• No results found

The distribution of asset poverty in the study area

Chapter 2 Poverty and food insecurity in a remote rural area of Papua New

2.4 The distribution of asset poverty in the study area

As noted in the introduction, there has been limited research in PNG that focuses on the characteristics of poverty within remote areas known to be disadvantaged. While there is no doubt that the study area is poor, not all households are equally poor.

Given the relatively low cash-flow in the study area and labour and goods market failures, an asset based measure was utilised to assess relative poverty. Imputing the kina value of assets, in the absence of markets and price information, was not

36

considered practical or meaningful. As noted in Chapter 1, the study area operates as an autarky. Using prices from an external area to value self-production would not be meaningful, particularly given that there is no intention to compare the data collected with data from other villages. In addition, there are theoretical reasons for using asset- based measures in these types of village level studies. Asset based measures are thought to be more representative of long-term welfare than consumption or income, which in subsistence environments may fluctuate seasonally (Filmer and Pritchett 2001, p.155). Asset based measures of poverty have been shown to be accurate at identifying poor households when compared with expenditure and consumption based measures (Filmer and Scott 2012, p.360). In the case of Papua New Guinea, Sahn and Stifel (2003) used Papua New Guinea Household Survey (1996) data on child nutrition outcomes to compare predictions generated from an asset index and expenditure measures. The ranking of household welfare based on the asset index was found to be better in predicting child nutrition outcomes than reported expenditures (p.484).

The asset index in this research was constructed drawing on a list of assets compiled with reference to the 2006 DHS, and the PNG economic and anthropological literature. Housing quality indicators were excluded on the grounds that there is very limited variation in house quality (see Table 2.2). Assets were grouped into three broad categories: livestock, land and consumer goods.

The livestock included in the index were pigs and chickens. Guyer (1997) notes that, in the absence of formal sector financial institutions, many of the assets of the poor need to be multi-purpose and that this quality is more important when income sources are fragile (p.118). Assets can be more than stores of value, or a means of producing an income stream, they can also be important in the context of social and cultural processes (Guyer 1997, p.123). In the Highlands, pigs are multi-purpose assets that have an important social and cultural role. Pigs are used as a store of protein, a means of compensation, and for strengthening relationships, settling marriages, curing the sick and in ceremonies associated with burying the dead (Boyd 2001, p.267). The multiple roles of pigs may make households reluctant to sell them when they need access to cash. Chickens, on the other hand, hold no cultural or ceremonial importance. They are also

37

of low value, so may be less useful than other livestock as a form of insurance against risks (Hesselberg and Yaro 2006, p.49).

Only the size of coffee gardens was included in the asset index. It was assumed that only households able to meet their subsistence needs would own coffee gardens. The size of coffee gardens was measured with a hand-held Global Positioning System (GPS). For the 29 households that had multiple coffee gardens, the closest was measured and then the size of others estimated relative to the first garden. Carletto et al (2015) find that self-reported estimates of garden size tend to over-report land area, and that this may be exacerbated in sloped areas, such as those in the study area (p.595-596). In addition, they find that the over-estimation problem is worse for those at the lower end of the size distribution (p.595). In the case of the study area, given GPS was used to measure the size of the first coffee garden, this source of bias is less likely. However, it is possible that there was overestimation of garden size by those with multiple gardens. Many households had a number of relatively small food gardens dispersed over a distance. This made it impractical to measure food garden size with the hand-held GPS. In addition, the absence of developed food markets suggests that the focus of households was primarily on subsistence production, not income generation, and that few households were able to generate a surplus. This in turn suggests that the size of food gardens would not be a good predictor of economic well-being.

Household goods were limited in the study area. Cooking utensils and gardening tools were included in the index as these were identified through discussions with NGO staff and researchers familiar with the area as being a means of differentiating between the wealth of households. Interviews with several of the polygynous household heads confirmed that ownership of cooking utensils and gardening tools was not shared between wives. And in the case of gardening tools, the number of adults in the household represented the number of potential garden labourers. Therefore, in the apparent absence of economies of scale, cooking utensils were divided by the number of wives in the household and gardening tools by the number of adults. Other household goods such as mobile phones and radios were also included in the asset index. Descriptive statistics of each of the items included in the index are at Table 2.6.

38

Table 2.6: Household ownership and factor loadings of assets included in the index

Asset type Average/ (Standard

deviation)

Per cent of

households Factor Score

Livestock

Pigs per household 1.3

(1.83)

No pigs 44.3 -0.38

1 to 4 50.6 0.30

5 or more 5.1 0.19

Chickens per household 3.4

(8.17)

No chickens 55.7 -0.37

1 to 5 26.6 0.22

6 or more 17.7 0.22

Coffee garden size (square metres) 3469.2

(3956.78)

No coffee garden 22.8 -0.29

Small (less than 1600) 24.0 0.05

Larger (more than 1600) 54.4 0.21

Consumer goods

Mobile phone 29.1 0.10

Radio 15.2 -0.01

Cooking utensils per spousea

4.1 (4.09)

No utensils 6.0 -0.26

1 to 3 54.4 -0.25

4 or more 37.9 0.36

Garden tools per adult 1.9

(1.06)

Less than 2 68.3 -0.22

2 or more 31.6 0.22

Sample size 79b

Notes: aWhere there is no spouse the figure includes all cooking utensils owned by the individual. b Sample size is 79 as one household refused to answer asset questions and three coffee gardens could not be measured.

39

To construct the asset index, the assets were grouped into categories based on the dispersion of assets. In most cases the boundary was around the median. The exceptions were mobile phones and radios, where a value of 1 was given if at least one item was owned and 0 otherwise. Once grouped, other categories were given a score of 1 if they applied to a household and 0 if they did not. Principal Components Analysis (PCA) was then used to construct the asset index, as recommended by Filmer and Pritchett (2001). Using PCA, each component is a linear weighted combination of the original variables (OECD 2008, p.65). The first component explains the largest amount of variation in the data and subsequent components explain additional variation (Vyas and Kumaranayake 2006, p.460). The first component is used to represent household wealth. To construct the index the factor scores generated under the first component are used as weights for ownership of each type of asset. Weighted asset scores are then summed and the resulting index is standardised to have a mean of 0 and a standard deviation of 1.

The factor scores are also shown in Table 2.6. High positive coefficients in the factor score indicate that ownership of that asset is likely to mean ownership of other assets (Howe et al. 2008). Negative coefficients indicate that ownership of that asset is correlated with owning few other assets. (Moser and Felton 2007, p.4). Thus, a factor score of -0.38 for owning no pigs indicates that households that don’t own pigs are unlikely to have many other assets. The distribution of asset index scores is shown in Figure 2.1.

40

Figure 2.1: Distribution of asset scores for study area

The distribution of scores is relatively wide and there is limited ‘clumping’ together of scores around a narrow value (McKenzie 2005, p.235; Vyas and Kumaranayake 2006, p.461). Although there are several instances of households sharing the same score, it is possible to divide the asset index scores into groups for further analysis. Summary statistics for the asset index scores split into quintiles are in Table 2.7.

41

Table 2 7: Summary statistics asset index scores in study area, by quintile

Asset index in quintiles

Number of

households Per cent

Minimum score Maximum score Mean Standard Deviation Difference in meana Lowest 16 19.3% -1.80 -1.00 -1.47 0.33 Low-mid 16 19.3% -0.92 -0.33 -0.50 0.16 0.97 Middle 16 19.3% -0.27 0.43 0.07 0.26 0.57 Mid-upper 16 19.3% 0.44 1.00 0.60 0.18 0.53 Upper 15 18.1% 1.07 1.84 1.39 0.28 0.79 Total 79

Note: aDifference in mean is defined as the absolute value of mean of the quintile above minus the mean of the current quintile.

Source: Author’s calculations.

The standard deviations of asset index scores is largest at both the upper and lower bounds of the score range. Differences in the means also show that asset scores are unevenly spread. An even distribution of assets between quintiles would lead to equality in the differences in mean between quintiles (Vyas and Kumaranayake 2006, p.464). However, column 7 in Table 2.7 shows a large difference in means between the middle- upper and upper quintile as well as between the middle and lower-middle quintiles. This suggests that the groups in middle of the asset distribution are quite similar, and different from groups at the upper and lower end of the distribution.

42

Table 2.8: Breakdown of Asset Index Score in study area, by quintile

Distribution of assets by asset index quintile

Lowest Low-mid Middle Mid-upper Upper

Livestock No pigs 100.0% 75.0% 25.0% 12.5% 6.7% Between 1 and 4 0.0% 25.0% 68.8% 87.5% 73.3% 5 or more pigs 0.0% 0.0% 6.3% 0.0% 20.0% No chickens 100.0% 75.0% 43.8% 43.8% 13.3% Between 1 and 5 0.0% 25.0% 25.0% 37.5% 46.7% 6 or more chickens 0.0% 0.0% 31.3% 18.8% 40.0%

Coffee garden size (square metres)

No coffee 56.3% 18.8% 25.0% 12.5% 0.0%

Small coffee garden (less than 1600) 12.5% 37.5% 31.3% 18.8% 20.0%

Large coffee garden (more than 1600) 31.3% 43.8% 50.0% 68.8% 80.0%

Consumer goods

Own a mobile phone 12.5% 37.5% 31.3% 50.0% 20.0%

Own a radio 12.5% 12.5% 18.8% 25.0% 6.7%

No utensils 25.0% 6.3% 0.0% 0.0% 0.0%

Between 1 and 3 68.8% 81.3% 75.0% 43.8% 0.0%

4 or more utensils 6.3% 12.5% 25.0% 56.3% 93.3%

Less than 2 tools 100.0% 50.0% 81.3% 68.8% 40.0%

2 or more tools 0.0% 50.0% 18.8% 31.3% 60.0%

Source: Author’s calculations.

Ownership of assets was analysed by quintile (Table 2.8). Livestock ownership is skewed towards the upper quintiles. 20 per cent of the households in the upper quintile own five or more pigs whereas in three out of the remaining four quintiles no households own this number of pigs.

The distribution of ownership of between 1 and 4 pigs is more even between quintiles. Although no households in the lowest quintile own any pigs, a quarter of households in the low-mid quintile own between 1 and 4 pigs. This number increases as overall asset ownership increases.

Chicken ownership is also skewed towards households with more assets. None of the households in the lowest quintile own any chickens. A lack of chicken ownership may indicate that the households concerned are too poor to enter into tradeable asset accumulation.

43

When household goods are considered, the bottom two quintiles are the only two groups with households owning no utensils. All households in the bottom quintile own less than two gardening tools per adult. The relative lack of household goods ownership in the bottom two quintiles may reflect poor access to cash via coffee production. This is particularly the case for the bottom quintile, where 56 per cent do not own coffee gardens.

More generally, coffee garden ownership appears to be a prerequisite for a high level of assets, as all the households in the top quintile own a coffee garden. However, it does not appear to be a guarantee of asset accumulation. 31 per cent of households in the bottom asset quintile had large coffee gardens. Given the uncertainty related to coffee transport and trading in the study area, ownership of a coffee garden represents potential rather than actual cash income. The ability of a household to translate this potential into actual cash income may depend on environmental factors, transport availability and human and social capital, to the extent that this helps to negotiate transport arrangements and coffee sales.