Farming Without a Livelihood: Structural Differentiation and Household Income Among Paddy Farmers in West Java, Indonesia

Abstract

Smallholder farming across the Global South increasingly fails to provide a full livelihood. West Java, Indonesia, shows the paradox sharply since paddy farming there is financially viable yet insufficient to live on for most who practise it. Using a survey of 1037 households in four regencies and a structure and agency framework, this study asks how inherited conditions of land and water separate farm performance from household livelihood. Paddy farming returns more than it costs in every size class, with a revenue-to-cost ratio rising from 1.65 to 1.93, while cost per hectare does not differ significantly by farm size. Profitability nonetheless fails to become sufficiency. Annual income covers 79% of household expenditure on small holdings against 153% on large ones, and 61% of small and 48% of medium farms fall short of their own needs. The disadvantage concentrates below 0.3 hectares, where returns approach break-even. Farmer-group membership and access to technical knowledge do not vary with farm size, so what differs is not whether farmers act but what their actions can reach. When farming no longer reproduces the household, in what sense does the farming population remain a population of farmers?

1. Introduction

Across the Global South, farming is ceasing to be what it long was, the activity that keeps a rural household alive. Smallholders still hold their plots and still plant them, but a growing share of what they need to survive now comes from wages, trade, and migration rather than from the land. This is what scholars have called deagrarianisation, the de-linking of rural livelihoods from agriculture even where cultivation continues [1,2]. It changes not only how rural households earn, but what they are because, when the farm no longer feeds the family, the category of farmer itself begins to loosen. The tension runs particularly deep in the rice economies of monsoon Asia, where dense smallholder populations work shrinking and fragmented plots. Recent work in Thailand finds rice smallholders persisting rather than consolidating, with variation among them explained by resource access and household labour rather than by scale [3], while global census evidence shows farms below 2 hectares making up 84% of all farms while holding roughly 12% of farmland [4]. Indonesia, with millions of sub-hectare paddy households, is a central case of this pattern.

Indonesia sharpens the general pattern into a national puzzle. Rice is not merely a crop here, but a pillar of state legitimacy, and the paddy smallholder has been the target of sustained public effort for more than half a century, from the Green Revolution’s seed and fertiliser packages to the input subsidies and price supports that persist today [5,6]. That effort rests on an assumption rarely examined, that the paddy farmer is first and last a farmer, whose welfare rises and falls with the crop. Decades of research complicate the picture without dislodging the assumption. Studies have traced the fragmentation of holdings, the widening reach of off-farm income, the ageing of the farmer population, and the endurance of rural poverty alongside rising national output [2,7,8]. That coexistence, a nation feeding itself while many of the households growing its rice remain poor, is where the development agenda strains against its own terms. Food security and household welfare are related but not interchangeable outcomes, a tension the SDG framework itself carries between SDG 2 and SDG 1, and a focus on extreme poverty can leave moderate poverty and inequality out of view [9]. Read together, this literature establishes that Indonesian smallholders are differentiating, that the conditions under which they farm and the returns they draw vary widely, and that farming secures a shrinking share of household survival. What it has been slower to ask is what those facts, taken together, mean for how the farming population should be understood.

West Java is where that question can be put most sharply. The province is among Indonesia’s largest rice producers and produces on the backs of smallholder households working predominantly sub-hectare plots. It also spans, within a relatively compact area, the irrigated lowlands of the northern coast and the rainfed uplands of the interior, so the conditions under which paddy farming takes place vary widely within a single administrative and policy jurisdiction. A province that reliably contributes to national rice supply while many of the households producing it remain dependent on income beyond farming provides a particularly useful setting in which to examine the relationship between farm performance and household livelihood [10,11,12].

This is not a new observation. Studies across Indonesia have documented for decades that paddy contributes only a minority share of total household income for many smallholder families [13]. What has been less examined is what that share means. The livelihood literature has tended to frame income diversification as adaptation, a deliberate strategy through which farming households spread risk and supplement seasonal income [14,15,16,17]. That framing carries an implicit assumption that farming is the foundation and other income sources are supplementary. When off-farm income becomes not a supplement but a principal means through which households reproduce themselves, the assumption no longer holds. The result is a farming population maintaining plots that generate positive financial returns while depending substantially on income earned elsewhere. Whether such a population should still be understood as farmers in any sociologically meaningful sense is a question the literature has largely deferred [18,19]. Recent work has begun to press it. Giller et al. [20] argue that the economic viability of farming is now a problem globally rather than locally; Lowder et al. [21] show that farm sizes continue to shrink even as economies grow and argue that viability must be considered within diversified household livelihoods rather than at the farm gate; and Ambarwati et al. [22] show that youth disengagement from Indonesian smallholder farming is associated with structural barriers to land access rather than simply attitudes. Otsuka [23], writing on Indonesian agriculture, draws a different policy conclusion and calls for larger average rice farms on cost grounds.

What the debate lacks is a measure. The distance between a farm that pays and a household that cannot live on it is asserted more often than it is quantified, and without a quantity, it cannot be compared between places or examined against the conditions associated with it. Two things must therefore be held apart. Farm viability is a property of the enterprise, whether the crop returns more than it costs to grow, measured here per hectare. Household sufficiency is a property of the household, whether what the household earns in a year covers what it spends, measured per household. The two are routinely reported together and treated as though they were the same condition, which makes the distance between them difficult to see. This study measures both for the same households, using farm production accounts to assess viability and recorded household income and expenditure to assess sufficiency.

The study draws on a structure and agency framework, in which structures constrain and enable what actors can do while being reproduced or altered through what actors actually do [24,25,26,27]. The framework is useful here because the question is not simply whether smallholder incomes are low, but whether the difference in livelihood outcomes is associated with conditions that farmers inherit or with the actions they take within those conditions. Structural conditions are defined here as factors largely outside a household’s control within a production cycle, principally the area it operates and its access to water. Agency refers to actions undertaken within those conditions, including participation in farmer organisations, investment in equipment, acquisition of technical knowledge, and engagement in work outside the household’s own farm.

The empirical basis is a survey of 1037 paddy farming households across four regencies of West Java: Subang, Indramayu, Tasikmalaya, and Purwakarta. The analysis addresses three questions. First, how are land and water conditions distributed across households and regencies? Second, how does farm viability vary with those conditions? Third, how far does household income cover household expenditure, and does this coverage vary with structural position in the same way as farm viability? The analysis also examines whether households in different structural positions differ in the forms of agency they exercise. The aim is not to treat structure and agency as competing explanations, but to examine how household action takes place within materially differentiated farming conditions.

2. Materials and Methods

2.1. Study Sites and Sampling Frame

The study was conducted in four regencies of West Java, Indonesia: Subang, Indramayu, Purwakarta, and Tasikmalaya. Subang and Indramayu are among the province’s major rice-producing regencies and are located on the northern coastal plain, with relatively favourable access to irrigation infrastructure, rural roads, and major markets. Purwakarta and Tasikmalaya are located further inland and include more heterogeneous and relatively less favourable agro-economic environments. The four regencies were selected purposively to capture contrasting rice-farming contexts, ranging from the productive lowlands of the northern coastal plain to the more heterogeneous upland areas. The purpose of this selection was to enable comparison across contrasting agro-economic settings rather than to obtain a statistically representative sample of West Java as a whole.

The sampling frame consisted of the RDKK (Rencana Definitif Kebutuhan Kelompok), an administrative register of farmer-group members used in the distribution of subsidised agricultural inputs and maintained by agricultural extension services at the regency level. Within each regency, farmers appearing on the available RDKK lists were assigned unique identification numbers and selected by simple random sampling without replacement. Selection was not stratified by farm size, allowing the distribution of farm sizes in the sample to emerge from the random selection process. The unit of sampling was the rice-farming household, represented by the household member responsible for farm management.

The sample size was determined primarily by the resources available for fieldwork rather than by a population-based sample-size formula. Approximately 1000 interviews were initially targeted, with additional interviews conducted during fieldwork to compensate for refusals and incomplete questionnaires. Approximately 1100 questionnaires were collected across the four regencies. Of these, 1038 contained sufficiently complete information on production costs, harvests, and household expenditure to be considered for analysis. One further exclusion, described in Section 2.3.1, leaves an analytical sample of 1037. This is well above the minimum of 384 that Krejcie and Morgan [28] indicate at 95% confidence and a 5% margin of error. The resulting sample provides a substantial number of observations for pooled statistical comparisons, while the smaller numbers in some regency-by-farm-size cells require caution when interpreting subgroup estimates. At 95% confidence, the pooled sample carries a margin of error of 3.0%, ranging from 4.7% in Subang (n = 439) to 8.7% in Purwakarta (n = 128); the smallest regency-by-size cell, the 13 large-farm households of Tasikmalaya, carries a margin of 27%. Cell sizes are therefore reported in the relevant tables, and results for particularly small cells are interpreted as indicative rather than representative.

The number of interviews allocated to each regency was not proportional to the population of rice farmers. The RDKK lists were maintained at the village and farmer-group levels, and consolidated and comparable counts of registered farmers across regencies were not available to the research team at the time of sampling. Consequently, the number of farmers selected in each regency reflected the size and availability of the administrative lists obtained during fieldwork. The original administrative lists were used solely as sampling frames and were not retained by the research team. No population weights were applied in the analysis. Accordingly, comparisons among regencies should be interpreted as comparisons within the study sample and should not be interpreted as estimates of the distribution of rice farmers across West Java.

2.2. Data Collection

Structured interviews were conducted using a questionnaire developed in Bahasa Indonesia and piloted with farmers outside the final sample. It covered socio-demographic characteristics, land tenure and irrigation, farm production costs by component, yield and farm gate price, household expenditure by category, and access to extension, technology, and credit. Household expenditure was recorded monthly for routine items and annually for occasional ones.

Enumerators were trained and familiar with local farming conditions. Labour was recorded as person-days by activity, separately for men and women, as reported by the farmer, with no standard coefficients applied.

Informed consent was obtained orally from every participant, who was told the purpose of the study, the voluntary nature of participation, and the right to withdraw. Responses were coded to protect respondent identities.

2.3. Data Analysis

The analysis distinguishes two questions: whether the farm enterprise is economically viable and whether household income is sufficient to cover household expenditure.

2.3.1. Farm Economic and Livelihood Measures

Questionnaires missing a cost block, the harvest record, or the expenditure block were excluded because the expenditure-coverage measure requires all three. Of the 1038 records that remained, one holding of 35 hectares was excluded as an outlier, being an order of magnitude larger than the next largest, leaving 1037 observations. The effect of this exclusion on the large farm class figures is reported in the note to the farm performance table.

Male and female person-days were valued separately at the prevailing regency wage. Operations contracted as a lump sum for the whole plot (borongan) were recorded at the contracted amount, while harvest labour paid as a share of output (bawon, ceblok) was valued at the reported share. Machinery was charged as a seasonal rental equivalent per hectare, applying the regency rental rate to owned implements. Seed, manure, urea, ZA, TSP, NPK, and pesticides were costed at reported purchase prices. Land tax and irrigation dues were recorded for the 188 households reporting them and were not imputed to the rest. Household-reported farm gate prices were used where available; otherwise, the village, sub-district, or regency median was applied, in that order.

Farm performance was measured per hectare and reported by farm size class, with the median household revenue-to-cost ratio reported alongside the class aggregate. Family labour was valued at the prevailing local wage and included as a farm cost. Land opportunity cost was excluded because most households cultivated land they owned and the analysis concerns whether existing smallholders can live from the land they hold. Rent information was available for only 30 households and was not collected in Indramayu; the observed value of approximately Rp21 million per hectare per year is therefore reported only as an order of magnitude and not included in costing.

The rice year comprises two 4-month seasons. The first is observed, while the second is modelled using information from qualitative interviews, in which farmers described it as a gamble rather than a repeat of the first. Land preparation, machinery, and land tax are held constant across seasons. Fertiliser, pesticide, and associated labour are scaled by an input ratio; seed is held constant because nursery preparation is carried out in full; and harvest labour and output are scaled by a yield ratio. Expected output is then adjusted by the crop failure rate. Parameters are assigned by household according to water reliability, at 0.8, 1.0, and 0 for households with reliable water and 0.6, 0.25, and 0.5 for those without, for the yield ratio, the input ratio, and the crop failure rate, respectively. In total, Eleven 11 alternative settings are reported in the Supplementary Material.

The wage charged to the farm returns to the household as income, following the double-entry logic of farm household accounting [29]. Annual household income therefore comprises farm net return from both seasons, the retained value of family labour, and income earned outside the household’s own farm. Family labour supply is assigned from the field evidence. Households below 1 hectare are credited with up to 55 person-days per season, the 90th percentile of labour use on holdings of 0.3 hectares or less, while holdings above 1 hectare are credited with none, as owners at that scale reported supervising rather than working. Lower- and upper-bound treatments of family labour are reported in the notes to the income table. Expenditure coverage is defined as annual household income divided by annual household expenditure and is reported both as a class aggregate and as a median. It is not expressed per hectare because household living costs do not scale directly with farm area.

2.3.2. Structure and Agency Operationalisation

Farms are classified by operated area into small (<0.5 ha), medium (0.5–1 ha), and large (>1 ha), with a finer split of the small class at 0.15 and 0.3 hectares also reported.

Irrigation type and pump ownership were recorded in separate questionnaire blocks and are not mutually exclusive. A household is classified as water secure if it has technical, semi-technical, or lift irrigation or owns a pump. Simple irrigation does not qualify because the questionnaire defines it as a supply that can neither be regulated nor measured.

Agency is operationalised through farmer-group membership, pump ownership, the main source of technical knowledge, access to training and field guidance, and the type of activity undertaken outside the farm. Selling labour and operating a business or income-generating asset are reported separately. Off-farm income is not used as an explanatory variable for expenditure coverage because it is already included in the coverage measure. Four of the five agency variables cover 603 to 747 households rather than the full sample because the later survey round that collected them did not include Indramayu.

2.3.3. Statistical Analysis

Shapiro–Wilk and Levene tests indicated departures from normality and variance homogeneity, so non-parametric methods were used for group comparisons. Differences across three or more groups were assessed using Kruskal–Wallis with epsilon squared, differences between two groups using Mann–Whitney with rank-biserial correlation, and categorical associations using chi-square with Cramer’s V.

Ordinary least squares (OLS) models with heteroskedasticity-robust standard errors regress the revenue-to-cost ratio, expenditure coverage, and yield on log farm size, water access, irrigation type, and regency. These models assess whether observed differences are associated with particular factors when considered jointly, rather than estimating causal effects.

2.4. Limitations

The survey is cross-sectional. It records current operated area and tenure, but not initial land endowment, inheritance, or land transactions over time. Because operated scale is not fixed, the analysis identifies associations between current farm size and current livelihood adequacy without establishing the direction of that relationship.

Because the sampling frame consists of farmer-group members, farmers outside such groups are underrepresented and may differ in their access to inputs and information. Landless agricultural workers do not appear in the frame because they hold no land to register. Roughly three-quarters of the wages paid by households in the sample are received by workers who are not observed by the survey, placing a substantial group of agricultural workers outside the analysis.

3. Results

3.1. Sample and Structural Conditions

The analysis covers 1037 households. Subang accounts for 42% of the sample, Tasikmalaya for 23%, Indramayu for 22%, and Purwakarta for 12%. The four regencies differ considerably in operated farm size (Table 1). In Tasikmalaya, 74% of households cultivate less than half a hectare, compared with 31% in Subang, where 29% operate more than 1 hectare (chi-square = 140.7, p < 0.001, Cramer V = 0.260).

Table 1. Sample distribution by regency and land size class.

Regency	Small (<0.5 ha)	Medium (0.5–1 ha)	Large (>1 ha)	Total	% of Sample
Subang	137	174	128	439	42.3%
Indramayu	138	66	26	230	22.2%
Tasikmalaya	177	50	13	240	23.1%
Purwakarta	55	50	23	128	12.3%
Total	507	340	190	1037	100.0%
Row share	48.9%	32.8%	18.3%

Size composition differs significantly across regencies (chi-square = 140.7, df = 6, p < 0.001, Cramer V = 0.260). Mean holding: Subang 1.07 ha, Indramayu 0.56 ha, Purwakarta 0.78 ha, Tasikmalaya 0.38 ha; provincial mean 0.76 ha, median 0.50 ha.

Most of the farmers (70%) are between 35 and 60 years of age, while a quarter are over 60 and continue to cultivate (Table 2). Only 4.5% are under 35, and nearly half have farmed for more than 25 years. Formal schooling is limited for most respondents. In addition, 54% completed primary school or less, while 3% hold a college or higher degree.

Table 2. Socioeconomic and agrarian characteristics of respondents.

Characteristic	West Java	Subang	Indramayu	Tasikmalaya	Purwakarta
Age (n = 1037)
<35	4.5%	5.2%	5.7%	2.1%	4.7%
35–60	70.4%	73.6%	73.0%	62.1%	70.3%
>60	25.1%	21.2%	21.3%	35.8%	25.0%
chi-square across regencies		22.6, p < 0.001, V = 0.104
Formal education (n = 1037)
Primary or less	54.4%	49.9%	55.2%	59.2%	59.4%
Junior secondary	20.6%	18.9%	26.1%	17.1%	23.4%
Senior secondary	21.6%	26.7%	18.7%	17.9%	16.4%
Tertiary	3.4%	4.6%	0.0%	5.8%	0.8%
chi-square across regencies		34.8, p < 0.001, V = 0.106
Farming experience (n = 1013)
1–10 years	17.8%	19.8%	11.3%	21.6%	15.9%
11–25 years	34.6%	37.2%	30.9%	29.3%	42.9%
>25 years	47.6%	43.1%	57.8%	49.1%	41.3%
chi-square across regencies		22.6, p < 0.001, V = 0.106
Land tenure (n = 1012)
Owner-operated	76.0%	76.0%	97.3%	68.4%	53.1%
Sharecropped	17.9%	12.2%	1.4%	30.3%	43.0%
Rented in	5.4%	10.1%	1.4%	1.3%	3.9%
Pawned in	0.6%	1.4%	0.0%	0.0%	0.0%
Other	0.1%	0.2%	0.0%	0.0%	0.0%
chi-square across regencies		171.2, p < 0.001, V = 0.237
Irrigation type (n = 1037)
Technical irrigation	31.7%	67.0%	10.0%	4.6%	0.8%
Semi-technical	15.3%	14.4%	10.4%	18.8%	21.1%
Simple irrigation	18.6%	4.6%	0.9%	61.3%	18.8%
Lift irrigation (pump/well/river)	3.3%	2.3%	0.0%	1.7%	15.6%
Rainfed	31.1%	11.8%	78.7%	13.8%	43.8%
chi-square across regencies		949.3, p < 0.001, V = 0.552
Pump ownership (n = 1037)
Owns a pump	24.3%	12.3%	79.6%	0.8%	10.2%
No pump	75.7%	87.7%	20.4%	99.2%	89.8%
chi-square across regencies		502.0, p < 0.001, V = 0.696
Effective water access (n = 1037)
Water secure	69.2%	88.4%	93.0%	25.8%	42.2%
Not water secure	30.8%	11.6%	7.0%	74.2%	57.8%
chi-square across regencies		393.0, p < 0.001, V = 0.616
Off-farm employment (n = 1037)
Has off-farm work	58.1%	56.5%	50.9%	73.8%	47.7%
No off-farm work	41.9%	43.5%	49.1%	26.2%	52.3%
chi-square across regencies		35.3, p < 0.001, V = 0.184

Mean age 53.4 years (median 53); mean farming experience 25.9 years (n = 1013). Note: Irrigation type and pump ownership come from two different questionnaire blocks and are not alternatives to each other. “Effective water access” combines them: a household is water secure if it has technical, semi-technical, or lift irrigation or owns a pump. This matters in Indramayu, where 181 households farm rainfed land, but 167 of them own a pump, and in Tasikmalaya, where only two households own a pump. Percentages are computed on valid answers; the valid n is given per block. Across the five irrigation types, yield differs with a large effect (Kruskal–Wallis p < 0.001, ε² 0.188), while the revenue-to-cost ratio and expenditure coverage differ far less (ε² 0.026 and 0.012).

Almost half of the households cultivate less than 0.5 hectares, a third operate between 0.5 and 1 hectare, and 18% operate more than 1 hectare. The mean operated area is 0.76 hectares, and the median is 0.50 hectares. Most of this land is owner-cultivated. Of the 1012 households with recorded tenure, 76% farm their own land, 18% sharecrop, and 5% rent (Table 2).

Water access varies substantially across regencies. Because irrigation type and pump ownership were recorded separately, both were considered in classifying water security. Overall, 69% of households have reliable access to water, and 31% do not, with a large difference between regencies (chi-square = 393.0, p < 0.001, Cramer V = 0.616). Of the 319 households classified as not water secure, 190 use simple irrigation, and 129 cultivate rainfed land. The questionnaire defines simple irrigation as a supply that can neither be regulated nor measured.

Irrigation type alone gives a different picture from the combined measure. In Indramayu, 79% of households cultivate rainfed land, yet 167 of the 181 rainfed households own a pump, resulting in 93% of households being classified as water secure. In Tasikmalaya, only 33 of 240 households cultivate rainfed land, but 147 use simple irrigation, and few own pumps; consequently, 178 households, or 74%, are classified as not water secure.

Pump ownership is concentrated in Indramayu, where 80% of households own a pump, compared with 12% in Subang, 10% in Purwakarta, and 1% in Tasikmalaya. Overall, 24% of households own a pump.

3.2. Farm Economic Viability

Total cost per hectare is similar across the three farm size classes (Table 3), at Rp15.82 million on small farms, Rp16.62 million on medium farms, and Rp16.40 million on large farms (Kruskal–Wallis p = 0.860). The composition of costs, however, differs by farm size. Labour accounts for 79.1% of total cost on small farms and 63.7% on large farms, while machinery increases from 6.7% to 12.7% and purchased inputs from 13.9% to 23.5%. Labour use decreases from 162 to 111 person-days per hectare (p < 0.001, ε² = 0.113).

Table 3. Farm cost structure and returns per hectare, wet season, by land size class.

Item (Rp per Hectare)	Small (<0.5 ha)	Medium (0.5–1 ha)	Large (>1 ha)	p	ε2
Male labour	7,200,502	7,948,887	7,111,809	0.016	0.006
Female labour	4,110,848	3,685,010	2,576,789	<0.001	0.163
Contract labour (borongan)	1,202,130	901,878	764,495	<0.001	0.036
Total labour	12,513,480	12,535,775	10,453,092	<0.001	0.035
Machinery and equipment	1,067,430	1,494,375	2,078,001	<0.001	0.084
Seed	273,835	283,602	322,846	<0.001	0.017
Fertiliser	1,458,231	1,467,006	2,475,507	<0.001	0.098
of which urea	610,954	659,149	1,629,504	<0.001	0.160
Pesticides	461,980	811,001	1,055,780	<0.001	0.137
Total purchased inputs	2,194,046	2,561,609	3,854,132	<0.001	0.131
Land tax and other charges	45,548	28,302	15,688	0.027	0.005
Total Cost	15,820,504	16,620,061	16,400,914	0.860	0.002
Revenue (value of output)	26,065,686	30,197,729	31,639,556	<0.001	0.111
Net return	10,245,182	13,577,668	15,238,642	<0.001	0.109
Cost shares
Labour	79.1%	75.4%	63.7%	-	-
Machinery	6.7%	9.0%	12.7%	-	-
Purchased inputs	13.9%	15.4%	23.5%	-	-
Tax and other	0.3%	0.2%	0.1%	-	-
Physical and ratio measures
Labour use (person-days/ha)	162	146	111	<0.001	0.113
Yield (kg GKP/ha)	5163	5833	6163	<0.001	0.122
Farm gate price (Rp/kg)	5049	5177	5134	<0.001	0.012
R/C ratio (aggregate)	1.65	1.82	1.93	<0.001	0.062
R/C ratio (median household)	1.51	1.77	1.85
Per household (not per hectare)
Households (n)	507	340	190
Mean holding (ha)	0.26	0.74	2.16
Mean total cost	4,039,724	12,293,957	35,362,356
Mean revenue	6,655,804	22,337,438	68,218,712
Mean net return	2,616,080	10,043,481	32,856,356

Note: p-values are Kruskal–Wallis tests on household-level per-hectare values across the three classes; ε² is the effect size (0.01 small, 0.06 moderate, 0.14 large). Family labour is valued at the prevailing local wage and charged as a cost here, so the net return is the return to land and management only.

Female labour costs also decrease with farm size, from Rp4.11 million per hectare on small farms to Rp2.58 million on large farms (p < 0.001, ε² = 0.163), the largest effect within the labour-cost components.

Urea expenditure increases from Rp0.61 million per hectare on small farms to Rp1.63 million on large farms (p < 0.001, ε² = 0.160).

Revenue per hectare differs across size classes, at Rp26.07 million, Rp30.20 million, and Rp31.64 million, respectively (p < 0.001), corresponding to yields of 5163, 5833, and 6163 kg per hectare. Farm gate prices are comparable across classes. The revenue-to-cost ratio therefore increases with farm size, reaching 1.65, 1.82, and 1.93 in aggregate and 1.51, 1.77, and 1.85 at the median household (p < 0.001, ε² = 0.062). Paddy farming has a revenue-to-cost ratio above one in every size class and regency in the sample, which indicates that it is profitable.

The association between farm size and the revenue-to-cost ratio differs by regency (Table 4). It is significant in Indramayu (p < 0.001, ε² = 0.165), Purwakarta (p < 0.001, ε² = 0.176), and Tasikmalaya (p < 0.001, ε² = 0.060), but not in Subang (p = 0.94). Subang accounts for 42% of the sample and 67% of all large farms, and its small farms obtain returns per hectare comparable to those of large farms.

Table 4. Cross-regency comparison by land size class.

Item	Subang			Indramayu			Tasikmalaya			Purwakarta
	S	M	L	S	M	L	S	M	L	S	M	L
n	137	174	128	138	66	26	177	50	13	55	50	23
Labour share of cost	79%	81%	67%	77%	57%	41%	81%	78%	67%	82%	79%	66%
Machinery share of cost	7%	6%	11%	6%	18%	22%	5%	8%	11%	10%	10%	10%
Input share of cost	14%	13%	22%	16%	25%	36%	14%	15%	22%	9%	11%	24%
Cost (Rp million/ha)	15.5	18.8	17.7	20.2	17.7	16.4	12.2	12.0	10.9	16.1	12.1	10.3
Yield (kg/ha)	5628	6287	6300	5511	5610	5962	4376	4927	4908	5150	5383	5961
R/C (aggregate)	1.83	1.75	1.82	1.46	1.70	2.01	1.63	1.87	1.99	1.71	2.36	3.01
R/C (median)	1.82	1.72	1.77	1.28	1.74	1.88	1.49	1.74	2.19	1.61	2.48	3.50
Within-regency test of the size effect (Kruskal–Wallis on household R/C across the three classes):
Regency					N			p			ε2
Subang					439			0.936			−0.004
Indramayu					230			<0.001			0.165
Tasikmalaya					240			<0.001			0.060
Purwakarta					128			<0.001			0.176
West Java (pooled)					1037			<0.001			0.062

Note: S = small, M = medium, L = large. This table reports farm performance only; household income and expenditure are reported separately. Pooled West Java figures by land size class are given in Table 3 and are not repeated here. The median R/C values entering each within-regency test are the R/C (median) row above.

Costs differ more between regencies than returns. Cultivation costs Rp17.87 million per hectare in Indramayu and Rp17.77 million in Subang, compared with Rp11.96 million in Purwakarta and Rp11.87 million in Tasikmalaya (p < 0.001, ε² = 0.156). Median household revenue-to-cost ratios are 1.95 in Purwakarta, 1.78 in Subang, 1.58 in Tasikmalaya, and 1.55 in Indramayu (p < 0.001).

The size gradient is concentrated among the smallest holdings (Table 5). The 143 households cultivating less than 0.15 hectares, representing 14% of the sample, use 258 person-days per hectare and have a revenue-to-cost ratio of 1.07. The ratio increases to 1.55 among households cultivating 0.15–0.3 hectares and then remains relatively stable at 1.88 for 0.3–0.5 hectares, 1.82 for 0.5–1 hectare, and 1.93 above 1 hectare. Excluding the smallest class reduces the effect size for farm size from 0.188 to 0.030, indicating that the observed size-related difference is concentrated below 0.3 hectares.

Table 5. Farm performance by detailed land size class.

Land Size Class	n	Share	Mean ha	Person-days/ha	Cost/ha (Rp)	Yield (kg/ha)	R/C (Aggregate)	R/C (Median)	Negative net Return, MT1 (n)	Negative net Return, Annual (n)
<0.15 ha	143	13.8%	0.107	258	22,159,859	4741	1.07	1.06	64	81
0.15–0.3 ha	169	16.3%	0.232	158	15,690,426	4839	1.55	1.57	17	33
0.3–0.5 ha	195	18.8%	0.385	145	14,595,298	5418	1.88	1.83	9	25
0.5–1 ha	340	32.8%	0.740	146	16,620,061	5833	1.82	1.77	9	28
>1 ha	190	18.3%	2.156	111	16,400,914	6163	1.93	1.85	9	20

Kruskal–Wallis across all five bands: p < 0.001, ε² = 0.188 (large effect). Excluding the smallest class: p < 0.001, ε² = 0.030 (small effect, n = 894). Note: The five detailed classes are nested within the three land size classes used in Table 3 and Table 4. The first three together form the small class (<0.5 ha), the fourth is the medium class, and the fifth is the large class.

Part of the higher labour use on the smallest holdings reflects a measurement floor since an operation cannot be recorded as less than one person-day regardless of plot size. Losses are also concentrated among the smallest holdings. With family labour valued at the market wage, 108 households record a negative net return in the wet season and 187 across the year. Of the 108 wet-season losses, 64 occur among households cultivating less than 0.15 hectares.

The last two columns count the households whose farm runs at a loss, that is, whose net return is negative once family labour is charged at the market wage.

3.3. Household Income and Expenditure Coverage

Annual farm net return increases from Rp3.35 million on small farms to Rp14.14 million on medium farms and Rp50.66 million on large farms (Table 6). Including the retained value of family labour and income earned outside the household’s own farm gives total household incomes of Rp23.43 million, Rp37.89 million, and Rp70.83 million, respectively. Household expenditure increases from Rp29.58 million to Rp46.25 million across the size classes.

Table 6. Annual household income and coverage of household expenditure.

Item (Rp per Household per Year)	Small (<0.5 ha)	Medium (0.5–1 ha)	Large (>1 ha)	West Java
Net return, wet season (MT1)	2,616,080	10,043,481	32,856,356	10,591,942
Net return, dry season (MT2, modelled)	733,565	4,096,214	17,800,246	4,963,045
Farm net return, annual	3,349,646	14,139,695	50,656,602	15,554,987
Imputed value of family labour retained	5,399,638	8,878,647	0	5,550,971
Off-farm income	14,678,225	14,876,294	20,175,474	15,750,376
Total household income	23,427,509	37,894,636	70,832,076	36,856,333
Household expenditure	29,581,928	34,746,496	46,247,363	34,328,684
Expenditure coverage (aggregate)	79%	109%	153%	107%
Expenditure coverage (median household)	80%	102%	117%	96%
Households below full coverage	61%	48%	41%	53%
Farming as share of total income	37%	61%	72%	57%

Note: Household income follows the double-entry logic of farm household accounting. Family labour is charged to the farm business at the prevailing local wage, which is why the farm net return above is a return to land and management only, and the same amount returns to the household as income. Households below 1 hectare are assumed to supply family labour up to a ceiling of 55 person-days per season, the 90th percentile of labour use on holdings of 0.3 hectares or less. Holdings above 1 hectare are assumed to hire all field labour and to supervise only, consistent with what farmers described in interviews, so no family labour value is imputed to them.

Expenditure coverage reaches 79% on small farms, 109% on medium farms, and 153% on large farms in aggregate. At the median household, coverage is 80%, 102%, and 117%, respectively (p < 0.001, ε² = 0.040). In addition, 61% of small farm households and 48% of medium farm households do not cover their household expenditure, compared with 41% of large farm households. When family labour is charged to the farm without returning its value to the household, coverage is 61%, 84%, and 153%; when all family-labour payments are returned to the household, it is 81%, 135%, and 249%.

Two bounding alternatives were computed and are not reported in the table. Charging family labour to the farm and not returning it to the household gives coverage of 61%, 84%, and 153% for the three classes, the lower bound. Returning every labour payment to the household, including payments to hired workers, gives 81%, 135%, and 249%, the upper bound. Coverage differs significantly across classes on all three treatments (main: p < 0.001, ε² 0.040; lower bound: p < 0.001; upper bound: p < 0.001).

Farming accounts for 37% of household income among small farms, 61% among medium farms, and 72% among large farms. Removing the second season reduces expenditure coverage to 77%, 97%, and 115% across the three size classes. Across the 11 parameter settings tested, coverage ranges from 70% to 82% on small farms, 79% to 116% on medium farms, and 74% to 170% on large farms (Supplementary Material). The results are most affected by assumptions concerning crop failure.

3.4. Structure and Agency

Most institutional and agency measures do not differ significantly by farm size (Table 7). Farmer-group membership is 52%, 55%, and 56% across the three size classes (chi-square p = 0.522), while pump ownership is 25%, 24%, and 24% (p = 0.907). The main source of technical knowledge does not differ by farm size (p = 0.168), nor does reported access to training (p = 0.723). Farmer-group members also do not differ from non-members in revenue-to-cost ratio or yield (Mann–Whitney p = 0.475 and p = 0.708).

Table 7. Institutional access and agency by land size class.

Variable	Small (<0.5 ha)	Medium (0.5–1 ha)	Large (>1 ha)	All	n	chi-Square (p)	Cramer V
Farmer-group membership					732	1.30 (0.522)	0.042
Member	51.5%	55.2%	56.4%	53.8%
Not a member	48.5%	44.8%	43.6%	46.2%
Irrigation pump ownership					1037	0.19 (0.907)	0.014
Owns a pump	24.9%	23.5%	24.2%	24.3%
No pump	75.1%	76.5%	75.8%	75.7%
Main source of technical knowledge					747	11.64 (0.168)	0.088
Self-taught	58.6%	51.2%	53.1%	54.9%
Government extension	23.1%	21.7%	20.6%	22.1%
Farmer group or institution	14.7%	18.5%	21.2%	17.4%
Social media or internet	3.0%	7.1%	4.4%	4.7%
Other	0.6%	1.6%	0.6%	0.9%
Access to training					631	2.07 (0.723)	0.040
Easy	61.7%	55.6%	58.0%	58.8%
Difficult	24.2%	28.7%	26.1%	26.1%
Not available	14.1%	15.7%	15.9%	15.1%
Access to field guidance					632	24.43 (<0.001)	0.139
Easy	57.9%	69.0%	76.1%	65.7%
Difficult	14.7%	14.4%	15.9%	14.9%
Not available	27.3%	16.7%	8.0%	19.5%
Type of off-farm strategy (households with off-farm work only)					603	52.63 (<0.001)	0.209
Selling labour	57.2%	36.8%	22.4%	46.9%
Own business or asset	25.2%	42.0%	42.1%	32.2%
Salaried employment or transfer	7.4%	14.4%	23.7%	11.4%
Other	10.2%	6.9%	11.8%	9.5%
Does membership pay? (Mann–Whitney, members vs. non-members, n = 732)
Outcome	Member (median)	Non-member (median)		p		rank-biserial
R/C ratio	1.82	1.76		0.475		0.031
Yield (kg/ha)	5500.00	5714.29		0.708		−0.016
Cost per ha	14,620,350.00	15,082,803.57		0.112		−0.068
Coverage	0.94	0.89		0.149		0.062

Note: Percentages are computed on households that answered the item; the valid n is given in the table. The agency block (training, field guidance, group membership, and knowledge source) was asked in a later survey round and is not available for the whole sample, which is why n varies.

Access to field guidance does differ by farm size (p < 0.001, Cramer V = 0.139). A larger difference appears in the type of activity undertaken outside the household’s own farm. Among the 603 households reporting off-farm work (58% of the sample; Table 2), 57% of small farm households sell their labour, and 25% operate a business or income-earning asset, compared with 22% and 42%, respectively, among large farm households (chi-square p < 0.001, Cramer V = 0.209). The ratio of wage labour to own-account work is 2.27:1 among small farms, 0.88:1 among medium farms, and 0.53:1 among large farms.

Yield differs across irrigation types (p < 0.001, ε² = 0.188), whereas differences in the revenue-to-cost ratio and expenditure coverage are smaller (ε² = 0.026 and 0.012, respectively). In ordinary least squares models with heteroskedasticity-robust standard errors, log farm size is significant for the revenue-to-cost ratio, expenditure coverage, and yield (Supplementary Table S1). Irrigation type remains jointly associated with yield and with the revenue-to-cost ratio after farm size and regency are controlled, but not with expenditure coverage (joint p = 0.447). The combined water-access measure adds no further explanation, and regency effects remain significant in all three models.

4. Discussion

4.1. Farm Viability and Household Reproduction

Indonesian agricultural policy has long operated on the assumption that the barriers facing paddy farmers are primarily technical or behavioural and can therefore be addressed through productivity programmes, technology adoption, or individual capacity building [5,6,30,31]. The evidence from 1037 farming households across four regencies in West Java points to a different problem. Farm size and location are strongly associated with whether farming can support the household, while measured differences in farm costs and several forms of institutional engagement are comparatively limited. The argument is not that farmer agency is irrelevant, but that agency operates within material conditions that differ substantially across the farming population.

The central finding is a gap between farm viability and household sufficiency. Paddy farming is financially viable in every size class and regency studied, with a revenue-to-cost ratio above 1 throughout and increasing from 1.65 on small holdings to 1.93 on large ones. At the same time, 61% of small farm households and 48% of medium farm households do not cover their household expenditure even after the second cropping season, and off-farm income is included (Table 6). The issue is therefore not simply whether farming generates a positive return. It is whether the return generated by the amount of land operated is sufficient for the household that depends on it.

This distinction gives measurable form to a pattern long described in the literature on deagrarianisation and agrarian differentiation [2,18,32]. Here, farm performance and household expenditure are measured for the same households, making visible the distance between a farm that pays and a household that cannot live from it. Cost per hectare does not differ significantly across size classes, while farmer-group membership and the main source of technical knowledge also show little systematic variation by farm size (Table 3 and Table 7). These findings provide little evidence that the gap is primarily explained by differences in the measured dimensions of farm management or institutional engagement. It is more closely associated with the amount of land operated and the conditions of the regency in which it is operated.

The floor matters more than the gradient. Below 0.3 hectares, the revenue-to-cost ratio increases from 1.07 among households cultivating less than 0.15 hectares to 1.55 among those cultivating 0.15–0.3 hectares. Above 0.3 hectares, it changes little, ranging from 1.82 to 1.93 across the remaining classes (Table 5). Excluding the smallest class reduces the effect size for farm size from 0.188 to 0.030, while losses are also concentrated among the smallest holdings. The result is therefore not a simple gradient in which every increase in farm size produces a corresponding improvement in farm efficiency. Rather, the principal disadvantage is concentrated at the very bottom of the size distribution.

This complicates both the modernisation argument that consolidation necessarily drives efficiency [23] and the smallholder-idealist argument that smaller farms are inherently more productive [33]. Between 0.3 and more than 1 hectare, the differences in return per hectare are relatively small. Ricciardi et al. [34] similarly find higher yields on smaller farms but no conclusive difference in profits by farm size. What the present results add is a lower scale at which the economic position of the household changes markedly. The survey cannot establish how individual holdings reached this scale, but the result is consistent with the longer history of land fragmentation in Java associated with inheritance, population pressure, and limited redistribution [7,35,36,37]. Better farm management cannot, by itself, increase the amount of land available to a household.

4.2. Water, Diversification, and Differentiation

Water reveals a different dimension of the same problem. Irrigation affects production more clearly than household livelihood adequacy. Yield differs across irrigation types, while the corresponding differences in revenue-to-cost ratio and expenditure coverage are much smaller (Table 2; Supplementary Table S1). Once farm size and regency are considered together, irrigation contributes relatively little to explaining expenditure coverage. The household question is therefore not simply whether water is available, but what other conditions determine the return that a household can obtain from the land it operates.

The Indramayu case shows what a conventional irrigation classification can miss: 79% of its farmers work land recorded as rainfed, yet 80% of households own a pump, and 93% are classified as water secure when irrigation and pump ownership are considered together. Limited canal access is therefore partly compensated by private expenditure. Machinery cost in Indramayu is Rp2.85 million per hectare, compared with Rp1.70 million in Subang and Rp0.87 million in Tasikmalaya, and within Indramayu, pump owners spend Rp3.18 million per hectare compared with Rp1.46 million among households without pumps. What this expenditure buys is broadly comparable production rather than a clear advantage, since neither yield nor the revenue-to-cost ratio differs significantly between pump owners and non-owners.

The legacy of uneven irrigation investment [38,39,40,41], therefore, appears in these data less as a simple yield gap than as a difference in who bears the cost of obtaining water. Where public irrigation provides reliable access, part of that service is external to the farm budget; where access is less reliable, households can substitute private pumping. The result is consistent with the broader argument that resource-constrained farmers may have to spend more simply to achieve comparable production conditions [42,43]. The persistent regency effects suggest that these differences extend beyond irrigation to other place-specific conditions, including infrastructure, markets, and access to off-farm employment.

This history reaches beyond irrigation. Fragmented holdings, uneven water infrastructure, input dependence, and a persistent smallholder income gap are not natural features of the agrarian environment. They are the accumulated residue of successive modernisation projects, most decisively the Green Revolution [38,44,45,46,47,48], whose benefits were distributed unequally. Farmers with prior access to land, capital, and irrigation pulled ahead, and those without fell further behind [45,49,50]. The present data suggest that those advantages and disadvantages are still operating rather than being eroded by market forces or technological progress. The floor at 0.3 hectares, the private cost of water in Indramayu, and the composition of off-farm work are consistent with a development trajectory that prioritised aggregate output over the equitable distribution of productive conditions.

The same question of position appears in livelihood diversification. Much of the rural livelihoods’ literature treats diversification as a strategic portfolio for spreading risk [15,16,29,51,52,53,54]. That interpretation fits only part of this sample. Farming supplies 37% of what a small farm household lives on, compared with 72% for a large farm household (Table 6), and the form of off-farm activity differs by farm size. Among households with off-farm work, 57% of small farms sell their labour, and 25% operate a business or income-earning asset, compared with 22% and 42%, respectively, among large farms (Table 7).

For the smallest holders, therefore, off-farm work is often less a portfolio strategy than a means of covering the gap left by the insufficiency of their own farm. The distinction is close to Bernstein’s [18] concept of a reproduction squeeze, in which households retain access to land and continue to cultivate it while also selling labour because the farm cannot provide sufficient household income. Among larger farms, off-farm activity is more often associated with own-account enterprise and income-generating assets. Habibi [55] makes a similar distinction among smallholders in Sumatra, arguing that those compelled to sell labour should be considered in relation to the broader classes of labour rather than within an undifferentiated category of peasantry.

The age structure gives this differentiation a further dimension. Farmers under 35 comprise only 4.5% of the sample, while a quarter are over 60, and nearly half have more than 25 years of farming experience. This is regularly treated as a succession problem [32,56,57,58] to be met with targeted engagement programmes [22,59,60], a framing that places the burden of renewal on young people’s motivations rather than on the conditions they are asked to enter. The present data cannot establish why younger people enter or leave farming, and no causal claim is made here. They do, however, show the economic conditions against which such decisions are made. Ambarwati et al. [22] locate youth disengagement on Java and Flores in structural barriers to land access rather than in attitudes, and Coleman et al. [61] reach a similar conclusion in the Mekong Delta, where farming itself is unattractive as a career while agricultural services are not. The small farm is not simply a farm with low productivity; for many households, it is a productive but insufficient base for household reproduction. This provides a material context for the wider literature on the declining centrality of farming to rural livelihoods in Southeast Asia [2,32].

These results also complicate a straightforward reading of the development goals. West Java meets the terms of SDG 2 by producing rice reliably across every class of holding, while a majority of the households doing the producing cannot cover their own needs from it, which is the poverty SDG 1 was written to end. Woodhill et al. [9] identify precisely this blind spot, where a focus on extreme poverty leaves moderate poverty and inequality invisible. National food security and household freedom from poverty are two achievements here rather than one, and what holds them apart is what SDG 10 names.

4.3. What Remains of the Farmer?

The findings raise a question that the survey cannot answer by itself, and this section offers an interpretation arising from the results rather than a conclusion demonstrated by them.

If paddy farming covers only a minority of a smallholder household’s needs and the remainder is earned on construction sites, in markets, or in the industrial corridors of the province, in what sense is that household still a farming household? The question matters because the answer differs according to how the household is viewed. In agricultural policy, these households are farmers because they hold land and cultivate paddy. In livelihood terms, however, many reproduce themselves substantially outside agriculture. Bernstein’s [18] questions of who owns, who labours, and how households reproduce themselves therefore provide a useful way of looking beyond the administrative category of the farmer.

Yet the category does not simply dissolve. These households continue to farm, and to farm profitably, even when farming ceases to sustain the household. That persistence is not necessarily economic irrationality, nor simply a residue of tradition. Land remains an asset, paddy generates a positive return, and cultivation continues alongside wage labour and other income-generating activities. In this respect, the findings are consistent with van der Ploeg’s [62,63] argument concerning the durability of peasant farming under conditions in which a narrow economic calculation might suggest its disappearance. West Java’s smallholders occupy an intermediate position, neither households whose reproduction depends entirely on farming nor rural workers who have left the land altogether.

This is also where the structure-and-agency perspective becomes important. The agency evidence does not suggest that smallholders are passive. Farmer-group membership, pump ownership, and the main source of technical knowledge do not vary substantially by farm size (Table 7). What differs is the form that livelihood action takes. Smallholders are more likely to sell labour, while larger farms are more likely to engage in own-account activities or income-generating assets. Structure, therefore, does not eliminate agency, but it shapes the opportunities and returns attached to different forms of action [24,25,26,27].

The implication is not that the farmer category should be abandoned, but that it is insufficient on its own. A household can be a farmer institutionally, cultivate a profitable crop economically, and yet depend substantially on labour outside the farm for its reproduction. Treating these as equivalent conditions risks obscuring the differentiation within the smallholder population. It also helps explain why policies centred on yields, inputs, or technology may improve the performance of the farm without resolving the livelihood position of the household.

The study cannot determine how these positions develop over time. Because the survey is cross-sectional, it cannot establish whether limited land produces livelihood insecurity or whether economically constrained households become concentrated on smaller holdings through inheritance, rental, purchase, or other land transactions. Longitudinal research would be needed to establish these trajectories. What the present evidence does establish is a distinction that is often blurred in both policy and analysis. A profitable farm is not necessarily a sufficient livelihood. The question of what remains of the farmer, therefore, lies not in whether smallholders still cultivate land, but in how farming, wage labour, and other forms of livelihood are combined within an increasingly differentiated agrarian economy.

5. Conclusions

Paddy farming in West Java does not cease to be economically viable when it ceases to be sufficient as a livelihood. The two can coexist because they operate at different scales. The farm generates a return on the land, while the household must live from the total return generated by the amount of land it controls. Once that distinction is made, the apparent puzzle of the smallholder becomes less a question of why farmers fail to make their farms productive than of why productive farms are no longer large enough to sustain the households attached to them.

This points to a change in how smallholder agriculture should be understood. The critical problem is not simply the efficiency of the farm, but the relationship between the scale of production and the reproduction of the household. Below roughly 0.3 hectares, farming may remain profitable while becoming incapable of providing a sufficient livelihood. Households consequently do not necessarily leave agriculture; they combine cultivation with wage labour, businesses, and other sources of income. Farming persists, but its place within household reproduction changes.

These findings attach to instruments West Java already operates, and they are stated here as targets for those instruments rather than as predictions of what implementation will deliver. The protection of paddy land falls under Regional Regulation No. 27 of 2010 on sustainable food agricultural land [64], and the floor identified here gives that instrument a sharper test, whether protected land remains in units a household can live from, not only whether it stays agricultural. The regeneration of the farming population is addressed by the province’s Petani Milenial programme [65], which recruits and trains young farmers; the present findings mark the limit of that approach because youth disengagement tracks the economics of farm scale rather than skills or motivation. The fertilizer price differences documented above fall within the provincial management of subsidised fertilizer under Regional Regulation No. 5 of 2011 [66]. Differentiated support follows from the findings directly, water infrastructure where households now pump privately, as in Indramayu, and land access where holdings are smallest, as in Tasikmalaya. How far any of this travels from regulation to field is a separate question, and the distance between policy design and delivery in smallholder agriculture is well documented [5,47].

The farmer, therefore, cannot be defined adequately by the possession or cultivation of land alone. A household may remain a farming household institutionally while becoming increasingly dependent on labour and income outside its own farm. What appears administratively as the persistence of smallholder farming can thus coexist with a deeper process of agrarian differentiation. The relevant question is no longer simply whether smallholders can remain farmers, but whether farming remains sufficient to reproduce them as farmers.

This is the broader significance of the West Java case. Smallholder persistence should not be mistaken for smallholder sufficiency. It is also why meeting the food security promise of SDG 2 at the provincial scale has not delivered the poverty and inequality promises of SDG 1 and SDG 10 at the household scale. Where land remains productive, but the livelihood built upon it does not, the future of smallholder agriculture depends not only on improving what happens within the farm, but on whether the wider agrarian system provides households with a viable basis for remaining on the land.