EThe effect of human resource

Transcription

1 L Volume nalmd Journal east asia Labour and Management in Development 2, Number 2 EThe effect of human resource development on household income in selected poor areas of rural China 99-2 Asia Pacific Press at the AUSTRALIAN NATIONAL UNIVERSITY

2 This work is copyright. Apart from those uses which may be permitted under the Copyright Act 1968 as amended, no part may be reproduced by any process without written permission from the publisher. ISSN ISBN X is a PhD candidate in the Population and Human Resources Program with the Faculty of Social Sciences, Flinders University, Adelaide. He is a former Associate Professor in the School of Economics, Peking University. Acknowledgment The author would like to acknowledge the useful comments and suggestions of Dr Ross Steele, Dr Alaric Maude, and three other anonymous referees. Abbreviations FAO Food and Agriculture Organisation HRD human resource development ILO International Labour Organization OLS Ordinary Least Squares UNDP United Nations Development Programme UNESCAP United Nations Economic and Social Commission for Asia and the Pacific UNESCO United Nations Educational, Scientific and Cultural Organisation UNFPA United Nations Fund for Population Activities Labour and Management in Development Journal, Volume 2, Number 2 2

3 Although there has been intensive research on the effect of human resource development (HRD) on economic development in developing countries (World Bank 1980, UNESCAP 1993, UNDP 1996), many relationships in this field are still unclear or unknown (Behrman 1990:127 30). The literature on human resource development in the Asia Pacific or its individual countries is limited (Low 1998). China is a low-income country with the largest single body of human resources in the world, but attention in China has not yet been focused on this issue, especially at the micro-level. Chinese researchers and government officials have often emphasised poor natural resources as a major cause of poverty in rural China. More recent research has shown that many poor areas of China have adequate natural resources and that the key factor responsible for the economic backwardness in these areas is the poor quality of human resources (Cao 1991; Wang and Bai 1991). This argument is, however, mainly based on case studies and is not underpinned by hard empirical evidence. Farming systems in China are changing from a socialised centrally planned basis to a system based on the household economy. Despite this, there has been little income research at the household level, mostly due to the difficulty in obtaining micro-level data. China has three major geographic regions Eastern, Central and Western and physical and socio-economic conditions are quite different in each region. Some earlier research has focused on individual poor counties of rural China (Tong 1991; Zhang 1991; Zhang 1996), but wide-ranging research is necessary to identify the differences in household income levels between regions. The main objective of this research is to identify the effect of human resource development on household income levels, explore the contribution of human resources to household income differentials and ascertain the ways in which household per capita net income can be increased. Although human resource development has a wider meaning, which refers not only to the means of development but also to the end results of development (UNDP 1990), in this paper, it mainly refers to improvements in the education, health, skills, mobility and occupation of members of the labour force. The narrower focus was adopted here to establish a workable framework for analysis as well as to concentrate on the issues that are strongly related to the determinants of household per capita net income in rural China. Both the quality and quantity of human resources will be discussed in this study, but emphasis will be on the quality of human resources. The effect of natural and capital resources on income levels will also be discussed briefly. This analysis uses data from a survey of 23 poor counties of rural China, and it enables us to analyse the effect of human resources on household income by region. Review of previous research on the effect of HRD on income Human resources have both quantitative and qualitative dimensions. The number of people, the proportion of people entering useful work, and the hours worked are essentially quantitative characteristics. The Labour and Management in Development Journal, Volume 2, Number 2 3

4 skill, knowledge and similar attributes that affect particular human capabilities to do productive work are qualitative components (Schultz 1961:8). Schultz (1971:36) claimed that health, skill and education are major human capabilities which need investment, and that migration of individuals to adjust to changing job opportunities is also a kind of human capability. Since permanent migration was not common in rural China during , temporary migration (mobility) will be considered as a human resource indicator in this study. The proportion of working time in nonagriculture is also considered as a human resource indicator; it demonstrates occupational mobility, the ability to adjust to changing job opportunities. Such occupational mobility can only be fostered with investment, but can be highly beneficial. The literature on the effects of HRD on income is quite extensive (Becker 1964; Mincer 1974; Schultz 1981). Some of it has focused on the effect of one component of human resources such as education or health, while other research has explored the relationship between human resource variables and income or studied the determinants of household income. Welch (1970) proposed that education influenced production and income, especially in the case of farmers, by enhancing the productivity of a farmer s inputs including labour by lowering the costs of acquiring and using information about production technology, and by facilitating more rapid entrepreneurial responses to changing markets and technological developments. Michael has claimed that [i]nvestments in education enhance one s earnings capacity, increase one s time value in the labour market, and raise one s full money income (1974:127). Empirical studies on China have also proved that the educational attainment levels have a highly significant effect on household income in rural areas (Nee 1989; Griffin and Zhao 1993). The World Development Report (World Bank 1993) emphasised that investment in health improved the quality of human resources and increased household income. It found that health improvements, especially those which enhanced a woman s health so she could better fulfil her role as a mother and household manager, boosted household earnings and financial security. Behrman and Deolalikar (1988:701) concluded that health status and nutritional intake had a positive impact on agricultural productivity and farmers income. In the poor areas of China, a strongly negative correlation was found between household average per capita income and the incidence of malnutrition (World Bank 1992:91). Agricultural extension assists farm people, through educational procedures in improving farming methods and techniques, increasing product efficiency and income, bettering their levels of living and lifting the social and educational standard of rural life (FAO, UNESCO and ILO 1984:1). Training enabled labourers to work more efficiently and increased the return to individuals. There was convincing evidence that participating in an agricultural extension program had a significant effect on output. Rates of return to investment in extension activities in most of the developing countries were much higher Labour and Management in Development Journal, Volume 2, Number 2 4

5 than the standard cut-off points of other viable investments, indeed they were found to be as high as 80 per cent in some Asian countries (Birkhaeuser, Evenson and Feder 1991). Higher migration rates lead to improvements in the distribution of income and to reductions in the extent of poverty (Stark 1978:90). Griffin (1976) argued that urban-to-rural remittances played an important role in rural development, and that internal migration would improve the distribution of income in rural areas and accelerate capital formation and technical change on small peasant farms. Maude (1981) suggested, based on his research in Malaysia, that income resulting from mobility was a major contributor to the economy of some households and a useful supplement for many others, but, for the rural economy as a whole, the net benefits resulting from mobility were relatively small. In rural China, mobility greatly increased farmers incomes, enhanced the quality of labour, and indirectly reduced the level of fertility; but it also caused some rural households to have a shortage of quality labourers for local crop production (Zhang 1991, 1996). Non-agricultural employment has been found to be a major cause of rising agricultural real wage rates in underdeveloped areas in India and other developing countries (Unni 1998). According to Taiwan s experience (Chinn 1979), allocating family labour to non-farm sectors and earning income from non-farm activities can raise the real standard of living in farm households, and non-farm income sources have played an important role in reducing income inequality in rural areas. In China, the post-1978 reform of agricultural marketing was a decisive shift. Farmer s occupations were no longer limited to agriculture. Instead, they could diversify into other production and business activities. There was an increase in the proportion of household income derived from non-agricultural sources (Watson 1988; Nee 1989). The rural non-farm sector could then play a major role in bettering the living standards of rural populations, especially where adverse land person ratios have imposed significant constraints on agriculture growth (Mukhopadhyay 1990:26). There is considerable evidence of a strongly negative correlation between household size and income per person in developing countries (Lipton and Ravallion 1995). As Lipton has noted birth prevention, especially in rural areas with scarce land, may well yield more income per head than any other investment, especially since it permits the family to save and invest, instead of using its substance to feed more children (1980:85). Household income per capita is inversely correlated with the size of household. That is, smaller families tend to have a higher standard of living than larger ones (Griffin and Zhao 1993:12). But, if the household size is too small, the household can have too few workers. The implementation of the household responsibility system in the countryside in the 1980s increased the importance of having a large number of potential workers in the household. The level of mechanisation in agriculture was still very low in the underdeveloped regions, and many farming Labour and Management in Development Journal, Volume 2, Number 2 5

6 jobs had to be done by hand instead of machine. As a result, the number of workers was decisive for rural households production, and was a key determinant of household income (Khan 1993:111). Non-human factors such as capital, land and location also affected household income. In order to explore the separate effects of human resource development, many researchers have studied the determinants of household income with the aid of multiple regression methods. For example, Hossain and Sen (1992) explored the relationship between household income and other input variables in a survey of 1,112 households in rural Bangladesh. The results showed that the size of land owned by the household, the level of educational attainment, and the number of workers in the household all had a significant effect on rural household income. Lack of land was a major disadvantage for the rural poor, but the diffusion of new technology in agriculture possibly acted as a compensatory mechanism for limited land resources by increasing land productivity. Khan (1993) constructed household income models for rural China from a survey of 10,258 rural households in 28 Chinese provinces in 1988, which covered both developed and poor areas. He claimed that the net income of a household was correlated with the industry in which the labourer was employed, level of educational attainment, whether or not the family was connected with the communist party, value of productive capital, land area, the proportion of cultivated land which is irrigated, use of chemical fertiliser and the degree of market orientation of the household s productive activities. He also emphasised the effect of provincial dummy variables on the dependent variable, and noted that sometimes the provincial dummy variable could change the sign or magnitude of the regression coefficient. Khan developed useful models to predict household income levels in rural China, but the level of human resource development was not his specific focus. It seems that different researchers have found different results regarding the determinants of household income in rural China, often because they conducted their research in different regions and based on different sample sizes (Khan 1993; Meng and Wu 1995; Wu, Richardson and Travers 1996). It was common, however, to find that educational level had a positive effect on income level. Although there was plenty of international and national research on the effect of education on household income, there were very few empirical studies of the combined effect of human resources such as education, health, skill, mobility and occupation on household income. The inclusion of most human resource variables in a model enables us to identify the effect of human resources on household income more accurately. Data and methodology The Survey of Demographic and Economic Development in 23 Poor Counties in Rural China provided a large set of individual and household level data. It was conducted at the beginning of 1990 by the Institute of Population Research, Peking University, and supported financially by the United Nations Fund for Population Activities (UNFPA). Twenty-one institutes of population research Labour and Management in Development Journal, Volume 2, Number 2 6

7 in China, and 23 Family Planning Commissions, were also involved in the project. The survey, conducted in 1989, contains social and economic information on about 23,000 households or 100,000 individuals and is a cross-sectional data set. The 23 counties were selected from 699 counties designated as poor by the central and provincial governments. 1 About 1000 households were assessed in each county in the survey. Stratified random sampling with a proportional three-stage allocation was adopted in the survey. The three sampling stages comprised poor townships, poor villages and households, and the three stratified groups poor, intermediate and well-off households were drawn randomly from the sampled villages in the proportions of 5:3:2. In order to ensure the quality of the survey, the data were gathered from the surveyed respondents in face-to-face interviews. The questionnaire consisted of about sixty questions on individual and household socioeconomic situation in , which covered the household s income and expenditure, human resources, land and capital resources, population and family planning, and people s opinions on poverty alleviation. Generally speaking, the quality of the survey data was good and the sampling was representative. There were, however, some unavoidable errors and biases in the survey. After logical examination of each variable used in this study and removal of extreme cases in accordance with common practice, the real effective sample size in this study was 20,453 households (further details of this survey can be found in Zhang (1992)). Household per capita net income (measured in yuan/person) is treated as a dependent variable in this paper and is defined as gross income minus expenditure on production divided by household size. Gross income includes cash income from the sale of products, gross value of selfconsumption of farm products, income from wages, pensions and other compensations, subsidies from government or collective units, and other income. This measure of per capita net income is the most common and authoritative indicator of a household s income level and capability to invest in production and to change its living conditions. Independent variables in our empirical models are listed in Table 1. They include five human resource quality variables, two human resource quantity variables, two natural (land) resource variables, three capital resource variables and two location dummy variables. Although bivariate analysis is utilised in this study, multiple linear regression is our main research method. We will denote I as the household per capita income, and X i as the ith independent variable. The multiple linear regression model is I = b 0 + b 1 X 1 + b 2 X b k X k + m The unknown regression coefficient b i can be estimated by the method of ordinary least squares (OLS), as well as the t-statistic, R-squared coefficient and F-statistic (Gujarati 1988). Using this model, we can identify which independent variable has a significant influence on the dependent variable by examining the t-ratio for the coefficient. Labour and Management in Development Journal, Volume 2, Number 2 7

8 Table 1 Main independent household variables used in the empirical models Variable Natural (Land) resources Arable land Hilly and forest areas Capital resources Current input Description Arable land areas rented by a household. Unit: mu. Hilly and forest areas rented by a household. Unit: mu. Annual current capital input, including the expenditure on fertilisers, seeds, raw materials and others. Unit: 100 yuan. Machinery Weighted amount of machinery. A tractor is given a weight of 4 compared to the other types of machinery such as carts, pumps, ploughs and others. Unit: number. Livestock Quantity of human resources Household size Workers Weighted number of livestock. Cattle, horse, mule and donkey are the standard units, and pig and sheep are given a weight of Unit: number. Number of people living in a household. Unit: persons. Number of people aged 15+ working at least 90 days out of one year. Unit: persons. Quality of human resources 3 5 years of education 1 if all workers had an average 3 5 years of formal education, 0 otherwise. 6 8 years of education 1 if all workers had an average 6 8 years of formal education, 0 otherwise. 9+ years of education 1 if all workers had an average 9+ years of formal education, 0 otherwise. Health status Skill status 1 if all workers were in good health, 0 otherwise. 1 if any worker had some special skills except for growing field crops, 0 otherwise. 1 6 moves 1 if the workers had a total of 1 6 moves for economic purposes to other villages during , 0 otherwise moves 1 if the workers had a total of 7 20 moves for economic purposes to other villages during , 0 otherwise. 21+ moves 1 if the workers had a total of 21+ moves for economic purposes to other villages during , 0 otherwise. 1 20% in non-agriculture 1 if the workers spent 1 20% of total working time on non-agriculture, 0 otherwise % in non-agriculture 1 if the workers spent 21 50% of total working time on nonagriculture, 0 otherwise. 51+% in non-agriculture 1 if the workers spent 51+% of total working time on non-agriculture, 0 otherwise. Location dummies East Centre Selected county 1 if a household was in the Eastern Region, 0 otherwise. 1 if a household was in the Central Region, 0 otherwise. 1 if a household was in the selected county, 0 otherwise. Notes: a. Mu is the unit of land area in China, and it is approximately equal to 1/15 hectare. b. Yuan is the unit of Chinese currency, Renmibi. Official foreign currency exchange rate: US$1 was equal to 3.7 yuan during and 4.7 yuan in 1990 (State Statistical Bureau 1995). Labour and Management in Development Journal, Volume 2, Number 2 8

9 There are two ways to assess the relative importance of independent variables (Norusis and SPSS 1993:341 43). One is to calculate the beta coefficients, and another way is to consider the increase in R 2 when a variable is entered into an equation that already contains the other independent variables. The increase is R 2 change = R2 - R 2 (i) where R 2 is the square of the multiple (i) correlation coefficient when all independent variables except the ith are in the equation. A large change in R 2 indicates that a variable provides unique information about the dependent variable that was not available from the other independent variables in the equation. This method will be employed in our research to compare the importance of land, capital and human resources for household income. We will assign relative importance to a group of variables instead of one by observing the change in the value of R 2. Household per capita net income by level of human resources The average household per capita net income in our survey was 356 yuan, much lower than the overall average for rural China in 1989 (602 yuan). This is because this survey focused on poor areas of rural China. Household income levels varied across different regions. Net per capita income ranged from 464 yuan in the Eastern Region to 354 yuan in the Central Region and 277 yuan in the Western Region. In the Eastern Region, 24.7 per cent of households had a per capita net income of more than 600 yuan, but 19.7 per cent of households had an income lower than 200 yuan. Of the households in the Western Region, 45.5 per cent had a per capita income lower than 200 yuan and 7.6 per cent had a per capita income above 600 yuan. Clearly, substantial inequalities of household income existed even within the 23 poor counties. Poor resource endowment is obviously a factor in the lower average incomes in some regions, but some observers have argued that per capita income differences can be better explained by differences in the endowments of human rather than physical capital (Psacharopoulos 1973:17). Income differentials by quality of human resources The mean household per capita net income differed greatly according to level of workers education (see Table 2). If all respondents are considered together, the per capita net income increased from yuan in households averaging 0 2 years of education per worker, to yuan in those averaging 3 5 years per worker, yuan in those averaging 6 8 years per worker and yuan in households which had an average of nine or more years of education per worker. Except for households averaging nine or more years of education per worker in the Central Region, household groups with more years of education consistently had a higher per capita net income. The education variable caused greater income differentials in the Eastern Region than in other regions. Households averaging nine or more years of education per worker in the East had the highest per capita net income of yuan, 90 times higher than those with 0 2 years of education. Per capita net income in the Central Region was lower in Labour and Management in Development Journal, Volume 2, Number 2 9

10 Table 2 Differentials of household per capita net income by quality of human resources and by region (yuan per person) East Centre West Total Years of formal education 0 2 years (100) (100) (100) (100) 3 5 years (125) (127) (116) (126) 6 8 years (152) (147) (131) (155) 9+ years (190) (143) (157) (182) Any unhealthy workers No (100) (100) (100) (100) Yes ( 76) ( 74) ( 82) ( 76) Any skilled workers No (100) (100) (100) (100) Yes (144) (128) (147) (142) Frequency of mobility for economic purposes None (100) (100) (100) (100) 1 6 moves (119) (114) (110) (119) 7 20 moves (156) (138) (134) (147) 21+ moves (177) (147) (142) (159) Proportion of working time in non-agriculture 0 % (100) (100) (100) (100) 1 20 % (121) (110) (115) (112) % (147) (128) (128) (138) 51+ % (192) (150) (156) (184) Number of cases 5,347 8,876 6,230 20,453 Note: Indexes of income based on the income of the first category of each variable in each region are in parentheses, with the base value being set at 100. Source: Survey of Demographic and Economic Development in 23 Poor Counties of Rural China each educational category than in the Eastern Region, but it was higher than in the Western Region. The income of households with nine or more years of education was marginally lower on average than households with 6 8 years of education in the Central Region households, which indicates that more years of education do not always bring households a higher income. Per capita net income in the Western Region households with 0 2 years of education per worker was yuan, the lowest in all educational categories. However, the income of Western Region households with nine or more years of education per worker was yuan higher than that of Eastern and Central households in the 0 2 years of education category. So, not all Western households were poorer than households in other regions, and some would arguably be better off if human resources were better developed in this region. Overall the presence of unhealthy workers in a household resulted in a 24 per cent reduction in household per capita net income compared to a household without any unhealthy workers. The presence of unhealthy workers in the household resulted Labour and Management in Development Journal, Volume 2, Number 2 10

11 in a 24 per cent decrease in income in the Eastern Region and 26 per cent decrease in the Central Region. The per capita net income of Western Region households with unhealthy workers was only yuan the lowest of all categories and 18 per cent lower than in the Western Region households without any unhealthy workers. The presence of skilled workers resulted in a substantial difference of per capita net income. Overall, the households with skilled workers had a 42 per cent higher per capita net income than households without any skilled worker. In the Eastern and Western Regions, the income differentials caused by different skill status were more obvious, with the indices of income at 144 and 147. In the Central Region, the income increase in households with skilled workers was the smallest, but income in such households was still 28 per cent higher than that in households without any skilled worker. The more frequently households moved for economic reasons, the higher their household per capita net income. The mean per capita net income of all households without any history of mobility in the previous three years was yuan, which rose to yuan for all households with 21 or more moves, an increase of 59 per cent. In the Eastern Region, the households who had moved 21 or more times had the highest per capita net income of all categories at yuan, 77 per cent higher than households with no moves. In the Central and Western Regions, the increase in income was not as high as in the Eastern Region, but households who had moved 21 or more times still had higher incomes than those who had not those in the Central Region received an increase in 47 per cent and those in the Western Region received 42 per cent increase. Per capita net income also increased as the share of time spent in non-agricultural work increased. Income in households where no time was spent in non-agricultural work was yuan, but it reached yuan in all the households where 51 per cent or more of working time was spent in nonagricultural work. In the East, households in the highest category of non-agricultural work had an income of yuan, 92 per cent more than those in the reference category, the largest gain in income of any categories. In the Central Region in the per capita net income gain was smaller than in the East, but the differentials were still quite substantial. Per capita net income in households that spent more than 50 per cent of their time in non-agricultural work was 50 per cent higher than in households without non-agricultural work. In the Western Region, the income indices were marginally higher than in the Central Region. In the Western Region, those households that spent the most time in nonagricultural activities had an income of yuan, 1.56 times more than those that did not participate in non-agricultural work. To undertake non-agricultural activities, workers needed to be skilled and be able to access more market information. Nonagricultural work, however, is generally characterised by higher input output ratios and higher income. Processing of agricultural products transportation, and construction were the most popular nonagricultural industries of rural workers in our survey. Labour and Management in Development Journal, Volume 2, Number 2 11

12 Table 3 Differentials of household per capita net income by quantity of human resources, by region (yuan per person) East Centre West Total Number of workers 1 worker (100) (100) (100) (100) 2 workers (124) (115) ( 99) (110) 3 workers (127) (123) (109) (116) 4 workers (135) (131) (117) (123) 5+ workers (134) (123) (113) (114) Household size 1 person (100) (100) (100) (100) 2 persons (108) (117) ( 92) (105) 3 persons (114) (115) ( 88) (105) 4 persons (110) (111) ( 92) (103) 5 persons (104) ( 99) ( 88) ( 94) 6 persons (101) ( 96) ( 89) ( 92) 7+ persons ( 99) ( 92) ( 91) ( 85) Number of cases 5,347 8,876 6,230 20,453 Note: Indexes of income based on the income of one worker or one person in each region are in parentheses, with the base value being set at 100. Source: Survey of Demographic and Economic Development in 23 Poor Counties of Rural China Income differentials by quantity of human resources The number of workers was an important quantitative measure of human resources which appeared to affect per capita net income, but the direction of the relationship was not constant (see Table 3). Most households in our survey had two workers, and the 3-worker and one-worker households were the second and third largest groups. Overall, one-worker households had the lowest per capita net income yuan and per capita net income rose as the number of workers increased until it reached a maximum of yuan in 4-worker households. Having more than four workers in a household did not always result in higher per capita net income. A similar situation was also found in each region, but there was only a marginal decrease in income in households with five or more workers compared to 4-worker households in the Eastern Region. The indices of income by number of workers were greater in the Eastern than in the Western Region. So having a certain number of workers was more important for increasing income levels in Eastern households than those in other regions, because there were more employment opportunities with a higher salary rate in the Eastern Region (for further details on the differences between the three regions, see Wang (1994)). The change in per capita net income according to household size was not consistently in the one direction. Overall, it peaked at yuan in 2-person households and then gradually dropped to a low of Labour and Management in Development Journal, Volume 2, Number 2 12

13 yuan in households with seven or more members. The change of income was also not the same in each region. In the East, income was highest in 3-person households, but it was highest in 2-person households in the Central Region and in one-person households in the Western Region. Per capita net income in one and two-person households in the Central and Western Regions appeared higher than other households since these households usually had only one person if any, at all, who was not working (since the observations of households without any workers have been removed from our effective sample, the person in one-person households must be a worker, and there was at least one worker in two-person households). According to our survey, however, such households were very rare, accounting for 8.3 per cent and 7.5 of all households in the Central and Western Regions, and they were not representative. The lowest income occurred in households with more than six people in all regions except in the West. The change of income by household size in the Western Region was not regular. In the East and Centre, except in the case of one and twoperson households, it seems that per capita net income fell as household size increased. Generally, the differentials of per capita net income by household size were not very substantial. The largest difference of income between two household groups was only 85.8 yuan in the overall data and 60.8 yuan, 82.0 yuan, and 37.5 yuan in the Eastern, Central and Western Regions respectively (Table 2). Modelling the effects of human resources on household per capita net income The bivariate analysis used in the previous section is not able to identify the separate effect of each human resource variable on income level because it cannot isolate the effect of the other independent variables. Therefore, multivariate regression analysis is necessary. It is also essential to include other factors such as natural resources, capital and regional variables into the model when an attempt is made to predict the effect of the level of human resource development on household per capita net income because these variables also have an effect on the dependent variables. Although most researchers use multivariate regression models to explore the determinants of household income, the models to be developed in this section are different from others in some respects. First, our models are based on a household sample survey of poor areas of rural China, and the focus of this research is the characteristics and relationships among the relevant variables in the households of the poor areas. Second, land and capital resource variables are considered in our models, but human resources remain the focus. Third, the survey of 23 poor counties had a sample size of about 20,000 households in 22 provinces of China. It therefore enables us to construct models by region, thus deriving results that are more representative of poor areas in rural China. Table 4 gives the regression coefficients of the independent variables, standard errors and their level of statistical Labour and Management in Development Journal, Volume 2, Number 2 13

14 significance. F-values indicate that all four models are significant at the level and R-squared values indicate that independent variables can statistically explain per cent of the variation in the values of the dependent variables. For models with such large numbers of observations, these results are quite good. In the total sample model and the regional models, except for the categories of 1 6 moves and 1 20 per cent of working time in non-agricultural activities, the regression coefficients of all categories of the five qualitative variables of human resource development were statistically significant. It was found that the households whose workers had more than eight years of education could enjoy yuan more net income per capita than those having 0 2 years of education (the reference group). Households in the 3 5 years and 6 8 years of education categories also had 24.3 yuan and 65.2 yuan higher net income than the reference category. Households whose workers were all in good health enjoyed 67.9 yuan more net income per capita than those households with someone in poor health. The presence of at least one skilled worker in a household resulted in a 50.7 yuan increase in per capita net income compared to a household without any skilled workers. Per capita net income in households with a small number of moves for economic purposes was not significantly different from that in households where there was no such move, but 7 20 moves and more than 20 moves resulted in an increase in the per capita net income of 65.1 yuan and 76.4 yuan in households in the total sample model. Households who spent more than 50 per cent of their working time in nonagricultural work enjoyed a 57.0 yuan increment to per capita net income, but the income of households spending 1 20 per cent of their time on non-agricultural work did not differ from that in households without non-agricultural work. The categories of households spending more than 50 per cent of their working time in non-agricultural work and households having an average of 9+ years of education had the largest and the second largest regression coefficients in each model. The regression coefficients of all the significant human resource variables were much greater in the model of the Eastern Region than in models of other regions. The number of workers was positively related to the dependent variable in all models, but it appeared more important as a predictor in the model of the Eastern Region. In the total sample, an increase of one additional worker was associated with an average rise of 30.1 yuan per capita net income. In the Eastern model, the comparable income increase was 47.2 yuan. Household size had a strong negative influence on per capita net income if other independent variables including the number of workers were held constant. An additional person in the household reduced per capita net income by 39.6 yuan in the total sample model. Arable land was negatively related to per capita net income in the total sample model. This, however, was not the case in the regional models. Arable land had no significant effect on income in the model of the Western Region, but it had a positive effect on income in the Eastern and Central Labour and Management in Development Journal, Volume 2, Number 2 14

15 Table 4 Results of multiple linear regression analysis of household per capita net income by region (regression coefficients) Independent variable Eastern Central Western Total Region Region Region Sample model model model model Arable land (mu) (1.590) a (0.419) a (0.437) (0.223) a Hilly and forest areas (mu) (0.526) (0.566) (0.395) c (0.270) a Current input (100 yuan) (1.172) a (0.824) a (0.967) a (0.537) a Machinery (units) (3.674) a (2.218) a (1.701) a (1.358) a Livestock (units) (0.561) a (0.328) a (0.289) a (0.216) a Workers (persons) (4.912) a (3.377) a (2.959) a (2.176) a Household size (persons) (3.813) a (2.626) a (2.115) a (1.592) a 3 5 years of education (11.292) a (6.567) a (6.058) a (4.364) a 6 8 years of education (12.721) a (7.556) a (8.129) a (5.039) a 9+ years of education (18.625) a (13.126) a (19.398) a (9.185) a Good health (11.511) a (7.193) a (7.170) a (4.918) a Skilled workers (9.890) a (6.774) a (6.731) a (4.560) a 1 6 moves (10.967) (7.479) (7.699) (5.024) 7 20 moves (14.274) a (11.062) a (9.381) b (6.795) a 21+ moves (22.501) a (18.840) b (14.388) c (11.030) a 1 20% in non-agriculture (13.593) (8.315) (7.481) c (5.465) 21 50% in non-agriculture (11.442) a (7.812) a (7.898) a (5.128) a 51+% in non-agriculture (12.926) a (10.320) a (12.667) a (6.713) a East (5.312) a Centre (4.544) a Fengcheng (14.836) a Panan (15.194) a Fuding (15.253) a Sishui (14.541) c Longchuan (15.055) a Helin (13.376) a Tongyu (13.315) c Qinggang (14.639) a Yingshang (12.066) a Liuan (12.090) a Nankang (12.566) b Shangcheng (12.553) Zigui (12.102) b Mayang (12.450) a Xuanhan (9.993) Luodian (11.300) a Luchun (11.604) a Yichuan (10.903) a Dingxi (13.297) a Shufu (11.619) a Constant (20.195) a (13.700) a (12.847) a (8.375) a R-squared F-value a a 72.6 a a Number of observations 5,347 8,876 6,230 20,453 Note: The standard errors are reported in parentheses. Significance level: a p<0.001, b p<0.01, c p<0.05. Source: Survey of Demographic and Economic Development in 23 Poor Counties in Rural China Labour and Management in Development Journal, Volume 2, Number 2 15

16 Regions, implying that, in these two regions, the more land the households had, the more per capita net income they earned. The explanation of the difference between the results of the total sample model and the results of the individual regional models was that most counties with a higher level of income had less arable land but more of their income was derived from non-agricultural industries. When the county dummy variables were not included in the total sample model, the relationship between arable land and income was negative. When the county dummy variables were included in the regional models, however, the sign and magnitude of the coefficients of arable land changed. A similar phenomenon was observed in research on household income in rural China (Khan 1993:106). The effect of arable land on per capita net income was not substantial, and adding one mu may only increase the income by 5.6 yuan in the East and 1.9 yuan in the Central Region. The hilly and forest land variable was positively related to the dependent variable in the total sample model, and it was also significant in the Western model, which may be explained by the larger size of the livestock industry in the West; but the regression coefficients were also small and the effect of hilly and forest land on income was limited. The hilly and forest land variable was not significant in the Eastern and Central models. All the capital input variables had a highly positive association with per capita net income. In the total sample model, if the current capital input increased by 100 yuan, the household per capita net income would increase by 12.7 yuan. Each additional unit of machinery and livestock led to an income increment of 23.5 yuan and 16.8 yuan respectively. But the effect of each capital resource variable on income differed in magnitude in the various regional models. In the Eastern and Central models, the machinery variable had a larger coefficient than the other two capital variables, but the opposite applied in the Western model. Regional dummy variables were very significant in the total sample model, and households in the Eastern Region had yuan more per capita net income than households in the Western Region (reference group) on average due to the regional differences alone. The households in the Central Region had 74.9 yuan more per capita net income than those in the Western Region. The coefficients of county dummy variables indicate the differences in mean household per capita net income between the selected county and the reference county (Wanxian, Daixian and Longsheng were the reference counties in the Eastern, Central and Western regions, respectively). Except for two counties, all county dummy variables were statistically significant. The regression coefficients of county dummy variables were usually larger than most of the micro independent variables, because the county dummy variables acted as a surrogate for more than one factor related to the average income level of a county, such as quality of land, infrastructure conditions, and geographic location. Assessing the relative importance of human resources to household per capita net income The results of the regression models show that most of the human resources, capital resources, land resources and regional Labour and Management in Development Journal, Volume 2, Number 2 16

17 dummy variables had significant effects on household per capita net income. We do not know yet, however, what kind of resource was the most important to per capita net income. Since each variable in our empirical analysis can only represent one aspect of a group of resources, it was not valid to use one individual variable to measure the importance of human resources compared to capital or land resources. A composite index of human resources could be helpful in comparing the relative importance of human, capital and land resources, nevertheless it is difficult to construct an index which comprises several different human resource indicators without causing distortions. There has been little empirical research on the importance of human resources to household income in rural China, but international research on this issue has suggested that human resource development contributes significantly to the level of personal income. There is increasing evidence and recognition that what matters for development, more than natural resources and man-made physical capital, is the capability of people to be effective and productive economic agents in short, human capital. In the particular case of agriculture, most studies on the subject establish that the education and skills of agricultural people are significant factors in explaining the inter-farm and inter-country differences in agricultural performance, along with the more conventional factors such as availabilities of land and water resources, inputs and credit (Alexandratos 1995). It is difficult to decide whether human, capital or land resources are the most important determinant of household income in an absolute sense. When it is said that one group of resources is more important than others, it refers to their relative impact in certain conditions. It is unreasonable to deny the function of capital and land resources in rural economic development. In our models, capital resource variables were very significant in determining household per capita net income; though one additional unit had a smaller influence on the dependent variable than human resource variables, their beta coefficients were usually bigger. Although land resource variables were not significant in some models land was still an indispensable resource for most households in rural areas. Human resources skill and knowledge in the application of land and capital were crucial for realising the potential of these resources. The function of land and capital was limited, however, since both need human beings to utilise their capacity in order for output to be increased. If people cannot make full use of land and capital resources due to a lack of knowledge or ability, then their potential impact would remain unrealised. It is not appropriate to use regression coefficients as indicators of the relative importance of variables due to the different units in which the variables are measured. Although the Beta coefficients are more comparable, they are also affected by the correlations of the independent variables and do not in any absolute sense reflect the importance of the various independent variables (Norusis and SPSS 1993:342). Moreover, this study focuses on the effect of human resources a collective term that includes several variables. Therefore, we will assess the relative importance of human Labour and Management in Development Journal, Volume 2, Number 2 17

18 resources by the increase of R 2, as discussed previously. Table 5 shows the overall contribution of each kind of resource to the R 2 values in the models. Land resources made the smallest contribution in all models, and their R 2 changes were not in excess of 0.005, so they were less important than the other resources. The contribution of land resources to per capita net income was the greatest in the East, where land resources were scarce, and least in the West, where total land areas were adequate. The capital resources had a much larger R 2 change than land resources in all models, at in the total sample model, 0.042, and in the Eastern, Central and Western models. Qualitative human resource variables contributed much more to increased R 2 values than capital resources in the Eastern and total sample models, with R 2 changes of and 0.079, but less than capital resources in the Central and Western models at and If, however, the quantitative variables of human resources number of workers and household size are included, the R 2 changes in the Central and Western models increase to and larger indeed than the changes due to capital resources. From these findings, we can conclude that human resources, if they include both the quality and quantity aspects of human resources, are more important in determining household per capita net income than capital resources in poor areas of rural China. Comparing the contributions of the quality and quantity aspects of human resources, it was found that the quality of human resources contributed much more to R 2 increases than the quantity of human resources in all models. This was especially pronounced in the Eastern Region, where the R 2 increase due to the quality of human resources was about three times that due to the quantity of human resources. Among the five qualitative variables of human resources, however, the core three education, health and skill had a smaller contribution to R 2 change than the other two mobility for economic purposes and proportion of working time in nonagriculture especially in the Eastern Region. So, although human capital in a traditional sense was still essential to enhance household income, the other human factors related to the labour market and the market for agricultural production played a more important role in determining household income levels in rural China. Location dummy variables had a R 2 change of in the total sample, a quite substantial contribution though it was smaller than human and capital resources. Since geographic conditions varied greatly in different counties in the West, location dummy variables contributed relatively more in the Western model with an R 2 change of 0.062, a larger change than that due to capital resources. Conclusion Although the relationship between HRD and economic development has been studied intensively, further research is necessary at the micro level. This study demonstrates that, in the case of rural China, in the same or in similar external environments, the quality of human resources in a household was an important determinant of household income levels. Labour and Management in Development Journal, Volume 2, Number 2 18