Page 1
A socio-economic analysis of rice production
systems in the uplands of northern Vietnam
Sushil Pandeya,*, Dang van Minhb
a Social Sciences Division, International Rice Research Institute, P.O. Box 933, Manila, Philippinesb College of Agriculture and Forestry, Thai Nguyen University, Thai Nguyen City, Vietnam
Received 19 August 1997; accepted 23 June 1998
Abstract
Production of upland rice under shifting cultivation system is an important economic activity in the mountainous areas of
northern Vietnam. A comparative study of two districts with differential market access and population pressure was conducted
to highlight the effect of these variables on upland rice systems in northern Vietnam. Farmers in the district with a greater
population pressure have a shorter cropping and a shorter fallowing cycle than in the district with lower population pressure.
Farmers reported a high incidence of food shortage in both districts. Income from livestock and wages are important for food
purchases, especially in the district with a better access to market. Even in these upland districts, access to lowland ®elds is a
critical determinant of food security. # 1998 Elsevier Science B.V. All rights reserved.
Keywords: Upland rice; Slash-and-burn; Food security; Market access; Population pressure; Northern Vietnam
1. Introduction
Upland rice is a major food crop in the humid areas
of the forest zone of Vietnam. In 1991, the area under
upland rice was 0.5 million ha comprising eight
percent of the total area planted to rice in Vietnam.
Upland rice is a major subsistence crop of about 52
ethnic minority groups in Vietnam. Most upland rice
farmers practice shifting cultivation based on slash-
and-burn. The actual area under upland rice is much
larger because a part of the land remains as fallow. The
broader characteristics of the upland production sys-
tems in Vietnam are discussed elsewhere (Hong et al.,
1996; Sam, 1996; Cuc, 1996).
Upland systems in Vietnam as well as in other
countries of south-east Asia are undergoing major
changes with an increase in population pressure and
improved market access (Pandey, 1996). When popu-
lation pressure is low and market access is poor, an
extensive land use strategy is generally adopted. Shift-
ing cultivation based on slash-and-burn and long
fallow (fallow period of more than 20 years) is such
a strategy. Traditional slash-and-burn systems are
considered to be sustainable as the land is fallowed
for a long period after 1 or 2 years of cropping. The
fallow period helps restore the productivity of land.
With the closure of the land frontier in recent years and
continuing increase in population pressure, farmers
are forced to intensify land use by reducing the fallow
period. As a result, crop yields have declined provid-
ing incentives for further encroachment of forested
Agriculture, Ecosystems and Environment 70 (1998) 249±258
*Corresponding author. Tel.: +63-2-845-0563; fax: +63-2-891-
1292; e-mail: [email protected]
0167-8809/98/$ ± see front matter # 1998 Elsevier Science B.V. All rights reserved.
P I I : S 0 1 6 7 - 8 8 0 9 ( 9 8 ) 0 0 1 5 2 - 2
Page 2
areas. A cycle of intensi®ed land use and resource
degradation has thus ensued. Although shifting culti-
vation may be an ef®cient system of land use under
low population density, the practice is unlikely to be
sustainable in the face of increased population pres-
sure. The long-term effect of population pressure on
land use intensity has been well documented
(Boserup, 1965, 1981; Harwood, 1996; Giampietro,
1997).
Market access is another major determinant of land
use pattern. Limited access to market restricts inter-
action between upland and lowland economies in two
ways. Firstly, upland farmers cannot access food from
the lowland because of high costs of transport. This
forces upland farmers to grow subsistence crops even
though they may have a comparative advantage in
producing cash crops. Secondly, high transportation
costs also limit opportunities for enhancing income by
growing cash crops which are in demand in the low-
lands. With improved market access, upland farmers
are able to generate income from the production and
sale of cash crops such as vegetables and fruits which
is then used for purchasing food crops from the low-
land. In both the Philippines and Thailand, the area
under upland rice declined dramatically over the last
three decades as farmers switched to the production of
annual and perennial cash crops to exploit the oppor-
tunities provided by improved market access (Pandey,
1996). The effects of such changes in land use on food
security and environmental degradation depend on the
policy and institutional arrangements that are in place.
The objective in this paper is to document the nature
of the upland rice production system in the mountai-
nous areas of northern Vietnam as well as to assess the
nature of changes occurring in these systems in
response to increased population pressure and
improved access to market. Technological, policy
and institutional interventions to improve the sustain-
ability of upland systems are likely to be successful if
they are based on good understanding of the structure
of the upland economy and the likely pattern of its
change.
2. Study design and sampling
The study of transitions in production systems
ideally requires temporal data over a long period.
As such data are rarely available, a comparative
analysis of production systems subject to different
types and degrees of exogenous in¯uences is often
used to study the likely patterns of changes. The
present study employs this method.
Cao Bang province in northern Vietnam was
selected for studying the upland rice systems in the
northeast mountainous zone. Nguyen Binh and Ngan
Son1 districts of Cao Bang were selected for the study
in consultation with the provincial agricultural of®-
cers. Only those communities within the districts
where the production of upland rice is a major activity
were selected. Farmers in both districts practice `com-
posite swiddening' (Rambo, 1996) involving shifting
cultivation in the uplands and wet rice cultivation in
the lowlands. Although both districts are far from the
provincial town, the market access and the extent of
forest coverage are somewhat different in these dis-
tricts. Ngan Son has a better market access2 as it is
well-connected by a major road to the lowlands of
Thai Nguyen province. The extent of deforestation,
which can be considered as a proxy for land scarcity3,
is also higher in Ngan Son. Thus, Ngan Son could be
considered as the district with a better access to market
and a higher degree of deforestation with Nguyen
Binh having a poorer access to market and a lower
degree of deforestation. The comparative study will,
hence, be based on these two districts representing
different levels of market access and extent of defor-
estation.
1Recently, Ngan Son has been included in the newly-created
province Bac Kan.2The access to market can be variously measured by the physical
distance, the travel time, the cost of transportation, the size of the
market and the marketing margin. Here, the simple measure of the
physical distance from the district headquarter to the main highway
linking the province to Thai Nguyen is used.3The extent of deforestation, defined as the removal of forest
cover, depends on many factors such as logging concessions and
forest management policies in addition to population pressure. As
the two study districts are adjacent to each other, the effects of
logging and forest management policies on the extent of
deforestation are unlikely to be too different. The observed
difference in forest availability could hence be attributed mainly
to the utilization of forest for production of upland crops. As the
techniques for producing upland crops are generally invariant
across farmers in the surveyed districts, the extent of deforestation
could be considered to be a reasonable proxy for population
pressure.
250 S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258
Page 3
The selection of the respondents was done in two
stages. Firstly, from among the `bans' where upland
rice is grown under shifting cultivation, a limited
number with differential degrees of access were
selected for the study4. Altogether ten `bans' from
Nguyen Binh and six `bans' from Ngan Son were
included in the survey. Secondly, households were
randomly selected from the `bans' selected in the ®rst
step. Altogether, 39 and 31 farmers were interviewed
from Nguyen Binh and Ngan Son, respectively.
The interviewer team consisted of four persons with
good knowledge of the farming systems in mountai-
nous areas of northern Vietnam. In addition, two or
three local people were also included to serve as
guides and interpreters for translating local dialect
into Vietnamese. Interview schedules were pre-tested
before the survey began.
3. Results and discussions
3.1. A brief characterization of Cao Bang
The provincial centre of Cao Bang province is about
300 km north of Hanoi. The provincial population of
0.6 million consists of various ethnic minority groups
of which Tay, Nung, Dzao are the main groups (83%).
The population density is 72 person/km2. In terms of
agro-ecosystems, the province can be divided into two
sub-regions, viz., northeast rocky mountain area and
southwest acidic soil area. Corn is mostly grown in
northeast rocky mountains and upland rice is concen-
trated in the ®ve southwest districts of Bao Lac,
Nguyen Binh, Ngan Son, Babe and Thach An. Cao
Bang is one of the typical northern upland rice areas
with approximately 5000 households practising shift-
ing cultivation. Although upland rice area has declined
since 1990, some 3000 ha were grown in 1995
(Fig. 1). The province belongs to the monsoon tropical
climate region. The winter season (November to
March) is cold and mostly dry. The rainy season (April
to October) is hot. The average annual rainfall is
1700 mm.
3.2. Demographic information in surveyed
area
The inhabitants of the surveyed villages are
the ethnic minority Dzao, consisting of the Dzao
Do and the Dzao Tien sub-ethnic groups. The Dzao,
like the Hmong, practice shifting cultivation. As
opposed to the practice of shifting homestead as well
as ®elds which was common several years ago, the
Dzaos now are mostly sedentary and have swidden
®elds which are cropped on a rotational basis (Cuc,
1996).
The average household consists of 7.2 persons, with
43 percent being in the age group of 15 to 50 years
(Table 1). The ratio of adult (age group 15 to 50) to
children (less than 15 years of age) is almost 1:1. The
elderly comprise nearly 12 percent of the sampled
population.
Fig. 1. Area and yield of upland rice in Cao Bang province.
Table 1
Population distribution by age (number per farm household)
Age group Nguyen Binh Ngan Son Both
(Years)Number % Number % Number %
Less than 5 1.6 21 1.0 15 1.3 18
5 to 15 1.7 23 2.2 32 1.9 26
15 to 50 3.1 42 3.0 45 3.1 43
More than 50 1.0 14 0.5 8 0.8 11
Total 7.5 100 6.8 100 7.2 100
Average family size 7.5 6.8 7.2
4The village here is called `ban' in local language. `Ban' is the
smallest administrative unit that belongs to a commune. The
number of `bans' in each commune ranges from 10 to 15. The
number of households in each `ban' varies from 8 to 25.
S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258 251
Page 4
3.3. Land characteristics
Although the upland `bans' were selected for the
study, farmers in these `bans' do have some lowland
®elds in the lower part of the toposequence. Lowlands
comprise almost a quarter of the average holding
per household of 0.79 ha in Nguyen Binh. The pro-
portion of lowland is lower in Ngan Son at 15 percent
of the total area per household of 0.82 ha. The avail-
ability of lowland ®elds among upland farmers
comes mainly from government policy which
favoured the distribution of lowlands to each farm
household, wherever possible, in order to enhance
food security.
The land holding is not only small but is also
fragmented. On the average, each household has three
upland parcels and about one lowland parcel. Some
upland parcels are as small as 0.01 ha (Table 2). Most
upland parcels have very steep slopes. The slope of a
parcel was estimated by asking farmers to compare
various degrees of slope drawn on plain paper with the
slope of their ®elds. To verify the accuracy of this
method, farmer estimates of slope for ten farmers were
compared with the slope estimated by interviewer
using a slope meter. The correspondence between
the two estimates was quite good. Overall, more than
50 percent of the upland ®elds in both districts have
slopes in excess of 308.
3.4. Cropping pattern
In upland ®elds, a range of crops such as upland
rice, cassava, corn and sweet potato are grown
(Table 3). Rice is generally the ®rst crop grown after
slash-and-burn. Farmers prefer to grow corn or cas-
sava after growing upland rice for 1 or 2 years.
Generally, parcels having `good' soil (`smooth' soil
and ®elds not rocky) and located nearer to the house
are considered more suitable for upland rice. Corn is
mostly grown in rocky parcels. In the lowland, only
one crop of rice per year is taken and ®elds are
fallowed. As farmers grow wet rice in lowlands and
dry rice in uplands under shifting cultivation, the
farming practice is called composite swiddening
(Rambo, 1996). Upland rice and corn are the major
crops, occupying over 90 percent of the upland
area.
3.5. Rice production practices
In the case of lowland rice, seedlings are trans-
planted in puddled ®elds. The use of inorganic ferti-
lizers is minimal. However, farmers apply up to 4±5 t
of organic manure per ha5. The yield of lowland rice
for 1995 was 3 t/ha in Nguyen Binh and 2 t/ha in Ngan
Son (Table 4).
The labour input for upland rice production is very
high at around 3000 h/ha (Table 5). The time required
Table 2
Range and average area of parcels by land type
Land
type
Nguyen
Binh
Ngan
Son
Both
Lowland
Number of parcels/household 1.15 0.80 1
Range of number of
parcels/household
0±4 0±2 0±4
Average area of parcel (ha) 0.16 0.15 0.15
Range of area of parcel (ha) 0.03±0.45 0.03±0.40 0.03±0.45
Upland
Number of parcels/household 4.08 3.26 3.71
Range of number of
parcels/household
1±7 2±5 1±7
Average area of parcel (ha) 0.15 0.21 0.17
Range of area of parcel (ha) 0.01±0.55 0.01±0.80 0.01±0.80
Table 3
Percentage area under different upland crops
Crop pattern Nguyen
Binh
Ngan
Son
Both
(%) (%)
Corn-sweet potato/soybean 0.42 0 0.20
Fallow-upland rice 81.30 70.74 76.15
Cassava 4.24 5.35 4.78
Fallow-sweet potato 0.21 0.09 0.16
Fallow-corn 13.62 24.15 18.59
Fallow-taro 0.19 0 0.10
Total 100 100 100
5The figures reported by farmers are on fresh weight basis. It is
not possible to provide precise estimates on dry weight basis as the
moisture content varies widely depending on how the manure is
kept.
252 S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258
Page 5
to walk from the house to the ®eld is included in this
estimate6. The travel time from home to the ®eld of 30
to 60 min is very common (Fig. 2) with the maximum
time reported being 3 h. The practice of using
exchange labour is common for land preparation,
sowing and harvesting. Depending on the size of
the parcel as many as 40 farmers may work on one
®eld before moving to another ®eld.
Labour input for land preparation and weeding
account for two-thirds of the total labour use for
upland rice production. Land preparation is done by
using hoes. Farmers reported that without land pre-
paration, rice yields will be very low because of
hardness of soils and weed problems. Rice seeds
are broadcast and covered by using hoes. This appears
to be normal practice among the Dzao (Cuc, 1996).
Farmers in the study area as do not use dibbling
stick which is common in other slash-and-burn sys-
tems in Laos. This practice may have evolved in
response to increased intensity of land use for crop-
ping which necessitates tillage for breaking com-
pacted top soil as well as for removing weeds
before the crop is planted.
Weeding is done two to three times. Farmers con-
sidered weeding to be one of the most dif®cult tasks
for upland rice production. Reduced fallow period
may have led to increased labour intensity for weed-
ing. This observation is supported by data from Laos
which indicate that the number of weedings increases
very rapidly with a reduction in the fallow period
(Roder et al., 1995). The yield of upland rice in 1995
was around 1.4 t/ha (Table 4).
Most upland rice is harvested by panicle and
bundled. In most cases, bundles are carried to the
house for storage and are threshed only when rice is
needed for food preparation or for sowing. However,
for some varieties prone to shattering, threshing is
done in the ®eld and the grains are carried in bamboo
baskets.
Table 4
Area, yield and production of major food crops
Variable Nguyen
Binh
Ngan
Son
Both
Lowland rice
Area/household (ha) 0.18 0.12 0.15
Yield (kg/ha) 2960 2160 2658
Production (kg/household) 529 268 391
Upland rice
Area/household (ha) 0.49 0.48 0.48
Yield (kg/ha) 1442 1327 1392
Production (kg/household) 710 641 680
Corn
Area/household (ha) 0.08 0.17 0.12
Yield (kg/ha) 1370 1388 1381
Production (kg/household) 113 230 165
Table 5
Seed and labour input use for upland rice
Input Nguyen Binh Ngan Son
Seed (kg/ha) 106 97
Labour (h/ha)
Land preparation 1208 (35) 968 (32)
Sowing 456 (13) 368 (12)
Weeding 1096 (33) 1112 (37)
Harvest 656 (19) 584 (19)
Total labour 3416 (100) 3112 (100)
Note: figures in parentheses are percentages.
Fig. 2. Percentage of parcels by distance from house (in minutes).
6It was not possible to separate the travel time from actual time
worked in the fields as farmers sometimes stayed in the field hut
until a particular operation is completed but at other times they
returned home on a daily basis. Methods other than the memory
recall method used in the survey would be required to obtain the
required resolution. However, the key-informant survey indicated
that the travel time averages around 10±15 percent of the total time
estimated for each activity.
S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258 253
Page 6
De®ning a variety as the crop population that farm-
ers recognize and name as distinct units7, 30 upland
rice varieties and six lowland rice varieties were grown
in 1995 by the farmers surveyed. All upland rice
varieties are traditional but lowland rice varieties
are both improved and traditional. In addition, lowland
varieties are all non-sticky. Although most farmers
keep their own seeds for the next cropping season, the
practice of exchanging seeds among farmers is com-
mon.
Most households grew three or four varieties. Farm-
ers gave several reasons for growing multiple vari-
eties. Firstly, farmers believed that the risk of total
crop failure by drought and insect attack is reduced by
growing varieties with different characteristics. Sec-
ondly, farmers attempt to match the varieties to the
environmental niches. Different varieties are planted
in ®elds with different soil types. For example, `khau
mo' is mostly planted in parcels where soil is of poorer
quality but `khau lum pua' is only planted in good soil
in the ®rst year after slash-and-burn. Thirdly, output
characteristics also differ by varieties. For example,
every farmer grew at least one sticky rice and one non-
sticky rice variety. Non-sticky rice is consumed more
regularly and sticky rice is consumed in smaller
quantities occasionally and is also used for special
products such as rice cakes. Finally, growing rice
varieties with different phenologies helps stagger
labour demand for weeding and harvesting.
3.6. Fallowing
Fig. 3 presents the frequency distribution of parcels
by fallow period. Overall, the average fallow period in
Nguyen Binh is longer than in Ngan Son. Almost 40
percent of the parcels in Nguyen Binh have fallow
periods in excess of 15 years. In the case of Ngan Son,
more than 70 percent of the parcels have fallow
periods of less than 12 years with almost a quarter
having fallow periods of 4±6 years.
Fields are cropped for several years before being
fallowed again. In Ngan Son, almost 50 percent of the
parcels are cropped only once after slash-and-burn. In
Nguyen Binh, the number of cropping years is longer
with almost 60 percent of the parcels being cropped
for 2±3 years. Theoretically, the number of years of
cropping can be expected to depend on the relative
scarcity of land for new openings and the economic
trade-off between the additional labour required for
new opening versus the magnitude of yield reduction
in previously cropped ®elds. If the length of the fallow
period is used as an indicator of the relative scarcity of
land, a longer cropping cycle in Ngan Son could be
expected. However, most ®elds in Ngan Son are
cropped only once or twice. Perhaps the fallow length
in Ngan Son is not enough to build soil quality
suf®ciently to support more than one or two cropping
cycles. Thus land scarcity may lead to shorter fallow
periods as well as shorter cropping periods. The fallow
in Ngan Son appears to be more degraded as 58
percent of the fallow consists of bush and grass as
opposed to only 40 percent of such fallow in Nguyen
Binh. Thus Ngan Son shifting cultivators seem to be
driven by land scarcity towards a path of shortening
fallow cycle and a more rapid decline in land and
labour productivity. Such a pattern of land use is
clearly unsustainable.
Fig. 3. Percentage of parcels by fallow period.
7Clearly, this is an incomplete view of within-crop diversity
because it does not take into account the allelic or genotypic
diversity present in these varieties. Farmers may give the same
name to genetically distinct populations or different names to the
population with the same genetic makeup. Farmers' nomenclature
of rice varieties is based on salient features. For example, `khau
mo' means tolerant to drought, `khau nua' refers to sticky rice,
`khau thep' refers to non-sticky rice, or `khau lum pua' means very
good taste of cooked rice, `soam doang' means seed has awn, `khau
chet' means July rice, `khau pet' means August rice, etc.
254 S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258
Page 7
Is there a correlation between length of the cropping
cycle and the slope of the parcel? A higher rate of land
degradation from erosion in ®elds with greater slope
may reduce the number of years of cropping that can
be supported. In a regression of the number years of
cropping on the ®eld slope, the coef®cient associated
with the ®eld slope was positive and statistically
signi®cant ± thus, supporting this hypothesis. The
absolute value of the coef®cient associated with slope
was higher in Ngan Son than in Nguyen Binh indicat-
ing that for a given slope, ®elds in Ngan Son support a
shorter cropping cycle than in Nguyen Binh. This
again provides support to the earlier hypothesis of a
more degraded fallow in Ngan Son.
3.7. Food shortage
Most of the households in the study area reported
that they were unable to meet the family food require-
ment from their own production. More than 15 percent
of the farmers had food shortage8 in all of the past 10
years. The incidence of food shortage seems to depend
critically on the per capita availability of lowlands.
Households who reported food shortage in more than 8
years out of the last 10 years have half the size of
lowland ®elds in comparison with those who suffered
food shortage in none of the years or in only 1 year
during the past 10 years (Table 6).
The incidence of food shortage in any particular
year depends on three factors. Firstly, production may
not be adequate in years with adverse weather condi-
tions. Secondly, the incidence of food shortage
depends on the family size which determines both
the demand for food as well as the labour supplying
capacity. The demand for food grain depends on the
total family size whereas the labour supplying capa-
city depends on the number of adult family members.
Thus the incidence of food shortage is likely to be
higher among households which have a lower propor-
tion of adults. Thirdly, the area of land operated and
soil quality determine the production potential. Farm-
ers with smaller areas or poorer soils are more likely to
suffer from food shortage.
A regression model was used to determine the
relative importance of these factors in explaining
the variation in the level of food availability per
household which was measured by the total quantity
of crops such as rice, cassava, corn, sweet potato and
other root crops produced. Both the farm size and
family size have positive and statistically signi®cant
coef®cients (Table 7). A higher coef®cient associated
Table 6
Land holding of farmers facing food shortages of selected frequencies in the past 10 years
Number of years of shortage Number of farmers Average size of lowland
holding per capita (m2)
Average size of upland
holding per capita (m2)
9 or 10 years 18 154 934
1 year or none 17 309 1056
Table 7
Factors explaining food consumption
Independent variable Coefficient
Nguyen Binh district
Area of lowland 1422.30a
Area of upland 588.88c
Family size 96.22b
Income 26.98
Ngan Son district
Area of lowland 1627.29c
Area of upland 1347.22a
Family size 159.89a
Income ÿ53.57
Both districts
Area of lowland 1491.25a
Area of upland 1056.76a
Family size 109.88a
Income 23.92
a, b, cDenote statistical significance at the 1%, 5% and 10% levels,
respectively.
Definition of variable: Dependent variable ± food consumption per
household (in kg). Independent variable: area of lowland ± area per
family in ha, area of upland ± area per family in ha, income ± cash
income from all sources in million VND (VND 11 000�US$1).
8Shortage is defined as a situation when consumption require-
ment during the year cannot be met from produce from the farm so
that produce has to be bought or borrowed. Thus food shortage
does not necessarily mean consumption shortfall.
S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258 255
Page 8
with the size of lowland holdings compared with that
of upland holdings re¯ects the higher productivity of
the lowlands in food production. The effect of labour
supply in food production is captured by the family
size. The level of cash income does not have a
statistically signi®cant effect on food availability. In
a subsistence-oriented production system, cash
incomes are generally low and may not greatly affect
the family food supply.
Nearly 68 percent of households reported food
shortage in 1995. About 20 percent of the households
had food shortages for 4±6 months. Food shortages are
felt usually from June to September during which the
food stock from the previous harvest is almost
exhausted. During these lean months, food is either
borrowed from neighbours or purchased using income
generated mainly from labour work and from selling
animals and forest products (Table 8). Purchased rice
is mostly lowland varieties which are brought by local
traders from the lowland and hence are more easily
available than the upland rice. Fig. 4 indicates that the
incidence of food shortage is correlated more closely
with the average size of lowland ®eld per capita than
with the size of the upland ®elds. The quantity of rice
purchased to meet the shortage in production was
found to be positively correlated with the non-farm
income.
3.8. Household incomes
The average cash and non-cash income9 per house-
hold for 1995 was about US$550 (Table 9). At the
average family size of seven persons, this amounts to
about US$78 per capita. More than 60 percent of the
total income is non-cash with the value of rice
accounting for about half the total income. The share
of rice is higher in Nguyen Binh because farmers in
this district have slightly more lowland area. The share
of upland rice in the value of total rice output is 57
percent and 70 percent in Nguyen Binh and Ngan Son,
Table 8
Relative importance of alternative sources of income for buying
food
Source of income for buying food Percentage
response
Selling fruits 1
Selling forest products (wood, fuel wood,
mushroom, bamboo shoot)
20
Labour (gold mine diggers, sawyer wood,
carrying goods)
39
Selling animal (cattle, bufallo, pigs, chicken) 33
Salary 5
Grocery shop in village/wine processing 2
Total 100
Fig. 4. Relationship between the number of reported months of
food shortage in 1995 and the average size of land holding per
capita of the households reporting food shortage.
Table 9
Average annual income by sources (US$/household)
Source of income Nguyen
Binh
Ngan
Son
Both
Cash income
Selling animals 30.21 121.11 70.46
Garden product 2.80 0.44 1.75
Forest product 23.17 19.17 22.56
Wage labour 79.78 65.75 73.56
Selling rice 3.07 2.75 2.92
Sub-total 139.02 209.22 171.27
Non-cash income
Rice for consumption 281.59 206.59 248.18
Other product
(corn, cassava, poultry)
74.73 109.14 89.41
Sub-total 356.32 315.73 337.59
Total income 495.34 524.95 508.86
Note: VND 11 000�US$1.
9Non-cash income is calculated as the imputed value of farm-
produced goods consumed by the household. Market price was
used for imputation.
256 S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258
Page 9
respectively. Thus rice is still a major income-gener-
ating activity in upland systems although its impor-
tance may have declined over time with gains in
lowland rice productivity. Rice is also the major food
produced and consumed by the farm households
(Table 10).
Major sources of cash income are labour work, and
sale of livestock and forest products. Livestock pro-
duction is a major source of cash income in Ngan Son
as the conditions are more favourable for production
and marketing of livestock in this district. Ngan Son
has a large pastoral area used for raising cattle and
buffaloes. In addition, Ngan Son has good access to
the livestock market of Bac Thai province. Substan-
tially higher cash income from livestock has helped
Ngan Son farmers maintain their annual income at a
par with farmers in Nguyen Binh despite the fact that
the value of rice production in Ngan Son is 35 percent
lower.
Income from forest products include selling of fuel
woods, mushrooms, bamboo products and, to a limited
extent, timber. The relative importance of these indi-
vidual forest products depends on the proximity to a
road. Selling of fuel wood is more common among
farmers located closer to a road. Farmers felt that the
importance of these forest products as a source of wild
food as well as cash income has declined over time
because of deforestation.
Labour income is generated by working on a range
of activities such as carrying goods, working in gold
mines and retailing consumer goods. Most of these
activities are done within the locality. Overall, the
contribution of labour income to cash income is more
than one-third and together with income from live-
stock plays a critical role in enhancing the food
security of upland farmers.
3.9. Concluding remarks
The upland rice production systems of northern
Vietnam have low productivity despite a high level
of labour use. Farmers are extremely poor and face
food shortages of varying intensity. Despite the highly
mountainous terrain of the study area, farmers do have
lowland ®elds where the staple rice crop is grown.
Although most of the rice is produced in upland ®elds
under a slash-and-burn system, farmers rely on low-
land ®elds to augment food supply even in these
remote areas. Hence, an increase in the productivity
of lowland rice may enhance the food security of
farmers in upland areas directly. An improvement
in the productivity of lowland rice can also indirectly
bene®t upland farmers because they mostly purchase
lowland rice in de®cit years. Although not every
farmer in the upland may have access to lowland
®elds, the potential role of lowland rice in enhancing
food security should be recognized by agricultural
researchers and policy makers. In fact, an improve-
ment in lowland rice productivity following de-col-
lectivisation of rice production in 1988 has already
helped to improve the food security of upland farmers
and reduce the upland rice area in recent years.
Farmers generate substantial income from forest
products and labour work in upland areas. The con-
tribution of these sources of income is likely to be
higher among the smaller farmers. The extraction of
forest products is often a major source of cash for
women. Rural development policies which limit
access to forests without providing alternative sources
of cash income can, therefore, undermine the welfare
of the poorer and disadvantaged segments of upland
societies.
The comparison of the nature of production systems
of districts with differential access to market and
forest indicates that access to market plays an impor-
tant role in enhancing income and food security of
upland farmers. Farmers in the Ngan Son district,
which is characterized as having better access to
market, were able to maintain their income at par
Table 10
Average production and consumption of different commodities (kg/
household)
Nguyen Binh Ngan Son Both
Production
Rice 1262 899 1101
Corn 108 230 162
Cassava 18 24 22
Other root crops 19 0 11
Consumption
Rice
Self-produced 1244 1111 1185
Purchased 133 163 147
Cassava 204 244 222
Sweet potato 127 222 169
Corn 14 16 15
S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258 257
Page 10
with farmers in Nguyen Binh by relying on livestock
production despite a lower productivity of their rice
®elds. Ngan Son farmers were able to exploit the
existence of large pastoral ®elds for grazing cattle
and buffaloes which are sold in a nearby town. The
pastoral ®elds are normally classi®ed as barren land
implying low economic value of such land. The results
of this study show that such lands can have quite a
large economic value, provided access to markets can
be developed. Policies which seek to limit access to
such land of®cially classi®ed as barren land or to
convert it to other uses can thus adversely affect
farmers' welfare.
In comparison with Nguyen Binh, farmers in Ngan
Son are forced to reduce the fallow length and reduce
the number of crops in each cropping phase because of
a relative scarcity of productive land. A similar pattern
of reduced fallow might emerge in Nguyen Binh as
population pressure increases in the future. If access to
markets could be improved, farmers in Nguyen Binh
could similarly bene®t from interaction with the low-
land economy through livestock production. In upland
areas where farmers have a limited resource base
for food production, diversi®cation of economic
activity based on improved linkages with the lowland
systems can be a successful strategy for enhancing
food security.
Acknowledgements
The authors acknowledge the support of the Thai
Nguyen University in conducting this study. The
assistance of Mr Hoang van Phu, Thai Nguyen Uni-
versity, was instrumental in the initial planning of the
study. The authors also thank anonymous referees and
an Editor-in-chief for improvements to our earlier
version of this paper.
References
Boserup, E., 1965. The Conditions of Agricultural Growth. Allen &
Unwin, London, 121 pp.
Boserup, E., 1981. Population and Technological Change: A Study
of Long-Term Trends. Chicago University Press, Chicago, 255
pp.
Cuc, L.T., 1996. Swidden agriculture in Vietnam. In: Rerkasem,
B. (Ed.), Montane Mainland Southeast Asia in Transition.
Chiang Mai University, Chiang Mai, Thailand, pp. 104±
119.
Giampietro, M., 1997. Socioeconomic pressure, demographic
pressure, environmental loading and technological loading in
agriculture. Agric. Ecosys. Environ. 65, 201±229.
Harwood, R.R., 1996. Development pathways toward sustainable
systems following slash-and-burn. Agric. Ecosys. Environ. 58,
75±86.
Hong, N.H., Quoc, N.G., Xuan, V-T., 1996. Upland rice production
in Vietnam: present situation and prospect for development. In:
Piggin, C., Courtois, B., Schmit, V. (Eds.), Upland Rice
Research in Partnership. IRRI Discussion Paper Series No. 16,
International Rice Research Institute, Philippines, pp. 61±
70.
Pandey, S., 1996. Socio-economic context and priorities for
strategic research on Asian upland rice ecosystems. In:
Piggin, C., Courtois, B., Schmit, V. (Eds.), Upland Rice
Research in Partnership. IRRI Discussion Paper Series No. 16,
International Rice Research Institute, Philippines, pp. 103±
124.
Rambo, T., 1996. The composite swiddening agroecosystem of the
Tay ethnic minority of the northwestern mountains of Vietnam.
In: Rerkasem, B. (Ed.), Montane Mainland Southeast Asia in
Transition. Chiang Mai University, Chiang Mai, Thailand,
pp. 69±89.
Roder, W., Phengchanh, S., Keoboulapha, B., 1995. Relationships
between soil, fallow period, weeds and rice yield in slash-and-
burn systems of Laos. Plant and Soil 176, 27±36.
Sam, D.D., 1996. Shifting agricultural practice today in Vietnam.
In: Rerkasem, B. (Ed.), Montane Mainland Southeast Asia in
Transition. Chiang Mai University, Chiang Mai, Thailand,
pp. 31±50.
258 S. Pandey, D. van Minh / Agriculture, Ecosystems and Environment 70 (1998) 249±258