rice in volume terms made up 17% of total world rice exports in 1997. However, this corresponded to 5% in value terms (which puts Vietnam as the sixth largest exporter of rice in value terms in 1997). The lower value share is due to the fact that Vietnam entered the world rice market at a time when rice prices were falling. This meant that Vietnam was forced to sell its rice at a discount price in order to remain competitive (Nielsen, 2002). 8 For a more detailed account of trade-related changes in Vietnam see Niimi, Vasudeva-Dutta and Winters (2002), also part of the ‘Impact of Trade reforms and Trade Shocks on Household Poverty Dynamics’ project. 3 Poverty Dynamics in Rural Vietnam between economic reforms and poverty reduction in Vietnam is indisputable, it is not clear how the macroeconomic changes that took place in Vietnam during the 1990s were transmitted to each individual household and why some households have benefited more than others from the reforms. As shown in table 1, the dramatic decrease in poverty that occurred in Vietnam between 1992 and 1998 did not benefit all households equally and consumption inequality increased in the country by 12.5% during the period (Litchfield and Justino, 2002). In addition, despite the large decrease in the poverty headcount between 1992-93 and 1997-98 and the large economic achievements, a significant number of Vietnamese households (28.73% overall; 34.19% in the rural sector) remained poor in 1997-98 and some households fell into poverty (4.74% overall; 5.34% in the rural sector) (table 2). These were, as before, households in the remote and mountainous Northern Uplands, ethnic minorities, large households, with a head employed in the agriculture sector and with low levels of education. In order to understand the relationship between economic reforms and poverty, it is important to discern the reasons why this has happened and to determine whether movements into poverty and failures to escape poverty are related in any way (and how) to the economic reforms implemented in Vietnam after 1986. Also we can not overlook the fact that, alongside the economic reforms, households and individuals may have been subject to other changes that result from normal family and community living. In a period of five years, household characteristics such as demographic composition, education attainment and so forth may have changed and households may have been subject to shocks (for instance, illness and weather and other covariant risks).9 This paper analyses the links between household poverty in Vietnam during a period of economic reform by examining how those reforms may have impacted on households’ incomes (and thus poverty) and comparing those effects with the impact in the levels and changes of specific household characteristics that do not have necessarily to be related to economic 9 In previous work (Litchfield and Justino, 2002), we found that households in Vietnam have undergone changes that may be associated with the reduction in poverty observed in the country between 1992-93 and 1997-98: there has been a significant decreased in the percentage of households headed by a younger individual (and thus more prone to be poor), the number of very large households decreased by over four percentage points, the number of households with a large number of young children also decreased significantly (indicating that fewer children were born), the education level of household heads and their spouses improved sharply and the number of household heads employed in sectors other than the agriculture sector increased. Although some of these changes may be associated with the economic reforms that took place in Vietnam after 1986, they reflect, to a 4 Poverty Dynamics in Rural Vietnam reforms.10 The paper aims not only to identify the winners and losers from the economic reforms but also to understand why and how some households have benefited more than others from the new economic policies. We provide thus both a framework to analyse the impact of trade and other economic reforms on household poverty dynamics and a methodology to distinguish these effects from other household characteristics. Together with Niimi, Vasudeva-Dutta and Winters (2002), this is one of the first studies to explore directly the empirical mechanisms of transmission between macroeconomic trade reforms and household poverty dynamics. This is a particularly important topic in view of increased globalisation of the world economy. One of the most visible manifestations of globalisation is the increase in international trade and the implementation of trade liberalisation policies and reductions in trade barriers. These policies are believed to provide opportunities for many people, especially those in developing countries. However some people can be constrained in their ability to share in the opportunities generated by increased international trade and may even suffer from trade reforms due to their vulnerability to trade shocks. We expect the results obtained in this paper to contribute towards the understanding of how national policymakers and international institutions can maximise poor peoples abilities to seize opportunities that economic openness may provide. In order to understand the direct causes and consequences of poverty, it is important to know whether poverty is simply a transitory state experienced by some households at one time or another or whether it is a persistent phenomenon for certain groups. This information is central for the design and targeting of policies aimed at reducing poverty. This paper focuses therefore not so much on changes in poverty rates per se but in movements in and out of poverty (poverty dynamics), thereby taking advantage of the important panel dimension of the Vietnamese household surveys. large extent, intergenerational changes within the household (panel attrition) as well as the impact of non-economic policies such as those promoting ‘one-child families’ and literacy-oriented programmes. 10 According with the 1997-98 commune survey, most communes attribute changes in their harvest patterns between 1992-93 and 1997-98 to weather changes and spreading of pests rather than change in agriculture policies. Only a few communes mention the introduction of new seeds and new techniques as one of the biggest reasons for increases or decreases in harvest between the two years. 5 Poverty Dynamics in Rural Vietnam The results discussed in this paper are based on household consumption data provided in the Vietnam Living Standards Measurement Surveys (VLSS) for 1992-93 and 1997-98. The VLSS data is obtained from nation-wide household surveys conducted in 1992-93 (October 1992 to October 1993) and 1997-98 (December 1997 to December 1998), and contains valuable information on 4800 households and 120 communes surveyed in 1992-93 and 6000 households and 150 communes interviewed in 1997-98.11 These surveys are particularly useful as they allow the construction of a panel of 4303 households interviewed in both years. Given that we only have available two time periods, it will not be possible to analyse the duration of poverty spells or to determine the extent of persistent poverty in Vietnam.12 However, the panel feature of the VLSS data allows us to trace across the five years those households that have remained poor, have moved out of poverty or have fallen into poverty and determine which particular household characteristics are associated with those movements. The analysis developed in this paper will be based on a panel of 3494 rural households.13 14 The paper is organised as follows. Section 2 provides a description of changes in variables directly affected by the economic reforms and how those changes may have impacted on the dynamics of household poverty in Vietnam. We focus on two important reforms: (i) liberalisation of agriculture markets (rice and other crops) and of prices of agriculture crops and inputs (particularly, fertilisers), which has affected household production decisions, and (ii) liberalisation of export markets, following the removal of export quotas and tariffs, which has influenced employment patterns of Vietnamese 11 These surveys were conducted by the Vietnam’s General Statistical Office and the Ministry of Planning and Investment, with financial assistance from the United Nations Development Programme (UNDP) and the Swedish International Development Agency (SIDA) and technical assistance from the World Bank. 12 See Jenkins (1998) and Devicienti (1992) for analyses of poverty dynamics in the UK when several survey years are available. 13 Our analysis focuses only on those households living in the rural sector. Our focus on the rural sector is justified by the fact that 78% of all households in Vietnam in 1997-98 lived in rural areas (80% in 1992-93) and 61% of all Vietnamese households were employed in the agriculture sector in 1997-98 (65% in 1992-93) (table 1). Furthermore, two of the key economic reforms implemented in Vietnam (reform of rice pricing and rice trade and de-collectivisation of the agricultural sector) were directed at the rural sector. 14 Although the two survey years are, individually, representative of the population, the rural panel is not representative of the total rural population (see Haughton, Haughton and Phong, 2001). However, our various calculations have shown that the poverty levels and the changes in the population shares in the household in the rural panel across several socio-economic characteristics do not differ significantly from the results for the whole rural population in each year. Thus, although we must be careful when extrapolating our results to the whole rural sector, we believe that the differences between our results and the total rural population results will not differ significantly. 6 Poverty Dynamics in Rural Vietnam households. In section 3, we investigate the relationship between poverty dynamics, reform-induced variables and other household characteristics (demographic characteristics, employment status, education levels, illness and weather shocks and commune-related variables). We consider not only the levels of those variables but also changes in household characteristics that took place between 1992-93 and 199798 and changes in reform-related variables. This relationship is modelled using a multinomial logit model, which allows us to estimate the probabilities of a household (i) staying poor in both year, (i) escaping poverty, (iii) falling into poverty or (iv) remaining above the poverty line in both years. We provide also a decomposition of the impact of the various household characteristics and changes to those characteristics on the probability of a household escaping or falling into poverty. Section 4 provides a sensitive analysis of the results obtained in the previous section to changes in the poverty line and compares the outcomes of the multinomial logit model with those of a consumption model. Section 5 concludes the paper. 2. ECONOMIC CHANGES AND POVERTY IN VIETNAM Economists know very little about the behaviour of households during economic reforms and their adaptation mechanisms to economic change. Winters (2000) and McCulloch, Winters and Cirera (2001) are two of the first studies to examine this issue by analysing the links between trade liberalisation programmes and household poverty. They explore how the effects of trade reforms trickle down to households via their direct effects on product and labour (and other factors) markets and, indirectly, through changes in labour revenue and public spending on social sectors. The effect of trade liberalisation on household poverty via the product markets operates by means of changes in prices,15 which affect nominal and real incomes of households in their capacity as producers or consumers of the product. Trade liberalisation policies can also affect household poverty via changes in wages and employment occurred in the labour markets and through changes in government transfers. 7 Poverty Dynamics in Rural Vietnam In the case of Vietnam, household incomes and consumption expenditures are likely to have been significantly affected by the large increase in the exports of agricultural crops (particularly, rice and coffee) and industrial products (especially, marine products, textiles and garments, processed food and footwear) that took place in the country during the 1990s (table 3). In this analysis we focus on two main mechanisms of transmission: (i) changes in the prices of agriculture crops and inputs (such as fertilisers), which resulted from the elimination of trade tariffs and increases in export quotas, and (ii) the creation of new employment opportunities in the main export industries. Production of agriculture crops The economic reforms implemented in Vietnam have triggered changes in the prices of commodities for which tariffs were removed and new export quotas were established. Changes in prices, in turn, are likely to have affected nominal and real incomes of households in their capacity as producers or consumers of agriculture crops, if price changes were transmitted to the internal markets.16 When the price of a certain good increases, producers of that good will benefit from the trade reforms, while consumers may lose out. The final effect depends on how each household responds to external economic shocks and is able to accommodate price changes by, for instance, switching to more profitable crops (Winters, 2000; McCulloch, Winters and Cirera, 2001). It is well known that some population groups are likely to benefit more than others from new economic incentives given that some groups are traditionally more vulnerable to shocks than others, both in terms of their exposure and their ability to adjust to them. As demonstrated by Glewwe and Hall (1998) using panel data from Peru, poor households are more vulnerable to macroeconomic shocks, either because they are more exposed (because, for instance, they rely on private and/or public transfers) or because they are constrained in their ability to adjust to new market incentives (due to, for example, lack of credit). Similar effects are observed in Vietnam, where, as we discussed above, some groups seem to have benefited more from the economic reforms than others. 15 Lowering tariff barriers is likely to reduce the price of imported goods, whereas export liberalisation may lead to higher prices for exported goods (Winters, 2000). 8 Poverty Dynamics in Rural Vietnam Table 4 illustrates the amount of production, area cultivated and unit values (obtained from sales) for a selection of agriculture crops cultivated in Vietnam. With the exception of non-rice food crops and livestock,17 the price of all agriculture crops produced in Vietnam increased significantly between 199293 and 1997-98. Table 5 portrays in further detail changes in the prices of rice (the most important crop cultivated in Vietnam) and coffee (the crop with the largest increase in production) between 1993 and 1998. The table shows a large increase in the international price of coffee and in the unit values of coffee. The change in unit values of coffee sold (farm price) is larger than the change in the retail unit values of coffee, which represents a benefit for households that produce coffee. The table shows also an increase in the domestic price of rice and paddy between 1993 and 1998. This increase was similar to the international price of rice and paddy and reflects the impact of trade reforms on the Vietnamese rice sector. The increase in the prices of rice and paddy is reflected at the household level, although the unit values of rice have increased by more than three times the increase in the price of paddy.18 Production decisions are also likely to be affected by changes in the cost of agricultural inputs such as fertilisers,19 which has decreased on average by 23% between 1992-93 and 1997-98 (table 5). The decrease in the cost of this important agricultural input would have benefited farmers and encouraged increases in the production of agriculture crops (table 4). 16 Governments may continue to fix the internal prices of goods which they have officially liberalised internationally (Winters, 2000). See McCulloch, Winters and Cirera (2002) for a theoretical analysis of these transmission mechanisms. 17 This was possibly due to smaller than expected increases in the demand between 1992-93 and 1997-98 for non-rice foods (meat, oils, fish, fruit and vegetables) (Benjamin and Brandt, 2002). 18 This result is similar to what others have found (see, for instance, Edmonds and Pavenik, 2002). We have also calculated the commune prices of rice and paddy since unit values, which are choice variables and not real prices, are typically affected by measurement errors (Deaton, 1997). We found that commune prices of rice had similar increases to unit values but commune paddy prices increased by 33.9% (table 5). It is not clear, however, that commune prices provide the best estimation of prices in the specific case of Vietnam. The commune prices reported in the household surveys refer to the average of three observations made in the commune surveys (which were done separately from the household surveys). According to the basic information on the questionnaires provided by the VLSS team, the observations did not, however, necessarily involve an actual purchase. As the anthropometrists were not local people, it is thus very possible that the prices quoted are not the true prices of the locality (VLSS, 2001: 21). Given this problems we will not focus in very much detail on the effects of unit values changes on poverty dynamics. 19 Fertiliser supply constraints were significantly reduced in Vietnam during the 1990s with the removal of restrictions to import fertilisers. After 1991, central and provincial state-owned enterprises that earned foreign exchange were allowed to import fertilisers directly (Benjamin and Brandt, 2002). 9 Poverty Dynamics in Rural Vietnam Rice is produced in large quantities in Vietnam and is the single crop whose production involves the most number of households. This is not a surprising result as rice represents the main staple for Vietnamese households, comprising 42% of all food expenditure (51% for households below the poverty line) and 75% of all calorific intake of a typical Vietnamese household (Minot and Goletti, 1998). Furthermore, some of the most significant and visible ‘Doi Moi’ reforms took place in the rice market, where many tariffs were eliminated and export quotas increased. The liberalisation of the rice markets, the consequent increases in the price of rice and decreases in the price of fertilisers would have thus created incentives for households to increase rice production. Table 4 shows that the quantity of rice produced in Vietnam increased by almost 35% between 1992-93 and 1997-98. There has also been significant intensification of rice production as the increase in rice production was accompanied by a decrease in the number of households producing rice and the area cultivated with rice.20 This is likely to reflect improvements in technology used in the cultivation of rice.21 In comparison to rice, the cultivation of fruit and vegetables is relatively small in Vietnam because the priority was to cultivate staple food in order to avoid famines and other food shortcomings (Thang, 2002). Fruit and vegetables were mainly cultivated for own consumption and seen as secondary to the cultivation of rice. Conditions for the cultivation of fruit and vegetables have, however, improved during the 1990s due to increased food stability, farmers looking for crops with higher values than rice,22 and benefits from government regulations (in the import of seeds, machinery, fertilisers, etc) (Thang, 2002). As a consequence, the production and area cultivated of non-rice crops increased significantly in Vietnam between 1992-93 and 1997-98, indicating an intensification of agriculture diversification, most likely due to improvements in market infrastructure, agriculture technologies and increased integration of 20 Note that the drastic decrease in the area cultivated with rice between 1992-93 and 1997-98 (93.56%) hides changes in land concentration. While in 1992-93, 96% of all households in the rural panel did not have any rice land, this percentage decreased to 59% in 1997-98. Furthermore, in 1992-93, most rice plots of households that cultivated rice were over 5000 m2 (61.4%), which explains the large average of land cultivated with rice in 1992-93, in comparison with the 1997-98 average. In 1997-98, most plots were between 100 m2 and 5000 m2. If we average the total area cultivated with rice over the whole panel we get a value of 382.39 m2 for 1992-93 and 251.60 m2 for 1997-98, which represents a lower decrease (34.2%) in the average land cultivated with rice. 21 In fact, rice productivity (measured by kilos of rice harvested per square metre of land cultivated with rice) for those households that produce rice increased by 14.3% between 1992-93 and 1997-98. 10 Poverty Dynamics in Rural Vietnam Vietnam in the world markets.23 The largest production increases were those of coffee (almost 200%),24 annual industrial crops and fruit crops. Although rice production grew at over 6% per year, production of annual industrial crops, coffee and fruit crops grew over three times that rate. Furthermore, whilst the area of land cultivated with rice decreased between 1992-93 and 1997-98, the area cultivated with nonrice crops increased on average by over 200% in the same period. Increases in the production of rice, coffee and other agriculture crops are likely to have affected the welfare status of Vietnamese households. Table 6 illustrates movements in and out of poverty for households that produce rice, coffee and other crops. The table shows that more rice producing households are poor in both years than households that do not produce rice. However, less rice producing rural households fell into poverty and more rice producing rural households moved out of poverty in 1997-98. Similar results were obtained for households that produce non-rice crops, particularly coffee.25 This suggests that, although the production of rice has benefited a significant number of Vietnamese households, increased diversification into other crops between the two survey years has contributed positively towards poverty reduction in Vietnam during the same period. Employment effects The new economic reforms implemented in Vietnam in the 1990s have created incentives for the establishment of new production units that, in turn, may have affected traditional employment structures and created new jobs. This effect is bound to be particularly strong in the main export industries (marine products, textiles and garments, processed food and footwear). A larger demand for marine products in 22 The price of rice dropped significantly world-wide since 1998 and even in 1992-93 and 1997-98 the unit values of paddy were lower than those of other crops. Recall also that farm unit values of paddy were significantly lower than retail unit values of rice, which may have led some farmers to swap to other more profitable crops. 23 This result has been obtained in other studies. Benjamin and Brandt (2002), for instance, suggest that households that lived in regions where it was not lucrative to cultivate rice may have decided to cultivate other crops. 24 Vietnam started to produce coffee intensively in the Central Highlands after 1995, following a boom in the world price of coffee (Thang et al., 2001). This resulted in a large increase in the exports of coffee from Vietnam, which grew at an annual average of 58.2% between 1993 and 1998 (table 3). 11 Poverty Dynamics in Rural Vietnam Vietnam during the 1990s has led some households to breed and export shrimps, prawns and other types of seafood as a complement to their normal farm activities. The exports of textiles and garments, footwear and marine products increased also significantly in the same period (table 3), reflecting the priorities assigned to the production of goods in which Vietnam would have a comparative advantage (due to the abundance of cheap labour) and a larger diversification of agricultural incomes.26 In 1997, the value added in the food processing sector was estimated to be about US$ 2.0 billion, which represented about 8.8% of Vietnam’s total gross domestic product and 35.5% of industrial value added (Minot, 1998). Furthermore, the exports of agricultural, fishery and forest products, almost all of which are processed in some way before being exported, account for around US$ 3.2 billion and grew during the 1993-1997 period at an annual rate of 20.7% (table 3). New employment opportunities in those sectors will, however, only benefit the poor if those displaced from traditional sectors are absorbed into the new growing sectors. If employment is concentrated sectorally or geographically, there may be negative consequences for inequality and poverty reduction. Transitional unemployment caused by the emergence of new markets and the disappearance of traditional markets is not a problem for those that have enough assets to smooth consumption through the adjustment period. However, it can be disastrous for the food and income security of poor vulnerable households that generally do not save and are thus unable to adjust to periods of transitional unemployment, however short these may be (Winters, 2000; McCulloch, Winters and Cirera, 2001). These households are also more likely to live in remote areas with difficult access to the new markets and be illiterate and poorly educated (and, consequently, unable to deal with the complexities involved in the establishment of new markets and new employment opportunities and their access rules). 25 The percentage of coffee and other crops producing households that fell into poverty in 1997-98 is significantly lower than that of households that produce rice, whilst the percentage of coffee producing households that moved above the poverty line in 1997-98 is significantly higher. 26 See Niimi, Vasudeva-Dutta and Winters (2002) for more detailed analysis of these changes. 12 Poverty Dynamics in Rural Vietnam Table 7 shows the changes in the number of members of rural households employed in the main exports industries.27 The first observation suggested by table 7 is the fact that only a very small number of rural households has any members working in one of the four export industries. Given the large changes in the exports of seafood, processed food, textiles and footwear, we expected a much higher impact in terms of employment changes. Although there have been increases in the number of households with members employed in the seafood, food processing and footwear industries, those changes have not been substantial. This can perhaps be explained by the fact that rural non-agriculture activities are usually limited to the production of handicrafts and the processing of raw materials for export goods. Consequently, the majority of these rural non-farm enterprises employs one or two workers (usually members of the family) and, as a result, job creation by these industries has not been high (Niimi, Vasudeva-Dutta and Winters, 2002). Despite small, changes in employment in the four main export industries have had positive welfare impacts on rural Vietnamese households. In general, households that had at least one member working in one of the main export industries had a higher probability of having fallen into poverty in 1997-98 than households with no members employed in the export sectors. However, the probability of escaping poverty (30.05%) is higher for households with members employed in the export sector. Of these, households with members working in the seafood industry had the highest probability of having fallen into poverty in 1997-98, whereas households with members working in the footwear industry registered the lowest probability of having fallen below the poverty line and the highest probability of having moved out of poverty in 1997-98. Summary 27 The variables that represent the number of household members working in export industries were constructed using the list of profession codes provided in the household surveys. Seafood includes in 1992-93 ‘fishermen, hunters and related workers’ and ‘aquaculture and fishing’ in 1997-98. Food processing includes ‘food, food stuff and beverage processors’ in 1992-93 and ‘food processing’ in 1997-98. Textiles includes ‘spinners, weavers, knitters, dyers and related workers’ and ‘tailors, dressmakers, sewers, upholsterers and related workers’ in 1992-93 and ‘textile and garment worker’ in 1997-98. Shoemaking includes ‘shoemakers and leather goods makers’ in 1992-93 and ‘leather and shoemaking industry worker’ in 1997-98. 13 Poverty Dynamics in Rural Vietnam The analysis developed in this section shows some tentative links between the economic reforms implemented in Vietnam after 1986 and changes in poverty status of rural Vietnamese households. The results discussed above showed significant welfare benefits for households employed in the four main export industries. The stronger links were, however, those reflected in the position of households as producers of agriculture crops: trade reforms that resulted in the increase of prices of agriculture crops and decrease in fertiliser prices have benefited households that produced rice, coffee and other crops in 1992-93. The results are particularly strong for producers of non-rice crops (particularly coffee). 3. MODELLING POVERTY DYNAMICS IN VIETNAM Poverty dynamics is usually analysed using models that assess the risk of a household or an individual remaining poor for a given period of time (see Jenkins, 1998). However, those models are inadequate to the situation at hand given that we only have available two time periods. In order to model movements in and out of poverty in Vietnam between 1992-93 and 1997-98 we had to resort to discrete outcome models. We model poverty dynamics in Vietnam using a multinomial logit model. Multinomial logit models are used to model processes that involve a single decision among several alternatives that cannot be ordered (for example, choices between modes of travelling, occupational choices, etc).28 Although there is not strictly speaking an element of choice involved in movements in and out of poverty, the process that underlines poverty dynamics between two periods can be divided into four mutually exclusive alternatives: (i) being poor in both periods, (ii) being non-poor in the first period and poor in the second period, (iii) being poor in the first period and non-poor in the second period and (iv) being non-poor in both periods. 28 This is what distinguishes the multinomial logit model from a simple logit model and from ordered logit models. The simple logit model is used to analyse several decisions between two alternatives (0 and 1), whereas ordered logits are used to model choices between several alternatives that can be ranked. 14 Poverty Dynamics in Rural Vietnam The multinomial logit model determines the probability that a household i experiences one of the j outcomes above. This probability is given by: Pr ob(Yi = j ) = e β 'j xi J ∑e , for j = 0, 1, 2, ..., J. (1) β k' xi k =1 where Yi is the outcome experienced by household i, β k are the set of coefficients to be estimated and xi includes aspects specific to the individual household as well as to the choices. The model is, however, unidentified since there is more than one solution for β 1, ..., β J that leads to the same probabilities Y = 0, Y = 1, Y = 1, ..., Y = J (Greene, 2000). In order to identify the model, one of the βs coefficients must be set to zero (the base category), and all other sets are estimated in relation to this benchmark. For convenience we have set β 1 = 0. In this case the probability function above becomes: Pr ob(Yi = j ) = e β 'j xi J 1+ ∑e , for j = 1, 2, ..., J. β k' xi k =1 (2) and Pr ob(Yi = 0) = 1 J 1+ ∑e . (3) β k' xi k =1 In the case of the analysis of poverty in Vietnam we have that J = 4, where Pr ob (Y = 0) is the probability of a household being poor in both years, Pr ob(Y = 1) is the probability of a household being non-poor in 1992-93 and poor in 1997-98, Pr ob(Y = 2) is the probability of a household being poor in 15 Poverty Dynamics in Rural Vietnam 1992-93 and non-poor in 1997-98 and Pr ob(Y = 3) is the probability of a household being poor in both years. The multinomial logit model described above implies that we can compute J log-odds ratios [ ] ln Pij Pi 0 = β j' x i . The log-odds ratios (also called relative risk ratios) can be normalised on any other [ ] probability, which will yield that ln Pij Pi 0 = x 'j ( β j − β k ) (Greene, 2000). For convenience we have decided to calculate ln[P = 2 P = 0] = φ ( Fi , Dit , Lit , I it , Eit , Ait , Sit , Citk ) (4) and ln[P = 1 P = 3] = φ ( Fi , Dit , Lit , I it , Eit , Ait , Sit , Citk ) , (5) which model, respectively, the probability of a household escaping and falling into poverty in terms of fixed household characteristics (region and ethnicity) ( Fi ), household demographic characteristics (household composition) in year t ( Dit ), occupation of the head of the household in year t ( Lit ), illness shock (number of work days lost due to illness) in year t ( I it ),29 education levels of the head of the household and spouse in year t ( Eit ), assets owned by the household (net income assets and remittances)30 in year t ( Ait , ) and institutional and infrastructure characteristics in year t ( C itk ). This variable represents the effects of access to electricity (as a proxy for access to development infrastructure not included in other variables) and commune characteristics.31 In order to control for noise in the data 29 This is a dummy variable that takes the value 1 if the head of the household lost more than 7 days of work due to illness in the month prior to the interview and 0 otherwise. 30 Net income assets represent total savings minus debts held by the household. We have decided to include remittances in our model as these have become an important source of household income in Vietnam during the later 1990s (Dang and Le, 2002). 31 Communes differ in terms of availability of schools, access to roads and transport, access to markets and availability of resources. Belonging to a certain commune is therefore likely to affect the poverty status of each household. Because commune 16 Poverty Dynamics in Rural Vietnam originated from sampling methods, we have decided to include a variable that represents the quarter in which the household was interviewed ( S it ). All these variables refer to household characteristics in 1992-93 and represent initial conditions. The ratios above give us, respectively, the risk of a household falling into poverty in 1997-98 in relation to staying non-poor in both years and the risk of a household moving out of poverty in 1997-98 in relation to staying poor in both years. e β j' xi in equation (3) is thus the relative risk ratio for a unit change in the variable x: a coefficient less than one means that the variables increases the probability of the household to be in the base category, whereas a coefficient of more than one increases the probability of the household being in the alternative state. The percentage of this probability is given by the coefficient minus one, multiplied by one hundred. For example, in table 8, the value of 1.329 for the Red River Delta means that an household living in the Red River Delta in 1992-93 (in relation to those living in the North Central provinces, our base for the dummy variable) has a 32.9% increase in the probability of having escaped poverty in 1997-98 (in relation to being poor in both years). Continuous variables have been normalised and can be interpreted as the percentage effect of a one standard deviation in the variable on the probability of a household escaping or falling into poverty. We have used three alternative multinomial logit models to analyse poverty dynamics in Vietnam between 1992-93 and 1997-98. The results are based on the rural panel of 3494 households. Equations (4) and (5) above represent our base model. This model is similar to that of Glewwe, Gragnolati and Zaman (2001) but differs in the choice of explanatory variables and the way they were constructed. The second model introduces variables directly related to the economic reforms ( Tit ): production of rice (the single most important crop in Vietnam), production of coffee (crop whose production increased the most between 1992-93 and 1997-98), a measure of agriculture diversification (area cultivated with non-rice characteristics are, to a certain extent, time-invariant (particularly in the short period of five years), we can control for commune effects on household poverty dynamics. In order to capture the characteristics of the commune to which the household belongs we have included in the model some institution characteristics of Vietnamese communes. Some of these characteristics reflect 17 Poverty Dynamics in Rural Vietnam crops), whether or not the household has members employed in at least one of the four main export sectors and the number of household members working in the main export industries (seafood, food processing, textiles and footwear). We analysed also other variables that account for technological changes possibly induced by the economic reforms. These are rice productivity (for households that produce rice), the area of land irrigated per capita and the use of fertilisers.32 We examined also the effect of land ownership since the land reforms implemented in Vietnam after 1986 have allowed farmers to own their land. This is a dummy variable that takes the value 1 if the household owns the land it uses. In addition we have included one other external shock variable: whether the commune where the household lives has been affected by a weather disaster (flood, draught, pests, typhoon and other weather-related disasters) between 1993 and 1998 ( Wi ). This variable should capture the influence of factors not related to the economic reforms on the agriculture variables. Models 1 and 2 include only the initial 1992-93 values for the explanatory variables. However, it may be appropriate to look at changes (i.e. returns to the variables), as well as the initial characteristics, in order to assess the impact caused by changes in the country’s technology between the two years into the analysis (see Glewwe and Hall, 1998 and Dercon, 2000). This would allows us to model the relationships between initial characteristics and poverty dynamics and between returns to those characteristics and movements in and out of poverty. The third model analyses the relationship between poverty dynamics and changes in household characteristics, as well as initial conditions, in order to allow for changes in the coefficients between the two years: ln[P = 2 P = 0] = φ ( Fi , Dit , ∆Dit , Lit , ∆Lit , I it , ∆I it ,Wi , Eit , ∆Eit , Ait , ∆Ait , Sit , Citk , ∆Cit , Tit , ∆Tit , ∆p r , ∆p c ) (6) the integration of each commune within the whole Vietnamese economy (access to road, market and post office), whereas other characteristics illustrate the availability of various resources within the commune (schools, food shop and clinic). 32 Rice productivity is defined by kilos of rice harvested per square metre of land. Land irrigation per capita is given by the number of square metres of land irrigation per person in the household. Use of fertilisers was constructed by adding up the 18 Poverty Dynamics in Rural Vietnam and ln[P = 1 P = 3] = φ ( Fi , Dit , ∆Dit , Lit , ∆Lit , I it , ∆I it , Wi , Eit , ∆Eit , Ait , ∆Ait , Sit , Citk , ∆Cit , Tit , ∆Tit , ∆p r , ∆p c ) (7) We added variables that represent change in the number of adults and children in the household between 1992-93 and 1997-98 ( ∆Dit ), changes in the education level of the head of the household and the spouse ( ∆Eit ), changes in net assets ( ∆Ait ), change in the access to electricity ( ∆C it ) and illness shocks ( ∆I it ). We have also included changes in the trade variables ( ∆Tit ): changes in the number of members of the household employed in the four main export industries,33 changes in land irrigated per capita, changes in rice productivity for those households that are rice producers, changes in land ownership, changes in the quantity of fertilisers used, change in the production of rice and coffee, changes in the area cultivated with non-rice crops and change in the unit values of rice ( ∆p r ) and coffee ( ∆p c ) (as discussed in the previous section). Given the significant effects that trade reforms in Vietnam have had on the country’s agricultural sector, we have also included a dummy variable that represents changes in occupation of the head of the household from the agricultural sector to other occupations and from other occupations to the agricultural sector. The inclusion of change variables may raise questions of endogeneity, particularly the inclusion of changes in the variables affected directly by the economic reforms. This is because these variables may be simultaneously a cause and a consequence of changes in poverty status. However, we felt that the information we obtained from the inclusion of these variables outweighed the possible endogeneity problems. In any case, endogeneity is partially controlled for by the simultaneous inclusion of initial values. amount (kilos) of different fertilisers (urea-nitrogen, phosphate, potassium, NPK and other fertilisers) used by the household in the 12 months prior to the interview. 33 Footwear gets dropped from models 2 and 3 because there are not any rural households from the panel working in the footwear industry that are poor. 19 Poverty Dynamics in Rural Vietnam The results for these models are given in table 8. The estimates in table 8 are already corrected for heteroscedasticity34 using White’s adjusted heteroscedasticity-consistent variances.35 The models have also been tested for independence of irrelevant alternatives since this is one of the most important assumptions underlying the multinomial logit model and requires that the various outcomes Y must occur independently of each other. If the independent of irrelevant alternatives assumption fails, the multinomial logit model cannot be applied to the data and another model must be used. We have used the Hausman test for independence of irrelevant alternatives and found that we could not reject the null hypothesis that the multinomial logit model provides the most efficient coefficient estimates. The table below provides a brief overview of the results estimated in table 8 (model 3) that were found to be statistically significant. Variables that increase probability of escaping poverty Variables that increase probability of falling into poverty Household characteristics • Living in the South East and Mekong River Delta • Older household head • Head with white collar job or employed in sales or production • All levels of education of the head • Spouse with technical and university education • Increase in net assets • Household with access to electricity and improvements in the access to electricity in 1997-98 • Living in commune with daily market, food shop and clinic Household characteristics • Belonging to ethnic minorities • Large number of adults and children in the household • Increase in the number of adults and children in the household • Weather shocks • Living in commune with a road, primary school and post office Reform-related variables • Head changed job to non-agriculture sector • Household members employed in seafood industry • Increase in the number of household members employed in all four main export industries • Quantity of fertilisers used • Larger rice productivity for net rice producers and increase in rice productivity • Increase in retail unit values of rice and coffee • Coffee production and increases in coffee production • Cultivation of non-rice crops and increased in the area cultivated with non-rice crops in 1997-98 34 Heteroscedasticity is common in household survey data (Deaton, 1997; Greene, 2000). The results in the statistical tables are not population-weighted. As discussed above, although the two household surveys are nationally representative, the panel is not necessarily representative of the whole population and we have not had access to the sampling information necessary to built adequate weights. 35 20 Poverty Dynamics in Rural Vietnam Variables that decrease probability of escaping poverty Variables that decrease probability of falling into poverty Household characteristics • Belonging to ethnic minorities • Large number of adults and children in the household • Increase in the number of adults and children in the household • Weather shocks • Living in commune with post office Household characteristics • Living in the Central Coast, South East and Mekong River Delta • Older household head • Head with a white collar job or employed in sales or production • All levels of education of the head and spouse (except spouse with lower secondary education) • Changes in the education level of the spouse of the head • Larger net assets and increase in net assets • Increase in remittances • Household with access to electricity and improvements in the access to electricity in 1997-98 • Living in a commune with upper secondary school, daily market, food shop and clinic Trade variables • Head changed job to non-agriculture sector • Household members employed in seafood industry • Increase in the number of household members employed in all four main export industries (except footwear) • Quantity of fertilisers used • Increase in amount of land irrigated per capita • Rice and coffee production and increases in coffee production • Increases in retail unit value of coffee • Cultivation of non-rice crops and increase in the area cultivated with non-rice crops in 1997-98 The results in table 8 are very consistent across the three models. However, the explanatory power of the multinomial logit model increases as further variables are added to the base model. The inclusion of trade variables into the base model explain an extra 20.7% of changes in poverty status of Vietnamese households between 1992-93 and 1997-8, whereas adding change variables improve the explanatory power of model 2 by 27.9% (and the explanatory power of model 1 by 54.3%). The table shows that households that live in the southern regions (South East and Mekong River Delta) have increased probabilities of escaping poverty. As expected, households with older heads,36 36 We believe that this result is associated with the fact that households with young heads are likely to have young children, which poses a financial burden on the household (Litchfield and Justino, 2002). One important cause of higher poverty for households with larger numbers of children is the high costs of education which have soared in Vietnam after the implementation of the new economic reforms (World Bank, 1999). 21 Poverty Dynamics in Rural Vietnam households with a head employed in a white-collar job, in sales and services or in production, households living in communes with better infrastructure and households with better educated heads and spouses have increased probabilities of escaping poverty, in relation to staying poor in both years.37 On the other hand, ethnic minorities and large households were more likely to having fallen into poverty in Vietnam between 1992-93 and 1997-98. Models 2 and 3 show that household poverty dynamics in Vietnam between 1992-93 and 1997-98 was significantly affected by variables associated with the economic and trade reforms that took place in the country during the 1990s. The models show that households with members employed in the seafood industry had increased probabilities of escaping poverty. Furthermore, households in which the number of members employed in all the main export industries increased between 1992-93 and 1997-98, had also increased probabilities of having escaped poverty in the latter year, in relation to remaining poor in both years. This result suggests that the trade and economic reforms have had important employment effects for Vietnamese households, even if only a small percentage of rural households was affected by those changes. These effects have benefited not only those households that in 1992-93 were already involved in the export industries, but in particularly those households that saw an increase in the number of members employed in those industries in 1997-98. The employment effects of the trade reforms were also felt strongly in the agriculture sector. The results in table 8 show that changes in the main occupation of the head of the household from the agriculture sector to any other sector increased the probability of the household escaping poverty by almost 72%. This effect may have been counterbalanced by the technological changes that took place in the Vietnamese agriculture sector between the two survey years, as the use of fertilisers, rice productivity and increases in rice productivity have increased the probability of Vietnamese households escaping 37 Most institution and infrastructure variables seem to have welfare-enhancing effects. Living in communes with access to a road (models 1 and 2), an upper secondary school (model 3 only), a daily market, a food shop and a clinic increase the probability of a household escaping poverty. The only exception is access to a post office, which seems to decrease the probability of households escaping poverty and increase the probability of households falling into poverty. This apparently 22 Poverty Dynamics in Rural Vietnam poverty. Although rice production did not seem to affect the probability of households escaping poverty, it decreased significantly the probability of households falling into poverty in 1997-98. This result seems to have been related to increases in the retail price of rice. Coffee production and increases in coffee production, as well as the cultivation of non-rice crops and increases in the area dedicated to the cultivation of non-rice crops, have increased and quite large effects on the probability of households escaping poverty. This result seems to suggest that, although Vietnamese households have been positively affected by changes in the rice market, they have benefited much more from agriculture diversification, induced by changes in the prices of cash crops (which are higher than rice prices - table 4) and decreases in fertiliser prices. This confirms the results we obtained in our descriptive analysis in the previous section. We have decomposed the poverty dynamics effects of the various household characteristics and traderelated variables by summing equations (6) and (7) for all households i in our rural sample and dividing each term by the sum of all changes (Dercon, 2000). Our analysis focused on the impact of those variables on the probability of a household escaping poverty as very few households in Vietnam fell into poverty in 1997-98. The results for this analysis are illustrated in table 9. The table shows that the largest effects on poverty dynamics in Vietnam between the two survey years were caused by initial level in education of the head and spouse of the head of the household and geographic location. This latter effect is likely to reflect the impact of economic changes in specific regions, which again illustrates the importance of trade reforms in household poverty dynamics in Vietnam between 1992-93 and 1997-98. In fact, both initial trade conditions and reform-induced economic changes contribute significantly towards changes in household poverty status in Vietnam between 1992-93 and 1997-98. Initial values in reform-related agriculture variables and reform-induced agriculture changes explain, respectively 8% and 17.7% of all movements out of poverty. These effects increase, respectively, to 9.1% and 17.8% if employment in the main export industries is included in this category. We have calculated that the largest contributions to the impact of trade-induced reforms on movements out of poverty were increases in perverse effect may be related to particular characteristics of communes that have post offices. Most of these are in the Mekong 23 Poverty Dynamics in Rural Vietnam coffee production and cultivation of non-rice crops. These results suggest that movements out of poverty in Vietnam between 1992-93 and 1997-98 were, to a relatively large extent, determined by the trade and economic reforms that took place in the country during the 1990s. We cannot, however, overlook the impact of education in explaining the reduction of poverty in Vietnam between those periods. Better educated households have a clear advantage over all other households, not only because they will be able to access better paid jobs, but also because they will be in a better position to take advantage of the opportunities that economic openness provides. 4. SENSITIVITY ANALYSIS Although the multinomial logit models analysed above show consistent results across three different model specifications, the model itself may be subject to criticisms. One of those relates to the interpretation of the model. This can be difficult when we estimate discrete variables, which are calculated not only in relation to the benchmark state but also in relation to their base value. In order to overcome this criticism we could estimate simple probit or logit models on the poor and the non-poor rural household samples separately. These models would give us the probabilities of poor rural households moving out of poverty and of non-poor rural households falling into poverty. We have analysed these alternative models but did not obtain any significantly different conclusions from the conclusions provided by the multinomial logit models (Justino and Litchfield, 2002; Litchfield and Justino, 2002). This is because the models are equivalent (except for sample size) for outcomes 2 and 3 of the multinomial logit models, with the advantage that the multinomial logit model estimates simultaneously the various error terms and thus introduces less measurement error into the analysis. Another possible criticism is the underlying assumption of unordered choices forced upon the data by the multinomial logit model. This model may not be the most adequate method for analysing household poverty dynamics between two periods if we believe that there is an implicit ordering of the states ‘poor River Delta, where quite a large percentage of households fell into poverty. 24 Poverty Dynamics in Rural Vietnam in both years’, ‘non-poor in 1992-93 and poor in 1997-98’, ‘poor in 1997-98 and non-poor in 1997-98’ and ‘non-poor in 1997-98’. Although forming a complete ranking of poverty outcomes involves arbitrary judgements, it may be reasonable to assume that each household will have a set of preferences regarding the four states. Niimi, Vasudeva-Dutta and Winters (2002) have tested a few possibilities using the methodology offered by the ordered logit models with not very satisfactory results.38 The multinomial logit model can also be criticised on the grounds that it introduces measurement errors into the analysis due to the fact that it relies on an arbitrarily defined poverty lines. Reducing a continuous variable, such as household expenditure, to a qualitative variable may “throw” information away on the variation in the dependent variable with respect to the variation in explanatory variables (Deaton, 1997). This is a particularly serious problem in analyses that use developing countries data, since large numbers of households may be concentrated around the poverty line. We have thus further tested the results of the multinomial logit models by using different poverty lines. We have considered a poverty line 10% above the official poverty line of 1160 thousand dongs in 1992-93 and 1790 thousand dongs in 1997-98 (model 4), and another 10% below those values (model 5). We have also tested the multinomial logit results to the use of adult equivalent scales, which we decided not to adopt in this paper (model 6).39 We found that models 4, 5 and 6 and model 3 (based on the official poverty lines) differ only very slightly. In general, the magnitude and sign of the various coefficients do not change a lot; only the statistically significance of some coefficients changes. One significant difference is the fact that in model 5, which considers fewer households as poor and thus concentrates on the very poor, the coefficient for the Central Highlands is now statistically significant. This seems to suggest that changes in coffee production had larger impacts on poorer households. Also change in spouse education is now statistically significant. This shows that the impact of this education variable on the reduction of household poverty in Vietnam is larger for the poorest households. 38 This is likely due to the fact that it may not be possible to order the four outcomes. Recall that the Hausman statistics in all multinomial logit models strongly indicated that the four outcomes are not interdependent. 39 We created expenditure per equivalent adult using the WHO equiv scale of 1 for adults and 0.65 for children. We also adjusted the poverty line so it is in per adult terms, rather than per person terms. To do this we multiplied the poverty line by 2800/2100 (which is the calorie needs of an adult over the calorie needs of an average person). 25 Poverty Dynamics in Rural Vietnam Although the multinomial logit models seem not to be very sensitive to the choice of poverty line and show very consistent results across different model specifications and different panel samples, we have also analysed a consumption model for Vietnam. An advantage of this type of models is the fact that the whole distribution is taken into consideration. This decreases the likelihood of information being ‘thrown away’, although it also implies that all conclusions we reach will refer to the whole distribution and not only the poor. We use two consumption models that examine, respectively, changes in household consumption between 1992-93 and 1997-98 using non-adjusted consumption expenditures and consumption expenditures adjusted to adult equivalence scales.40 These models are similar to the model used by Stefan Dercon to analyse the impact of economic reforms on rural household in Ethiopia (Dercon, 2000). The derivation of the model can be found in appendix A. Table 8 shows the results for these models. The variable on the left-hand side is the difference between the logarithmic function of household consumption expenditure in 1992-93 and 1997-98 (which can be approximated to the growth rate of household consumption expenditure). The variables on the right-hand side are the same as in model 3. These models allow us to calculate the different elasticities of consumption expenditure relative to household characteristics that remain constant over time (region, tribe and religion) and household characteristics that have changed over time. As long as we can relate some of these characteristics to the economic and trade reform implemented in Vietnam, the model allows us to determine the effects of those reforms on changes in real household consumption expenditure (Dercon, 2000). The difference between this model and Dercon’s is that we have included not only price changes but also other tradeinduced changes in our variable inputs, which gives us two mechanisms to analyse the effects of trade reforms on changes in household consumption expenditure: (i) changes in trade-related outcomes (employment in export industries and changes in the household production of rice and coffee and cultivation of non-rice crops) and (ii) changes in prices (in this case changes in the price of rice and coffee).41 40 We did not find any significant differences between the two models. 26 Poverty Dynamics in Rural Vietnam Although most variables behave in similar fashion in both consumption and multinomial logit models, the two types of models show different results for quite important variables: ♦ Models 6 and 7 suggest that living in the Central Coast and in the Mekong River Delta has a negative impact on the growth of consumption expenditure between 1992-93 and 1997-98, whereas model 3 shows that living in the Central Coast decreases significantly the probability of a household falling into poverty and living in the Mekong River Delta increases the probability of escaping poverty. ♦ Larger numbers of children are associated with increases in consumption expenditure between 199293 and 1997-98, when the same variable decreases significantly the probability of a household escaping poverty and increases significantly the probability of a household falling into poverty. ♦ The employment of the head in white collar jobs had quite a large positive impact on the probability of a household escaping poverty, though the variable is statistically insignificant in models 6 and 7. The same is true of employment of the head in sales. In fact, the coefficient of this variable is statistically insignificant and negative in the consumption models. ♦ Models 6 and 7 show that employment in the export sectors and having members of the household employed in the seafood industry contribute negatively towards the growth of consumption expenditure. Furthermore, whilst all export industries had a positive and statistically significant impact on the probability of escaping poverty, these coefficients are statistically insignificant in the consumption models. ♦ Some of the education variables (head with primary school education, spouse with upper secondary and technical/university education and change in head and spouse education) have negative coefficients in the consumption models. However, all these variables contributed positively (and most are statistically significant) towards the reduction of poverty in Vietnam between 1992-93 and 1997-98. ♦ Models 6 and 7 suggest that the use of fertilisers, land irrigation, rice production and cultivation of non-rice crops have a negative impact on the growth of consumption expenditure of Vietnamese 41 Dercon considers only changes in a composite price. 27 Poverty Dynamics in Rural Vietnam households between the two survey years. All these variables contributed positively towards the reduction of poverty of many households in the same period, according to the results of model 3. These results suggest that movements out of poverty are not necessarily related to increases in consumption across the whole distribution. We can easily visualise a situation whereby households with certain characteristics that escape poverty concentrate around the poverty line. A positive impact of those characteristics could then be sufficient to move these households above the poverty line. Simultaneously, a large number of households with the same characteristics and above the poverty line could have experienced decreases in their consumption levels between the two years without falling below the poverty line. We have investigated this hypothesis by re-examining the behaviour of the consumption models across different expenditure quintiles.42 The results for those models are provided in table 10.43 The table shows that the negative consumption coefficients of households living in the Central Coast are determined by the coefficients of households in the second and third quintiles. Poor households living in the Central Coast have a positive coefficient. The results for the Mekong River Delta are driven by the households in the first quintile. This is likely due to the high number of landless households in the Mekong River Delta, which have experienced a deterioration of their living standards (Lam, 2001). The quintile regressions also show that the impact of larger number of children is particularly damaging (and statistically significant) for households in the first three quintiles despite the overall positive coefficient. The quintile regressions show further that the negative coefficients of some of the education variables are driven by the upper quintiles. Higher levels of education have a particularly positive impact on the consumption levels of poor households (especially those in the second quintile).44 42 We divided households into five expenditure groups according to their consumption expenditure levels in 1992-93 (using sumdist) and run the consumption regressions on each of these five groups. The independent variable is the different between the logarithmic function of their consumption expenditure in 1992-93 and 1997-98. The explanatory variables are the ones used in models 6 and 7. 43 In order to keep our table appendix more manageable, this table includes only the variables discussed above. The full results are available upon request. 44 This confirms the results we obtained previously in model 5 (the multinomial logit model that uses a poverty line 10% below the official poverty line). 28 Poverty Dynamics in Rural Vietnam Although the employment in white collar and sales jobs increase probabilities of escaping poverty, these affect more households in the higher expenditure brackets rather than the very poor. This seems to suggest that employment in higher paid jobs impacts more significantly on the probability of households around the poverty line escaping (or not falling into) poverty. The same is true with employment in the export industries (particularly employment in the seafood industry), which has a positive impact on the consumption levels of households in the higher expenditure groups. This could be driven by the fact that workers that belong to poorer households will be employed by the worse-paying industries and take jobs that require less skills. Increases between 1992-93 and 1997-98 in the number of household members employed in the food processing, textiles and garments and footwear industries affects positively households in the second and third quintiles. These are households around and immediately below the poverty line. The impact of these variables on the poorest households is not statistically significant but the coefficient is also positive. Thus, although initial employment in the export sectors may not have had significant impacts on the consumption expenditure of poor households, increases in the household employment in those sectors had a positive effect on their consumption levels. The results for the agriculture variables are more disperse. Despite the overall negative coefficients, the use of fertilisers and of land irrigation has benefited households in the lower expenditure brackets. However, initial rice production has impacted negatively (the coefficient is also statistically significant) on the very poor households, whereas it has a large positive impact on the consumption levels of households in the upper income bracket. This may explain why the coefficient is statistically insignificant and below 1 in model 1 (probability of escaping poverty) but above 1 in model 4 (which uses a higher poverty line). On the other hand, cultivation of non-rice crops (in particularly, increases in the cultivated area of non-rice crops) has benefited poorer households, which further confirms our previous conclusion that increases in agriculture diversification in Vietnam had an important and positive impact on the reduction of poverty in the country between 1992-93 and 1997-98. 29 Poverty Dynamics in Rural Vietnam These results obtained in the quintile regressions clarify thus the main different between the overall consumption models and the multinomial logit models. They also provide a bridge between the two models. The conclusions yielded by the quintile regressions emphasise the fact that the multinomial logit model provides a very useful method to analyse poverty dynamics, without introducing significantly large measurement errors into the analysis. The model has in addition the advantage of focusing our attention directly on how the poor – and not the whole distribution – were affected by the economic and social changes that took place in Vietnam in the 1990s. 5. CONCLUSIONS This paper intended to show how, empirically, we can identify the mechanisms of transmission of effects induced by economic reform to household poverty dynamics. Our main objective was to isolate particular characteristics and policies that maximise the ability of poor households to seize potential opportunities brought about by economic liberalisation. The paper developed this analysis for the case of Vietnam, where the economic and trade reforms implemented in the late 1980s have been credited for the dramatic decrease in poverty observed in the country between 1992-93 and 1997-98. This analysis was possible due to the availability of a panel of households that were interviewed in both years. The paper analysed the transmission of economic reform effects to the household level using multinomial logit models that provided estimates for the relative risk of households escaping or falling into poverty. The results were generally consistent across all models estimated and were not sensitive to the choice of different poverty lines or model specification. We focused on two important reforms: (i) liberalisation of agriculture markets (rice and other crops) and of prices of agriculture crops and inputs (particularly, fertilisers), which has affected household production decisions, and (ii) liberalisation of export markets, following the removal of export quotas and tariffs, which has influenced employment patterns of Vietnamese households. The results of our analysis showed that the employment effects were in general positive and had significant welfare-enhancing effects for those households with members 30 Poverty Dynamics in Rural Vietnam employed in the main export sectors (seafood, food processing, textiles and garments and footwear). These effects have benefited not only those households that in 1992-93 were already involved in the export industries, but in particularly those households that increased the number of members employed in those industries in 1997-98. Although these changes would have affected only a relatively small number of rural households, this suggests that there may be scope for further poverty reductions in Vietnam by creating additional employment opportunities in the main export sectors for poor households. The stronger links were, however, those reflected in the position of households as producers of agriculture crops. The results of our analysis suggested that trade reforms that resulted in the increase of prices of agriculture crops and decrease in fertiliser prices have benefited households that produced rice, coffee and other crops in 1992-93. In particular, increases in agriculture diversification between 1992-93 and 1997-98 benefited significantly rural households in Vietnam: poverty decreased more notably amongst these households than amongst households that continued producing mainly rice or decided to increase rice production. This impact was especially strong for the poorest households. In face of these results, we think that the Vietnamese government must seriously re-evaluate some of its economic policies for the rice sector, particularly in terms of their impact on poor households that produce rice. Government policies should focus more on promoting further the diversification of rural incomes by encouraging the establishment of small-scale agro-processing or other farm and non-farm enterprises, thereby reducing the dependency of poor rural households on rice production. Special attention should also be paid to the reinforcement of universal literacy and education, given the large impact of levels of education on the poverty status and consumption levels of poor Vietnamese households. We must not also overlook the fact that the economic reforms have not prevented some households from falling into poverty. Those were households living in remote areas, belonging to minority ethnic groups, 31 Poverty Dynamics in Rural Vietnam with large number of members, low levels of education and whose main earnings derive from the agriculture sector. This is a particularly worrying result as the large bulk of the economic reforms implemented in Vietnam during the 1990s was concentrated in the agriculture sector. Although the economic reforms have yielded large benefits for some, a lot still remains to be done in order to ensure that some of the most vulnerable households are able to cope with the type of socio-economic shocks induced by economic reforms and that clusters of persistent poverty do not start to form amongst these households. This paper is still a long way from providing a complete understanding of the effects of economic reforms on household poverty dynamics. The paper, for instance, does not look at changes in wages or changes in exchange rates that would affect the prices of tradable goods versus non-tradable goods (which are bound to have large effects on movements in and out of poverty of most households). Furthermore, although the economic reforms that took place in Vietnam have been quite extensive, only very recently have important barriers been eliminated. Also after 1998 there was a drastic decrease in the world price of rice and coffee. These economic changes would not have been captured by the 1997-98 survey. The mechanisms we have identified provide nonetheless an important and seminal step towards the understanding of the microeconomic impact of economic reforms and the relationship between these effects and other social and economic events. 32 Poverty Dynamics in Rural Vietnam BIBLIOGRAPHY Baulch, R., Chuyen, T. T. K., Haughton, D. and Haughton, J., 2000. Ethnic Minority Development in Vietnam: A Socioeconomic Perspective. Development Research Group, World Bank, Washington D. C. Benjamin, D. and Brandt, L., 2002. Agriculture and Income Distribution in Rural Vietnam under Economic Reforms: A Tale of Two Regions. William Davidson Working Paper no. 519. The William Davidson Institute, University of Michigan. Burgess, S., Gardiner, K., Jenkins, S. and Propper, C., 2000. Measuring Income Risk. ISER Working Paper 2000-15. Institute for Social and Economic Research, University of Essex. CIEM, 2000. Vietnam’s Economy in 1999. Central Institute for Economic Management, Statistical Publishing House, Hanoi. Dang Nguyen Anh and Le Kim Sa, 2002. Rural-Urban Labor Migration in Vietnam: New Insights from the 1999 Population Census. Vietnam’s Socio-Economic Development, 30: 69-80. Deaton, A., 1997. The Analyis of Household Surveys: A Microeconomic Approach to Development Policy. Johns Hopkins University Press. Dercon, S., 2000. The Impact of Economic Reforms on Households in Rural Ethiopia 1989-1995. Centre for the Study of African Economies, Oxford University. Devicienti, F., 2000. Poverty Persistence in Britain: A Multivariate Analysis Using the BHPS, 19911997. Institute for Social and Economic Research, University of Essex. Dollar, D., 2002. Reform, Growth and Poverty in Vietnam. Development Research Group, World Bank. Economist Intelligence Unit, various issues. Country Profile of Vietnam, Laos PDR and Cambodia. Edmonds, E. and Pavenik, N., 2002. Does Globalisation Increase Child Labour? Evidence from Vietnam. National Bureau of Economic Research, Working Paper no. 8760. Glewwe, P. and Hall, G., 1998. Are some groups more vulnerable to macroeconomic shocks than others? Hypothesis tests based on the panel data from Peru”, Journal of Development Economics 56, 181-206. 33 Poverty Dynamics in Rural Vietnam Glewwe, P., Gragnolati, M. and Zaman, H., 2000. Who Gained from Vietnam's Boom in the 1990s?: An Analysis of Poverty and Inequality Trends. Policy Research Working Paper No. 2275. World Bank. Greene, W. H., 2000. Econometric Analysis (4th edition). New Jersey: Prentice-Hall. Haughton, D., Haughton, J. and Phong, N., 2001. Living Standards During and Economic Boom: The Case of Vietnam. UNDP and General Statistical Office, Hanoi. International Monetary Fund, 2000. Vietnam: Statistical Appendix. Staff Country Report No. 99/56, IMF. Jenkins, S., 1998. Modelling Household Income Dynamics. Institute for Social and Economic Research, University of Essex. Justino, P. and Litchfield, J., 2002. Poverty Dynamics in Rural Vietnam: Winners and Losers During Reform. PRUS Working Paper no. 10. University of Sussex, Brighton. Lam Thi Mai Lan, 2001. Landless Households in the Mekong River Delta (A Case Study in Soc Trang Province). Vietnam’s Socio-Economic Development, 27: 56-66. Litchfield, J. and Justino, P., 2002. Poverty Reduction in Vietnam: What Do the Numbers Tell Us?. PRUS Working Paper no. 8. University of Sussex, Brighton. McCulloch, N., Winters, A. and Cirera, X., 2001. Trade Liberalisation and Poverty: A Handbook. London: CEPR. Minot, N., 1998. Competitiveness of the Food Processing Sector in Vietnam: A Study of Rice, Coffee, Seafood, Fruits and Vegetables Subsectors. Appendix I of the Industrial Competitiveness Review, UNIDO, op. cit. Minot, N and Goletti, F., 2000. Rice Market Liberalisation and Poverty in Viet Nam. Research Report 114, IFPRI. Nielsen, C. P., 2002. Vietnam in the International Rice Market: A Review and Evaluation of Domestic and Foreign Rice Policies. Fødevareøkonomisk Institut, Rapport 132. Niimi, Y., Vasudeva-Dutta, P. and Winters, L. A., 2002. Trade Liberalisation and Poverty Dynamics in Vietnam. Mimeo, University of Sussex, Brighton. 34 Poverty Dynamics in Rural Vietnam O’Connor, D., 1996. Labour Market Aspects of State Enterprise Reform in Viet Nam. OECD Development Centre. Thang, N. et al., 2001. Impact of globalisation on poverty. A case study of coffee in Dak Lak province. Vietnam’s Socio-Economic Development 27, 35-55. Thang, B. T., 2002. Trade of Fruits and Vegetables in Vietnam. Vietnam’s Socio-Economic Development 30. Van de Walle, D. and Gunewardana, D., 2001. Sources of ethnic inequality in Vietnam. Journal of Development Economics 65, 177-207. VLSS, 2001. Vietnam’s Living Standards Survey 1997-98 – Basic Information. Poverty and Human Resources Division, World Bank. White, H. and Masset, E., 2001. Constructing the Poverty Profile; An Illustration of the Importance of Allowing for Household Size and Composition in the Case of Vietnam. Institute of Development Studies, Sussex. Winters, L. A., 2000. Trade Liberalisation and Poverty. PRUS Working Paper no. 7. Poverty Research Unit, University of Sussex. World Bank, 1999. Vietnam Development Report 2000: Attacking Poverty. Joint Report of the Government, Donor and NGO Working Group, Consultative Group Meeting for Vietnam, December 14 - 15. 35 Poverty Dynamics in Rural Vietnam TABLE 1 – CHANGES IN POVERTY AND CONSUMPTION EXPENDITURE MEANS BY SOCIO-ECONOMIC CHARACTERISTICS Category Headcount index Consumption expenditure means 1992-93 1997-98 % change 1914.5 2763.75 44.4 Population share (%) 1992-93 1997-98 100.0 100.0 1992-93 1997-98 % change 58.1 37.4 -35.6 All Vietnam Type of household Urban 24.9 8.6 -65.5 3092.3 4829.4 56.2 19.9 21.8 Rural 66.4 45.4 -31.6 1621.8 2166.4 33.6 80.1 78.2 Region Northern Uplands 78.6 58.6 -25.4 1378.7 1920.0 39.3 15.6 17.9 Red River Delta 62.8 28.7 -54.3 1827.7 2938.0 60.8 21.6 19.6 North Central 74.5 48.1 -35.4 1439.0 2196.6 52.6 12.8 13.8 Central Coast 49.6 35.2 -29.0 2034.7 2641.4 29.8 11.9 10.7 Central Highlands 70.0 52.4 -25.1 1515.7 1941.7 28.1 3.2 3.7 South East 32.7 7.6 -76.8 2844.9 5019.5 76.4 12.6 12.8 Mekong River Delta 47.1 36.9 -21.7 2112.7 2535.9 20.0 22.4 21.5 Ethnic group Vietnamese (Kinh) 55.1 31.6 -42.6 1981.5 2917.5 47.2 84.5 86.1 Chinese 11.8 8.4 -28.8 3542.7 5077.7 43.3 2.5 2.2 Other 86.4 75.2 -13.0 1177.4 1524.3 29.5 13.1 11.7 Household size 44.0 17.0 -61.4 2415.4 3753.2 55.4 12.2 12.1 Small (≤ 3 members) Medium (3-6 members) 55.7 30.8 -44.7 1989.4 2933.3 47.4 36.4 44.1 63.2 49.7 -21.4 1742.5 2318.5 33.1 51.4 43.8 Large (≥ 6 members) Occupation of the head1 White collar 23.6 9.9 -58.1 3043.7 4904.3 61.1 4.6 6.6 Sales 27.7 13.0 -53.1 2908.2 4337.7 49.2 8.1 9.5 Agriculture 69.0 48.2 -30.1 1566.2 2081.3 32.9 64.7 61.0 Production 45.9 25.9 -43.6 2288.5 3281.4 43.4 10.9 11.5 Other/no work 44.4 27.4 -38.3 2355.8 3340.2 41.8 11.7 11.3 Education of the head None 62.6 55.0 -12.1 1702.1 2090.2 22.8 36.4 8.4 Primary 56.7 42.2 -25.6 1946.6 2407.2 23.7 26.1 34.9 Lower secondary 64.0 38.0 -40.6 1788.9 2581.9 44.3 22.6 36.3 Upper secondary 44.5 25.1 -43.6 2448.5 3566.3 45.7 4.3 12.3 Technical and university 39.2 14.2 -63.8 2616.1 4588.2 75.4 10.6 8.1 Source: Authors’ own calculations from Vietnam Living Standards Surveys (VLSS) 1992-93 and 1997-98. Note: Consumption expenditure means are expressed at January 1998 prices (values for 1992-93 have been inflated by a factor of 1.456 suggested by the VLSS). 1. This variable was constructed using the list of profession codes provided in the household surveys. We have divided the occupation variable into white collar jobs (scientists, architects, lawyers, economists, academics, clerical workers, etc), sales and services (retail and wholesale workers, salesmen, hotel managers and workers, hairdressers, etc), agriculture (farmers, forestry workers, fishermen, etc), production (miners, masons, food processing workers, shoemakers, painters, etc) and those not working. The unemployed include those heads looking for a job and those who have no job. 36 Poverty Dynamics in Rural Vietnam TABLE 2 – POVERTY MATRIX FOR 1992-93 AND 1997-98 Category % of rural panel Always poor (%) Nonpoor 92-3 Poor 97-8 (%) Never poor (%) Poor 92-3 Nonpoor 97-8 (%) All Rural All Rural 27.34 29.68 39.19 31.08 92-93 97-98 All Rural All Rural 81.22 79.89 28.73 33.89 4.74 5.35 Vietnam Region Northern Uplands 16.54 16.54 47.15 54.84 4.53 4.84 27.30 25.61 Red River Delta 25.01 25.01 24.90 29.11 3.86 4.00 34.65 39.02 North Central 15.74 15.74 38.11 41.09 4.07 4.55 33.71 33.27 Central Coast 10.70 10.70 25.15 29.95 3.99 4.81 22.95 25.13 Central Highlands 3.29 3.29 41.74 41.74 2.61 2.61 25.22 25.22 South East 8.47 8.47 7.98 10.81 2.10 3.04 24.16 33.11 Mekong River Delta 20.23 20.23 23.89 27.16 8.57 9.76 18.74 20.37 Ethnic group Kinh 84.77 84.60 23.81 27.95 4.73 5.39 29.14 31.90 Chinese 0.37 0.34 5.73 28.00 1.27 0.00 5.73 8.00 Others 14.85 15.05 65.86 67.66 5.30 5.26 18.14 17.61 Household size 22.70 22.35 17.93 22.30 5.58 6.73 27.04 30.05 Small (≤ 3 members) Medium (3-6 members) 40.44 45.22 26.35 31.54 4.54 5.05 28.93 31.81 36.86 32.43 39.37 44.32 4.42 4.83 25.48 26.81 Large (≥ 6 members) Occupation of the head White collar 3.41 4.29 6.34 11.15 1.69 2.60 17.76 27.14 Sales and services 4.26 5.44 8.70 15.04 3.87 5.60 18.65 25.96 Agriculture 75.19 71.84 36.18 37.45 5.35 5.49 30.02 30.06 Production 7.81 8.21 20.33 27.86 3.98 4.46 28.73 33.57 Unemployed 9.33 10.22 21.35 30.31 4.23 6.00 22.58 26.50 Education of head None 39.30 10.85 37.33 41.44 5.35 5.71 23.81 23.97 Primary 24.53 37.72 29.22 33.20 5.88 6.53 26.77 28.83 Lower secondary 24.10 35.89 29.02 33.30 4.48 4.77 33.45 35.35 Upper secondary 3.98 9.62 18.81 25.68 2.54 3.37 23.13 28.42 Technical and university 8.10 5.92 13.00 20.41 2.23 3.27 22.52 30.82 Source: Authors’ own calculations from the Vietnam Living Standards Survey 1992-93 and 1997-97. 37 21.02 36.58 24.10 47.90 30.43 65.76 48.80 14.71 27.57 21.09 40.11 30.43 53.04 42.72 42.32 87.26 10.70 34.76 64.00 9.47 49.45 40.18 30.73 40.91 31.61 24.04 74.21 68.78 28.44 46.96 51.83 59.11 53.39 27.00 34.11 37.19 33.51 38.13 33.05 55.52 62.25 28.88 31.45 26.57 42.53 45.51 Poverty Dynamics in Rural Vietnam TABLE 3 – SELECTED ECONOMIC INDICATORS FOR VIETNAM 1991 1992 1993 1994 1995 1996 1997 1998 Av. ∆ 93-8 8.3 6.8 1.8 30.5 24.6 38.7 58.2 1 20.7 15834 13552 15414 16299 17764 19392 20783 20044 GDP (const, US$ million) 5.8 8.7 8.1 8.8 9.5 9.3 8.2 5.8 GDP growth rates (%) 195 217 225 246 264 279 265 GDPpc (constant, US$) 5.6 6.6 10.0 7.7 6.5 4.2 GDPpc growth rates (%) 30571.6 31262.0 32022.4 32856.9 33666.8 33978.3 34352.7 34800.1 Employment (millions) 2338 2541 3924 5826 8155 11144 11592 11499 Imports (US$ million) 2087 2581 2985 4054 5449 7256 9185 9360 Exports (US$ million) Rice 68 150 103 132 259 329 269 619 Coffee 28 76 113 351 614 375 469 603 Agricultural, forestry 1090 1276 1444 1905 2521 3051 3200 and fisheries Marine product 292 329 415 544 575 607 713 707 14.2 Textiles 28 38 63 104 150 178 220 210 35.5 Garment 131 344 502 657 858 1142 1313 1275 25.4 Footwear exports 7 30 139 300 511 782 1149 1212 109.2 Source: GSO statistics. Export values by commodity from World Bank Mirror Statistics. Employment statistics by MOLISA and supplied by CIEM. Notes: 1. Refers to average change between 1993 and 1997. TABLE 4 – CROP PRODUCTION, AREA CULTIVATED AND UNIT VALUES OF AGRICULTURE CROPS IN VIETNAM, 1992-93 AND 1997-981 No producers 92-3 97-8 Ch 2986 2779 -6.9 2497 2385 -4.5 1258 1186 -5.7 92 101 9.8 926 997 7.7 Production (kgs)10 92-93 97-98 Ch 1405 1894 34.9 1059 1109 4.7 1482 3521 137.5 525 1572 199.3 326 421 28.9 Area cultivated (m2) 92-3 97-8 Ch 9543 615 -93.6 1814 1639 -9.6 1326 1583 19.4 3498 10221 192.2 206 808 293.0 92-3 1.72 5.20 4.44 9.21 3.69 Unit values 97-8 Ch 1.92 11.6 1.57 -69.8 4.44 0.0 14.10 53.1 6.37 72.6 Paddy2 Other food crops3 Annual ind crops4 Coffee Oth per ind crops5 1837 2277 24.0 341 473 38.6 320 430 34.3 1.80 2.20 22.2 Fruit crops6 667 577 -13.5 517 464 -10.1 1121 1701 51.8 Agro-forestry7 3046 2810 -7.8 18 21 15.8 44.4 43.48 -2.1 Livestock8 217 631 190.8 937 1963 109.4 476 5024 955.4 Aquaculture9 Source: Authors’ own calculations from the Vietnam Living Standards Surveys 1992-93 and 1997-98. Notes: 1. Production and unit values represent the mean average for those households that produce the respective crop. Area cultivated is averaged over those households that cultivate the respective crop. Production values and unit values are deflated by regional prices and are expressed in January 1998 prices. 2. Paddy refers to the weighted average of ordinary rice (autumn, winter and spring rice), ordinary rice cultivated in swidden land, glutinous rice and speciality rice. Rice was converted into paddy by a factor of 0.66 suggested by the VLSS. 3. Other food crops include corn/maize, sweet potatoes, cassava, other staple foods, potatoes, cabbage, other leafy greens, tomatoes, water morning glory, fresh legumes, dried legumes, herbs and spices and other vegetables. 4. Annual industrial crops include soy beans, peanuts, sesame seeds, sugar cane, tobacco, cotton, jute, rush, other annual industrial crops and other annual crops. 5. Other perennial industrial crops include tea, rubber, black pepper, coconut, mulberry, cashew and other perennial industrial crops. 6. Fruit crops include citrus, pineapples, bananas, mangoes, apples, grapes, plums, papayas, litchis, sapodilla, custard apples, jackfruit and other fruit trees. 7. Agroforestry crops include mu oil tree, cinnamon, anise, pine trees, varnish trees, trees for wood, bamboo, fan palm tree, water coconut and other silviculture trees. 8. Livestock includes water buffalos, cattle, pigs, horses, goats, chickens, ducks, geese, bees, silk worms, deer, snakes and other animals kept as livestock. 9. Aquaculture includes: fish, shrimps and other water products. 10. Production refers to kilos of the respective crop harvested in the 12 months prior to the interview, except for agro-forestry crops, livestock and aquaculture. Production of agro-forestry crops refers to the value of amount of agro-forestry crops harvested in the 12 months prior to the interview (in thousand of dongs). We could have used number of trees cultivated by the household. However, this variable is evaluated in the household surveys using a mix of area of trees cultivated and number of trees. We were not able to convert area into number of trees. The best variable we could thus used to measure changes in production of agro-forestry crops was the value of that production. Livestock refers to the number of animals owned by the household at the time of the interview. Aquaculture refers to the value received from aquaculture in the 12 months prior to the interview (in thousand of dongs). The household surveys do not provide quantities of aquaculture production. 38 Poverty Dynamics in Rural Vietnam TABLE 5 - CHANGES IN RICE, COFFEE AND FERTILISERS PRICES (RURAL PANEL)1 1992 1993 1994 1995 1996 1997 1998 % ∆ 93-98 International rice price (91=100) 91.47 80.20 91.47 109.56 115.70 103.41 103.75 29.4 International coffee price (91=100) 75.40 83.42 177.01 178.07 143.85 222.99 159.36 91.0 Domestic retail rice price (91=100) 62.99 63.66 69.11 80.82 78.25 73.96 80.82 27.0 Domestic retail paddy price (91=100) 62.14 62.58 66.87 82.30 81.21 73.05 79.64 27.3 1.72 1.92 11.6 Unit values of paddy (farm price)2 Unit values of rice (retail price) 3.06 3.91 27.8 1.74 2.33 33.9 Commune paddy prices3 Commune rice prices 2.79 3.69 32.3 9.21 14.10 53.1 Unit values of coffee (farm price)4 Unit values of coffee (retail price) - 34.78 50.31 44.7 Unit values of fertilisers 2.49 1.92 -22.9 Source: World Bank (1999), Global Economic Prospects and the Developing Countries for international prices and State Pricing Committee for domestic rice and coffee prices. Unit values and commune prices were calculated by the authors from the VLSS 1992-93 and 1997-98. Notes: 1. The table provides the means of household unit values, commune prices and changes in unit values and commune prices. The values are per kilo. All prices are measured in 1998 prices and refer to the average price of one kilo of rice, coffee or fertiliser. 2. Unit values of rice and paddy were derived as the weighted average of the ratio between the total value of ordinary and glutinous rice (bought and sold) and their total quantities. There were some cases where we had observations for total values but not total quantities. We decided that these were recording errors and corrected them by replacing the missing values with the commune mean unit value of ordinary and glutinous rice. Given the importance of rice in the diet of all Vietnamese households, we also replaced the missing unit values of those households who had no value for rice consumption by the commune mean unit value. Rice was converted to paddy in 1997-98 by a factor of 0.66, recommended in VLSS (2001). 3. We assume the commune prices of rice to be the price faced by consumers and the commune prices of paddy (i.e. non-processed rice) to be the prices faced by rice producers. 4. The commune questionnaires do not provide coffee prices and we have not had access to the domestic retail prices of coffee. 39 Poverty Dynamics in Rural Vietnam TABLE 6 - POVERTY MATRIX FOR 1992-93 AND 1997-98: CHARACTERISTICS OF AGRICULTURE PRODUCTION Category % of rural panel Always poor (%) Nonpoor 92-3 Poor 97-8 (%) Poor 92-3 Nonpoor 97-8 (%) All Rural 92-93 97-98 All Rural All Rural Rice Producer 85.46 79.54 35.45 36.05 5.11 5.24 30.31 30.41 Not producer 14.54 20.46 13.65 23.71 3.92 5.89 20.66 26.25 Coffee Producer 2.63 2.89 22.49 23.32 4.31 4.66 34.93 35.75 Not producer 97.37 97.11 28.89 34.19 4.75 5.37 27.15 29.51 Other crops Producer 84.37 83.80 34.18 35.40 5.00 5.14 30.25 30.31 Not producer 15.63 16.20 14.85 25.90 4.08 6.47 19.92 26.35 Source: Authors’ own calculations from the Vietnam Living Standards Survey 1992-93 and 1997-97. Never poor (%) All Rural 29.13 61.76 28.31 44.15 38.28 39.21 36.27 30.93 30.57 61.15 29.15 41.28 TABLE 7 - POVERTY MATRIX FOR 1992-93 AND 1997-98: EMPLOYMENT IN EXPORT INDUSTRIES Poor 92-3 Never poor (%) Nonpoor 97-8 (%) 92-93 97-98 All Rural All Rural All Rural All Rural 8.53 9.19 13.70 18.74 6.14 7.92 25.58 30.05 54.58 43.30 Employed in export sector1 Seafood industry 3.03 3.18 25.71 27.65 10.71 11.52 25.36 28.57 38.21 32.26 Food processing 2.46 3.09 11.89 14.95 4.90 5.15 29.72 30.93 53.50 48.97 Textiles 3.43 2.89 7.48 13.12 4.08 6.79 22.00 28.51 66.44 51.58 Shoes 0.00 0.49 9.68 5.88 3.23 5.88 38.71 58.82 48.39 29.41 Not employed in export sector 91.47 90.81 30.69 35.86 4.56 5.10 27.57 29.64 37.18 29.89 Source: Authors’ own calculations from the Vietnam Living Standards Survey 1992-93 and 1997-97. Note: 1. Refers to households with at least one member employed in one of the four main export sectors. Category % of rural panel Always poor (%) 40 Nonpoor 92-3 Poor 97-8 (%) TABLE 8 - RELATIVE RISK RATIOS FOR ESCAPING AND FALLING INTO POVERTY AND CHANGES IN CONSUMPTION EXPENDITURE PNP Model 1 NPP PNP Model 2 NPP PNP Model 3 NPP PNP Model 4 NPP PNP Model 5 NPP PNP Model 6 NPP Model 7 Model 8 Geographic/ethnic characteristics Northern Uplands Red River Delta (North Central) Central Coast Central Highlands South East Mekong River Delta 1.043 1.329* 1.043 1.023 1.088 1.129 0.919 1.231 1.114 0.978 0.861 1.381 0.885 0.853 1.045 1.703** 1.102 1.016 0.795 1.287 1.024 1.036 1.105 1.580** -0.040 0.043* -0.037 0.046** 1.298 1.861* 5.544*** 2.409*** 0.347*** 0.178*** 0.053*** 0.096*** 1.186 0.751 4.133*** 1.880*** 0.381*** 0.511 0.075*** 0.168*** 1.381 0.472 5.229*** 2.121*** 0.300*** 0.549 0.064*** 0.176*** 1.167 0.710 5.506*** 1.768** 0.364*** 0.458 0.073*** 0.226*** 1.540* 0.974 7.915*** 1.855** 0.271*** 0.356** 0.049*** 0.206*** 1.238 0.672 5.384*** 1.840** 0.325*** 0.542 0.078*** 0.207*** -0.129*** -0.030 0.077** -0.183*** -0.125*** -0.019 0.090*** -0.172*** 0.524 0.411*** 0.359 2.709*** 0.350 0.430*** 0.690 2.594*** 0.404 0.474*** 0.815 2.177*** 1.831 0.484*** 0.686 2.870*** 0.797 0.735* 1.137 1.574** 1.806 0.430*** 0.474 2.130*** -0.072 -0.043* -0.054 -0.046* 0.981 1.190 1.015 1.159 0.945 1.156 1.047 1.014 0.998 1.129 1.033 1.047 0.055*** 0.054*** Head is male (Head is female) 0.969 1.010 0.915 1.191 1.006 1.187 1.021 0.936 0.988 1.226 0.968 1.017 0.001 0.003 Head age in 92-93 Change in head age Head age squared 1.340*** 0.611*** 1.353*** 0.605*** 0.844*** 1.354*** 0.837*** 1.399*** 1.213** 1.082 0.920 0.656*** 0.887 1.258*** 1.172* 0.993 0.872** 0.657*** 0.953 1.242*** 1.231** 1.200* 1.071 0.647*** 0.839 1.146** 1.031 0.856* 0.875** 0.754*** 0.904 1.280*** -0.011 0.006 0.009 -0.014 -0.007 0.019** No adults in 92-93 Change no. adults No children 92-93 Change no. children 1.077 1.081 0.976 1.494*** 0.727*** 2.365*** 0.668*** 2.965*** 0.583*** 0.519*** 0.443*** 0.343*** 2.026*** 1.397*** 4.603*** 2.489*** 0.569*** 0.528*** 0.458*** 0.352*** 2.122*** 1.274*** 4.516*** 2.651*** 0.623*** 0.566*** 0.442*** 0.331*** 1.941*** 1.318*** 4.766*** 2.552*** 0.548*** 0.519*** 0.630*** 0.424*** 2.265*** 1.255** 2.677*** 2.329*** -0.037*** -0.125*** 0.024** -0.160*** -0.050*** -0.134*** 0.043*** -0.109*** 2.054*** 0.441*** 2.004*** 0.462*** 2.047*** 0.433*** 1.897*** 0.508*** 1.827*** 0.399*** 1.961*** 0.426*** -0.033* -0.035* (Kinh) Chinese Other ethnicity Buddhist (Non-Buddhist) Demographic characteristics Human capital characteristics (Head no educ) Primary school Poverty Dynamics in Rural Vietnam Low sec school Up sec school Technical/university Ch head education 2.992*** 3.192*** 2.260*** 0.272*** 0.166*** 0.223*** 2.954*** 3.089*** 2.262*** 0.286*** 0.167*** 0.211*** 3.383*** 3.781*** 2.636*** 1.421*** 0.211*** 0.109*** 0.156*** 0.571*** 3.322*** 3.550*** 2.582*** 1.506*** 0.249*** 0.143*** 0.194*** 0.591*** 4.377*** 3.709*** 2.918*** 1.648*** 0.134*** 0.090*** 0.124*** 0.474*** 3.239*** 3.235*** 2.578*** 1.409*** 0.219*** 0.127*** 0.172*** 0.620*** 0.012 0.038 0.009 -0.010 0.009 0.037 0.005 -0.014 Hh w/ no spouse 1.043 1.002 1.083 1.008 1.083 0.841 1.016 0.769 1.164 0.980 1.085 0.713 -0.002 -0.002 (Spouse no educ) Primary school Low sec school Up sec school Technical/university Ch spouse educ 1.116 1.128 1.308 2.212*** 0.683** 0.934 0.672 0.190*** 1.200 1.180 1.364 2.609*** 0.643** 0.891 0.611 0.145*** 1.279 1.324 1.659 3.270*** 1.157 0.573*** 0.687 0.399** 0.109*** 0.648*** 1.324* 1.184 1.566 3.021*** 1.032 0.485*** 0.630* 0.309*** 0.140*** 0.817 1.360* 1.494* 1.271 2.279** 1.275* 0.614** 0.611** 0.498* 0.137*** 0.638*** 1.366* 1.388* 1.724* 3.045*** 1.145 0.525*** 0.606** 0.364** 0.120*** 0.756* 0.030 -0.009 -0.073 0.005 -0.014 0.032 -0.004 -0.057 0.008 -0.014 3.520*** 1.283 0.151*** 0.386*** 3.751*** 1.569 0.125*** 0.232*** 4.153*** 1.913* 0.115*** 0.215*** 3.929*** 2.212** 0.121*** 0.196*** 5.029*** 1.258 0.092*** 0.235*** 4.138*** 1.826* 0.120*** 0.243*** 0.026 -0.058 0.028 -0.059 1.038 0.670** 0.734 1.115 1.318 0.692* 0.496*** 0.943 1.778** 0.936 0.402*** 0.751 1.717** 1.185 0.430*** 0.617* 1.547 0.837 0.460*** 0.798 1.917*** 1.514* 0.360*** 0.566** 0.020 0.036 0.024 0.041 1.719*** 0.643** 1.431** 0.770 1.418** 0.727* 1.390** 0.751 0.077*** 0.081*** 0.781 1.259 0.955 1.201 0.759 1.547 0.737 1.398 -0.054* -0.061* Labour market characteristics White collar Sales (Agriculture) Production Not working Ch other occu9 978 Ch to agric in 97-8 Empl. export sector (Not empl exp sec) No hh workers in exp industries 92-3 Seafood Food processing Textiles Ch no work exp ind Seafood Food processing Textiles Footwear 1.480 0.495* 1.425 0.509 1.433 0.510 1.067 0.495 1.374 0.590 -0.089* -0.094** 1.083 0.933 1.015 0.823** 0.952 1.018 1.244** 1.025 1.185 0.695*** 0.859 0.874 1.133 1.115 1.030 0.774** 0.811* 0.927 1.370** 1.153 1.135 0.672** 0.787 0.912 1.153 1.044 1.060 0.731*** 0.821* 0.854 -0.019** 0.007 0.009 -0.018** 0.010 0.008 1.125* 1.151* 1.237** 1.112* 0.822*** 0.828*** 0.799** 0.928 1.003 1.229*** 1.182* 1.114*** 0.884** 0.786*** 0.830* 0.957 1.219** 1.150* 1.000 5.311*** 0.792*** 0.827*** 0.967 0.071*** 1.121* 1.167** 1.164* 1.092** 0.837*** 0.842** 0.802** 1.069 -0.009 0.010 0.006 0.006 -0.008 0.010 0.004 0.006 External shocks 42 Poverty Dynamics in Rural Vietnam No. wdays lost 92-3 (No work lost) Head ill in 1997-98 (Head not ill) 1.243 0.895 No weather dis 93-8 1.284 0.804 1.476 0.692 1.210 0.766 1.162 0.947 1.180 0.719 -0.029 -0.027 1.083 0.998 1.001 1.061 1.097 0.823 1.162 0.990 0.005 0.004 0.979*** 1.025*** 0.982** 1.024*** 0.985** 1.035*** 0.969*** 1.023*** 0.977*** 1.024** -0.006*** -0.006*** 1.557*** 0.583*** 1.089 0.914 1.813*** 1.216 1.017 0.988 0.534*** 0.892 0.846 0.860* 1.217 1.165 1.114 1.111 0.755 0.908 0.833 0.800*** 2.327*** 1.260 0.733* 0.967 0.471*** 0.856 1.059 0.883 1.354 1.203 1.099 1.023 0.703 0.930 0.780** 0.814*** -0.017 0.009 -0.028** 0.013* -0.021* 0.009 -0.027** 0.012 1.230* 0.891 1.157 0.458*** 1.404** 1.207* 0.976 1.157 1.166*** 1.018 0.927 1.101 0.511*** 0.897 0.986 1.029 1.319** 1.184 1.307 1.037 1.161*** 1.020 1.068 1.143 0.401*** 1.000 1.014 1.057 1.444** 1.597*** 0.995 1.076 1.159** 1.011 0.732* 0.855 0.415*** 0.985 0.948 1.065 1.288* 1.205* 1.221 1.072 1.099** 1.029 1.057 1.141 0.458*** 0.949 1.049 1.076 0.006 0.020 -0.032** 0.017 0.032*** 0.009 0.005 0.019 -0.029* 0.018 0.032*** 0.007 5.068*** 0.176*** 2.365* 0.092*** 3.305*** 2.104*** 1.326** 0.842 6.754*** 4.236*** 0.233*** 0.514*** 0.611*** 1.790 0.039*** 0.243*** 2.860*** 2.100*** 1.345*** 0.810 4.356*** 3.297*** 0.269*** 0.503*** 0.622*** 1.946* 0.048*** 0.312*** 2.414*** 1.894*** 1.302* 1.389 3.879** 2.546** 0.322*** 0.583** 0.580*** 0.985 0.019*** 0.181*** 3.264*** 2.054*** 1.328*** 0.802 6.819*** 4.328*** 0.248*** 0.517*** 0.605*** 2.138* 0.037*** 0.238*** 0.003 0.025*** 0.000 0.102** -0.214*** 0.004*** 0.002 0.026*** 0.001 0.098** -0.222*** 0.004*** 1.112 0.917 1.101 1.180 1.802** 1.080 1.004 1.431 1.535 1.252 0.074** 0.080** 1.735 1.354 2.422* 1.328 2.176 0.895 1.693 1.959 0.057 0.075 Agriculture variables Quant fert 1992-93 Ch in quant fert Land irrig pc 92-93 Ch land irrig pc Rice Productivity 92-3 Ch productivity Quant prod 92-93 Ch quant prod Ch retail uv Ch farm uv Other crops Quant prod cf 92-3 Ch coffee prod Ch retail uv coffee Ch farm uv coffee Cult non-rice 92-3 Ch area cultivated Assets Land in 1992-93 (No land) Ch in land rights Net assets in 92-93 Change in net assets 0.728 0.292** 0.748 0.251*** 0.624 1.887*** 0.246** 0.528** 0.502 2.243*** 0.206** 0.463*** 0.815 1.409 0.252** 0.530** 0.885 2.227*** 0.105*** 0.524** 0.030 0.054** 0.028 0.055** Hh rec remit 92-3 1.083 0.872 1.149 0.781* 1.061 0.859 1.043 0.769* 0.946 0.906 1.127 0.769* 0.014 0.014 43 Poverty Dynamics in Rural Vietnam (Not receive rem) Increase remittances (No increase) 1.229 0.744* 1.159 0.830 1.315* 0.750* 1.243* 0.870 0.060*** 0.060*** 1.907*** 0.181*** 2.266*** 0.152*** 2.209*** 0.213*** 2.365*** 0.146*** -0.019 -0.020 1.634*** 0.446*** 1.635*** 0.457*** 1.800*** 0.444*** 1.711*** 0.454*** 0.011 0.015 Infrastructure and institutions Hh w/ elect 92-3 (No electricity) Ch electricity access Commune vars 92-3 Road Primary school Low sec school Up sec school Post office Daily market Food shop Clinic 1.510*** 0.309*** 1.373*** 0.325*** 1.644*** 0.933 0.874 1.348 0.555*** 1.276* 1.456*** 2.092*** 1.466** 6.228*** 0.879 0.674* 2.584*** 0.640*** 0.460*** 0.517*** 1.540** 0.876 0.863 1.559** 0.613*** 1.498*** 1.442*** 2.076*** 1.505** 7.952*** 0.987 0.560*** 2.419*** 0.531*** 0.451*** 0.420*** 1.190 0.662 0.871 1.755*** 0.589*** 1.547*** 1.413** 1.928*** 2.289*** 8.376*** 0.969 0.535*** 2.607*** 0.458*** 0.496*** 0.484*** 1.108 0.926 1.026 1.778*** 0.535*** 1.663*** 1.343** 1.385 2.302*** 6.338*** 0.881 0.536*** 2.245*** 0.453*** 0.576*** 0.690 0.880 0.432* 0.815 1.788** 0.732* 1.394** 1.300 1.981*** 2.712*** 9.978*** 1.225 0.473*** 2.090*** 0.500*** 0.513*** 0.450*** 1.026 0.575 1.202 1.748*** 0.582*** 1.494*** 1.342** 1.789** 2.758*** 12.281*** 0.778 0.543*** 2.367*** 0.433*** 0.567*** 0.578* 0.149*** 0.291*** -0.081*** 0.011 -0.023 0.011 -0.018 0.007 0.151*** 0.284*** -0.082*** 0.012 -0.021 0.010 -0.021 0.005 (Interv 1stq 92-3) 2nd quarter 3rd quarter 4th quarter 0.950 1.015 0.892 1.470** 0.820 1.026 0.919 1.269 0.902 1.502** 0.647*** 0.986 0.911 1.287 0.800 1.364 0.613*** 0.942 0.980 1.262 0.726** 1.306 0.646** 0.906 0.973 1.441** 0.980 1.576** 0.625** 1.065 0.830 1.107 0.751** 1.394 0.563*** 0.898 0.058** -0.063*** -0.050** 0.058** -0.066*** -0.052*** Constant 0.174*** 1.283 0.227*** 1.455 0.087*** 1.448 0.035*** 2.038 0.198** 0.184 0.043*** 3.504 -0.067 -0.086 Seasonality Observations 3494 3494 3494 3494 3494 3494 3494 3494 3494 3494 3494 3494 3494 3494 Adjusted R-squared 0.184 0.184 0.222 0.222 0.284 0.284 0.288 0.288 0.278 0.278 0.276 0.276 0.252 0.228 Hausman statistic -22.73 -22.73 -6.86 -6.86 -33.14 -33.14 -2685.02 -2685.02 -125470 -125470 -112.87 -112.87 Notes: ***significant at 1% level, **significant at 5% level, *significant at 10% level. Model 1: base model. Model 2: includes variables affected by the trade reforms. Model 3: includes ‘change’ variables. Model 4: uses poverty line 10% above the official poverty line. Model 5: uses poverty line 10% below the official poverty line. Model 6: uses adult-equivalent adjusted poverty line. Model 7: consumption model using non-adjusted consumption expenditure estimates. Model 8: consumption model using consumption expenditure estimates adjusted to adult equivalent scales. 44 Poverty Dynamics in Rural Vietnam TABLE 9 – DECOMPOSITION OF POVERTY DYNAMICS EFFECTS (%) Variables Location Ethnicity and religion Initial demographic characteristics Initial education levels (head and spouse) Initial employment characteristics Number of workdays lost due to illness Weather shocks Initial agriculture conditions Initial infrastructure Initial assets Seasonality effects Probability of escaping poverty 27.4 -4.6 -8.0 33.6 4.7 0.5 -1.7 8.0 11.7 0.9 -0.6 Changes in demographic characteristics Educational changes Changes in employment characteristics Illness shock Agriculture changes Changes in infrastructure Change in assets Total Source: Own calculations based on estimates from model 3. 45 3.0 3.2 0.8 0.1 17.7 2.4 0.9 100.0 Poverty Dynamics in Rural Vietnam TABLE 10 – CHANGES IN CONSUMPTION PER EXPENDITURE QUINTILE (SELECTED VARIABLES) Model 7 Quint 1 Geographic characteristics Central Coast Mekong River Delta Quint 2 Quint 3 -0.129*** -0.183*** 0.022 -0.152** -0.090* 0.088 -0.097* -0.017 -0.033 -0.074 -0.042 -0.108 Demographic characteristics Head age in 1992-93 Number children 1992-93 -0.011 0.024** 0.025 -0.033* 0.041** -0.100*** -0.007 -0.053*** 0.014 -0.013 0.048 -0.048 Human capital characteristics Head with prim school educ Ch head education Spouse with upper sec educ Spouse with tech/univ educ Ch spouse education -0.033* -0.010 -0.009 -0.073 -0.014 0.004 0.006 0.025 0.025 0.001 0.077** 0.067** 0.097** 0.118 0.055** 0.041 0.001 -0.016 -0.012 -0.018 0.026 0.030 -0.042 -0.038 -0.012 -0.023 0.022 0.206** -0.046 0.067 Labour market characteristics White collar in 1992-93 Sales in 1992-93 Employed in export sector Workers in seafood ind Workers in food proc ind Workers in textiles ind Ch workers seafood Ch workers food proc Ch workers textiles Ch workers footwear 0.026 -0.058 -0.089* -0.019** 0.007 0.009 -0.009 0.010 0.006 0.006 0.036 -0.157 -0.257** 0.028 0.033 0.085** 0.001 0.015 0.014 0.005 0.164** 0.048 0.017 -0.000 0.037 0.015 -0.011 0.032*** 0.027 0.026*** 0.123* 0.056 0.015 0.000 0.026 0.011 0.021 0.036*** 0.054** 0.002 0.096 0.121** -0.137* -0.024 0.029 0.030 -0.024** -0.013 -0.006 -0.013 0.227*** 0.015 -0.045 0.062*** -0.006 0.012 0.016 0.009 -0.001 0.000 Agriculture production Quantity fertilisers 1992-93 Ch use of fertilisers Land irrigaped pc 1992-93 Ch land irrigated per capita Quantity rice produced Ch rice production Area cultiv w/ non-rice 92-3 Ch area cult w/ non-rice crops -0.017 0.009 -0.028** 0.013* -0.032** 0.017 -0.214*** 0.004*** 0.240*** 0.091** 0.017 0.009 -0.260*** -0.033 0.018 0.179** 0.093*** 0.045** 0.005 -0.014 -0.008 -0.008 0.005 0.145** 0.005 0.023 0.001 0.006 0.001 0.036 -0.077 0.003** -0.008 0.000 -0.030 0.022* 0.012 -0.002 0.206** 0.076** -0.038 0.010 0.002 0.026** 0.047** 0.021 -0.042 0.103 681 0.49 444 0.44 Observations 3494 813 801 755 R-squared 0.25 0.35 0.42 0.39 Notes: ***significant at 1% level, **significant at 5% level, *significant at 10% level. 46 Quint 4 Quint 5 Poverty Dynamics in Rural Vietnam APPENDIX A – CONSUMPTION MODEL The model assumes risk neutral households that maximise their expected profits and is based on a household’s joint income generation process described as q=f(x,k,u), in which q is total output, x is a vector of n variable inputs, k is a vector of j fixed inputs and u is a vector of m stochastic factors shared by all households (climatic conditions, commune characteristics, etc). x represents effectively the fixed effects in the model, whereas k represents the effects of variables that change between the years of the analysis. The household’s maximisation problem is then written as max E (Y ) = p.g ( x, k , u ) − p x .x , where p is the x,q output price and px represent the input prices. Assuming that the production function takes a Cobb-Douglas form q = a.x α k β u γ , we have that Y =a 1 1−α .α 1 1−α .(1 − α ).k β 1−α .p 1 1−α .px −α 1−α .u γ 1−α (1) Taking logs, aggregating (1) over all households, using subscripts I to denote different households and introducing a household-specific effect δ i for each household I, we have that (Dercon, 2000: 7): ln Yi = δ i + ln ai* + ϕ * ln p i + ∑ α n* ln p inx + ∑ β *j ln k ij + ∑ γ m* ln u im n j (2) m Allowing for these possible changes, considering periods t and t+1 as respectively before and after the reforms, and denoting ∆ as the difference in values between t+1 and t, differences in profits over time are: ∆ ln Yi = ∆ ln a i* + ϕ t*+1∆ ln pi + ∆ϕ * ln pit + ∑ α nt* +1 ∆ ln p inx + ∑ ∆α n* ln pintx + ∑ β *jt +1 ∆ ln k ij n + ∑ ∆β ln k ijt + ∑ γ * j j m * mt +1 n ∆ ln u im + ∑ ∆γ ln u imt * m m 47 j (3) Poverty Dynamics in Rural Vietnam The formulation above allows us to calculate the different elasticities of consumption expenditure relative to household characteristics that that remain constant over time (region, tribe and religion) and household characteristics that have changed over time. As long as we can relate some of these characteristics to the economic and trade reform implemented in Vietnam, then the model allows us to determine the effects of those reforms on changes in real household consumption expenditure (Dercon, 2000). The theoretical model (3) can be easily transformed into an econometric model: ∆ ln Yi = η + ϕt*+1∆ ln pir + ∆ϕ * ln pitr + β t1+1∆Di + ∆β 2 ln Dit + βt1+1∆Ei + ∆β 2 ln Eit + β t1+1∆Li + ∆β 2 ln Lit + β t1+1∆Ti + ∆β 2 ln Tit + βt1+1∆Ai + ∆β 2 ln Ait + β t1+1∆I i + ∆β 2 ln I it + ∆ ln Wi + ∑ ∆ k ln Citk + eit +1 k where: η = fixed characteristics (including interview quarter) p r = rice and coffee prices (producer and consumer) Dit = demographic characteristics (household size and household composition) in year t Eit = education characteristics in year t Lit = labour market characteristics in year t Tit = trade-related variables (employment in export industries, production of rice and coffee and measure of diversification) in year t Ait = household assets in year t C itk = institutional and infrastructure characteristics in year t I it = illness shock Wi = weather shock e = error term 48
© Copyright 2026 Paperzz