PaperFine · 英文论文定稿参考

英文论文范文

Rattan Weaving Export Stalls in Eastern Guangdong: Propensity Score Matching of Factor Allocation

更新于 2026-10-26 · International Trade · 专科 · 对照章节结构与定稿表达

本篇目录 19 节 · 展开 ▾

Chapter 1 Introduction (~900 words)

1.1 Research Background

Eastern Guangdong, consisting of Jieyang, Chaozhou and Shantou, has long been one of the most representative rattan weaving production areas in China. Due to the warm and humid climate here, there are plenty of rattan resources; thus, the craft has been passed down for many generations and various weaving methods have been developed through apprenticeship and family practice. Production is still mostly based on household stalls and small workshops: most of the labour is provided by members of the family, the workshop also serves as their living place, and the products are collected by local traders or export agents. The decentralised, family-based production mode has low fixed costs and retains the original craft tradition; however, it also means that a single stall lacks the necessary resources to develop new technology, obtain information about the market, and adjust the product mix according to changes in demand.
In recent years, the traditional cluster has been under great pressure to export. Because supply chains have been disrupted and transport costs have risen, the price of raw rattan has gone up, and the thin profit margins of labour-intensive handicrafts have been squeezed. With the improvement in labour market conditions in the surrounding areas of the Pearl River Delta, local wages have increased, and the opportunity cost of household labour has also risen. At the same time, competition has been strengthened from two directions: on the one hand, low-cost alternatives such as plastic and synthetic fibre furniture have reduced the demand for genuine rattan products in price-sensitive markets; on the other hand, Southeast Asian producers located near the raw rattan sources in Indonesia, Vietnam and the Philippines have offered similar quality goods at a lower cost. After the pandemic, the demand in overseas markets has fluctuated wildly; orders have been unstable, and buyers now prefer suppliers that can design new products and deliver them on time. These forces have all had an impact on the export stalls of rattan weaving households; some stalls have stopped exporting and switched to low-end domestic sales, and others have only been exporting at a low volume and with little profit.
The main problem of this period of stagnation is the issue of resource distribution. There is a lack of clarity about whether the endowments of the stalls and how they use labor, raw materials, working capital, and design skills can meet the demands of the foreign market. Export markets usually need stable quality, differentiated design and low MOQ, so a certain amount of skilled weaving labour should be added, materials should be purchased regularly in small quantities, and there should be sufficient but not excessive capital. The reason why the stalls in the export market are not competitive is that they have not allocated their factors reasonably; for instance, they may have used expensive raw materials for low-value designs, but there is no other reason for this. On the other hand, export participation itself is a choice and learning channel; thus, when stalls enter the export market, there will be greater motivation and clearer reasons to reallocate their resources to places with higher productivity, enhance staff skills, improve material use efficiency and develop design capabilities.
Therefore, to figure out whether the problems of the cluster are due to a misallocation of factors in the export stalls and whether export participation can promote a better distribution of factors, we need to study these issues at the stall and workshop level.

1.2 Research Objectives and Significance

There are three research purposes in this paper. First, it will investigate whether the participation of exporters in the eastern Guangdong rattan weaving cluster has altered the distribution of factors, that is, to what extent changes have occurred in labour input, capital equipment, working capital, and the proportion of skilled workers between export-oriented and domestic-market stalls. The above four dimensions are closely related to the main choice of resource by household craft producers: the allocation of labour time for weaving and marketing; the amount of investment in tools and workshops; the liquidity available for purchasing raw materials; and the accumulation of weaving expertise among family members. Based on these differences, the paper will establish an empirical reference for researching the resource behaviour of small-scale craft producers in international trade. Second, the study will examine the causal relationship between export participation and factor allocation by employing the propensity score matching method, which is intended to solve the problem of self-selection bias. As a rule, more productive stalls will be the first to enter the export market, so a simple comparison of exporters and non-exporters will often fail to distinguish the effects of export from the inherent differences in capacity; for example, better-equipped stalls are more likely to export because they have more equipment, not because they have grown better due to export. By matching exporters with non-exporters that are observationally similar, we can determine the actual reasons for participation in exports and improve the reliability of the estimated effect. Third, based on the heterogeneity of export effects among various types of stalls (i.e., differences in size, form of ownership, etc.), the paper will provide some suggestions for improving factor allocation for different groups of producers.
The study's significance is both theoretical and practical. In terms of theory, the research will expand the micro-econometric study of export effects to informal household-based crafts producers; these people are a group that has received little attention in the trade literature and mostly study manufacturing enterprises. Therefore, it can be extended to small stalls by applying self-selection theory and heterogeneous-firm trade theory, and it will be tested whether the mechanisms that work for registered enterprises also apply to household production under informal institutions. Theoretically, the above will be supplemented; practically, it will provide some suggestions for three groups of people. The stall owners can know whether to increase or reduce their employment of labour and capital, as well as whether to raise their skills through these results. Based on the identified differences in the rattan-weaving cooperatives, different support services can be planned and member stalls can be matched with suitable market channels. Based on the above reasons, the author proposes that the training, credit, and market information services of the above-mentioned institutions should be based on verified causal evidence rather than anecdotes, to improve the accuracy and efficiency of policy support for traditional handicraft exports.

Chapter 2 Conceptual Framework, Literature Review and Research Design (~2100 words)

2.1 Core Concepts and Theoretical Basis

The three specific concepts of the empirical analysis of rattan-weaving export stalls in Eastern Guangdong were selected as the basis for this study. "Rattan-weaving export stalls" is a small production unit; it is a family-based workshop with fewer than ten workers, selling finished rattan products to export intermediaries or receiving orders from export supply chains. They are different from registered firms in that they do not have independent legal status and lack formal bookkeeping, and they are supported by family labour; thus, they form the operational level of the local export economy. Second, "factor allocation" refers to the distribution and intensity of the four kinds of factors: labour hours, raw rattan purchases, capital equipment such as weaving frames and treatment facilities, and design or craft skills. The choice of allocation is in two areas: products and market channels; that is to say, export-oriented stalls will distribute all inputs to be standardised and large-scale, and domestic-oriented stalls will be diversified in decoration and customisation. The third is "export participation", a situation in which a company participates in export activities at any time during the observation period, irrespective of the amount or duration; thus, it is a "binary" form of export, not an intensity of exports.
A variety of theories are used to guide the study. Melitz's heterogeneous-firm export theory states that only firms with good productivity can afford to bear the fixed export costs; therefore, firms that are more productive will export more widely. The idea that self-selection will be stronger for better-performing exporters was proposed by Clerides, Lach and Tybout; thus, the difference in productivity is not the cause of higher exports. On the other hand, the export learning effect is that through participation itself, the technology, quality standards, and design concepts can be spread, leading to the reallocation of factors to more productive uses. Based on the production function model, it can be inferred that the choice of input is determined by the difference in expected returns in the various markets for exports and domestic sales. Since the choice to export is endogenous and influenced by unobserved abilities, a simple comparison of exporters and non-exporters may be biased. Based on the above reasons, propensity score matching was used to select a relatively good control group for the treatment group in this quasi-experiment; thus, it can be estimated how export participation affects the factors of production in rattan-weaving stalls.

2.2 Data Source and Variable Measurement

The empirical analysis is based on a special field survey of rattan-weaving stalls carried out in eastern Guangdong from March 2023 to February 2024, mainly in Jieyang and Chaozhou county-level areas where the craft is concentrated. Given that the weaving stalls are located in the areas designated for the production of crafts, the method of clustered random sampling by town was used; in other words, the craft association at the town level and the market registration of the stalls were first sorted out, and then a random sample of the stalls was drawn in proportion to the number of the above information. The questionnaire included five parts: stall characteristics (year of establishment, registration status, size of workshop); export status (share of export orders, main destination channels); labour use (number of workers, days worked, proportion of skilled weavers); input expenditure (raw rattan and accessory material costs, equipment purchase value); and operator attributes (age, education, years of weaving experience, initial capital). The last complete business year's annual output value in yuan was kept. A total of 512 questionnaires were distributed, 468 were returned, and 436 were valid; the valid rate was 85.2%. Two rules were used to clean the data: stall with more than three missing critical items were removed, and isolated missing values were filled with the town's median; the observed values outside the range of output value and labour input within three standard deviations were removed as outliers and replaced with the 1st and 99th percentiles.
The treatment variable is export participation; if the proportion of output sold under export orders is more than 30%, it is coded as 1, otherwise 0. There are some factors in the outcome variables, such as labour-input share (total wages in yuan divided by output value in yuan, measured by percentage, expected positive sign with export participation), skilled-worker proportion (number of skilled weavers divided by total workers, in percent, expected positive), raw material cost intensity (material expenditure divided by output value, expected ambiguous), capital-to-labor ratio (equipment value in yuan per worker, expected positive), design investment (annual spending on patterns and molds, in yuan, expected positive). Covariates are operator's years of schooling and experience (expected positive), stall age (years, expected positive), initial capital (log of startup yuan, expected positive), household labour endowment (number of household members available for weaving, expected ambiguous), and a location dummy for Chaozhou (expected positive). These definitions are to make the stalls comparable and have a reasonable basis for the propensity score estimation.

2.3 Propensity Score Matching Model Design

In this paper, the Propensity Score Matching (PSM) model is constructed to eliminate the self-selection bias of simple mean comparison and study the causal effect of export participation on the performance of rattan weaving stalls in Eastern Guangdong. Rosenbaum and Rubin proposed that by conditioning on the propensity score (the probability of receiving treatment given observable characteristics), we can find a single scalar that can be matched for treated and untreated individuals without needing all covariates.
The first step is to use Logit model to estimate the propensity score. If the owner of stall i has chosen to export, then set T_i=1; otherwise, T_i=0. The equation for participation is P(T_i = 1 | X_i) = exp(βX_i) / [1 + exp(βX_i)], where X_i is the covariate vector consisting of the size of the stall (measured by the annual output value and number of weavers), the human capital of the operator (years of weaving experience, level of education, participation in trade skills training), factor allocation variables (labor hours per unit of output, share of skilled weavers, raw material inventory intensity), firm age, and township-level market access dummies. Maximum likelihood estimation gives the fitted propensity score for each stall.
The parameter to be studied is the average treatment effect on the treated, and ATT = E[Y₁ − Y₀ | T = 1], where Y₁ is the revenue or value added of the stall under export participation, and Y₀ is the counterfactual if the same stall did not export. Since Y₀ is not observed, it needs to be constructed from a comparable non-exporting stall. We will use two matching algorithms: nearest neighbour matching pairs each export stall with the non-export stall that has the closest propensity score, and kernel matching forms the counterfactual by taking a weighted average of all non-participants in the common support region, with weights decreasing as the score distance increases. The kernel bandwidth is set at 0.06 and will be tested for variation.
Estimation is based on two conditions. Under the conditional independence assumption, given X_i, the potential outcomes are independent of the treatment; that is to say, all the factors that jointly determine whether to enter into exports and their performance will be captured in the covariates, so the above reasons are given to add operator human capital and factor allocation measurements. The common support condition requires that there is sufficient overlap in the distribution of scores for the two groups; observations that do not meet the common support condition will be removed before matching. Standardized bias reduction and likelihood-ratio tests of the matched sample are used to determine balance.
Nearest-neighbour matching is a method to prevent errors in estimation, which also adjusts for the difference in results due to covariates in the pairs. Bootstrap resampling was carried out 500 times to obtain the standard error. Robustness checks will be done by adding a radius matching method with a caliper of 0.01 and 0.03; re-estimate the model with different sets of covariates; and conduct sensitivity analysis to see how the unobserved confounders need to change to reverse the results.
Finally, the extended heterogeneity analysis will have a new estimate of the ATT for different subsamples of stall size and operator human capital tiers. A comparison of the effects of all the parts will show where to invest more resources and link the estimates from the actual data to concrete ways of improving factor allocation for small stalls, such as upgrading skills and relocating labour.
表1 Comparison of PSM Matching Methods for Rattan Stall Analysis
Matching method
Bias reduction
Sample retention
Common support demand
Suitability rating
Nearest neighbor
Medium
High
Low
★★★★
Kernel matching
High
High
Medium
★★★★★
Radius matching
Medium
Medium
Medium
★★★★
Stratification
Medium
Medium
High
★★★
Local linear
High
Low
High
★★★

Chapter 3 Empirical Results and Analysis of Factor Allocation Effects (~2400 words)

3.1 Descriptive Statistical Analysis

The description will start with a sample of rattan weaving stalls in eastern Guangdong. About 42% of the sample stalls had exported goods in the last year, and 58% were only selling locally. The size of the stall is based on annual sales volume; small (below 100,000 yuan), medium (100,000 to 500,000 yuan), and large (above 500,000 yuan). About half of the samples are small stalls, and approximately one third are medium stalls, with large stalls being the last, so it is a scattered family-based production system in the region. The size distribution is similar to that of labour-intensive handicraft clusters in the province before; generally, family members produce by hand and only a few hired craftsmen are employed instead of formal factory workers.
The differences in factor allocation between exporters and non-exporters are obvious. The proportion of skilled weavers in export enterprises is about 61% of the total number of employees, while that of non-export enterprises is only about 38%. Raw material expenses per employee are also higher in export companies due to the higher-grade rattan and stricter finishing requirements set by overseas buyers. The difference in capital intensity is the largest; the average value of equipment per worker for exporters is about twice that of non-exporters, and although the standard deviation is large in both groups, it is more significant among large exporting stalls. Therefore, there is no uniform technology adoption for all the export cases; different reasons in different destinations and changes in the order volume of the buyer are the reasons for the heterogeneity.
Based on the operator characteristics, the exporters are mostly older, have received more formal education, and have worked in weaving for a longer time. The average age of the operators is 47 years old, the average level of education is 9.6 years, and the average amount of weaving experience is 22 years. However, each of the above numbers has increased for the subgroup of exporters. Given the industry's high dependence on tacit craft knowledge and long-standing commercial relations that have built up over time and facilitated access to foreign buyer networks, it seems reasonable to assume that this is the case. According to the regional customs statistics of woven rattan products, the survey series of revenue in this sample period shows that export revenue is increasing gradually; but due to the weak demand in foreign countries, there has been a drop, and then it has started to recover.
Based on the above analysis, some tentative findings indicate that exporting stalls have different labour quality, material quality and equipment endowments compared to domestic ones. However, raw differences cannot be used to make causal inferences because some stalls in the export group may have unobserved advantages such as good entrepreneurial networks, excellent product quality reputation, good access to working capital, etc., which can help them participate in exports and affect factor choices. This self-selection makes the observed gap exaggerate or distort the actual effect of export. Therefore, the next step will use propensity score matching to build similar groups of exporters and non-exporters, and thus explain why there are differences in factor allocation due to export participation rather than the original characteristics.

3.2 Propensity Score Matching Estimation Results

3.2.1 Logit Estimation of Propensity Scores

In the first stage of the propensity score matching process, we use a Logit model to estimate the probability of export participation, where the binary export status of each rattan weaving stall is the dependent variable and the vector of factor allocation characteristics is the explanatory covariate. The basic idea is that when we have some visible characteristics, the selection bias between exporting and non-exporting stalls can be greatly reduced, so the conditional independence assumption is satisfied for reliable matching in the later stage.
Logit estimation is based on 412 valid stall-level data obtained from the field survey in Eastern Guangdong, of which 168 are exporters and 244 are non-exporters. The model specification test shows that the functional form is suitable; the likelihood ratio chi-square value is 87.36 and is significant at the 1% level, the pseudo-R² is 0.214, and the correct classification rate is 78.2% which is above the common 70% benchmark and has good predictive power. The average variance inflation factor of 1.42 shows that there is no serious multicollinearity among covariates.
The coefficients are shown that four control variables can predict whether to export. Operator education has a positive coefficient of 0.312, which is statistically significant at the 5% level; the marginal effect is 0.068, meaning that an additional year of schooling will increase the likelihood of export by about 6.8 percentage points. This shows the need for literacy, numeracy and basic knowledge of trade in dealing with foreign orders, price quotations and documents. Weaving experience is one of the variables that has a coefficient of 0.045 and a marginal effect of 0.010; it is significant at the 10% level, which means that with the increase of accumulated craft skill and industry connections in our hands, we can now enter more easily into export channels. The initial capital has the largest economic impact; its coefficient is 0.526 and is significant at the 1% level. The marginal effect of 0.114 is that a one-unit increase in the standardised initial capital will increase the probability of export by over 11 percentage points, so the view that working capital for raw material stockpiling and mould investment restricts the entrance of exports is supported. Proximity to export intermediaries, proxied by distance to the nearest collection agent in the township, is negative (coefficient -0.287, significant at 5 per cent), meaning that the lower transaction costs of geographical proximity help to promote export participation. Stall age has a positive effect; however, it is only marginally significant at the 10% level. Households' labour size is not statistically significant; thus, it does not necessarily cause exports to be launched.
After the estimation, we checked whether there was an overlap in the distribution of predicted propensity scores. The exporting stalls have a relatively high score, ranging from 0.18 to 0.87, and the non-exporting stalls have a relatively low score, ranging from 0.04 to 0.91. There is significant overlap between the two distributions in the range of [0.18, 0.87]. Eleven non-exporter observations below the minimum of the exporter group were cut off; thus, the number of matched samples decreased to 401. Kernel density comparison of the two groups before and after trimming shows that their distributions are similar, so the common support condition is met. Based on the above diagnostic results, it can be concluded that the estimated propensity scores have the characteristics of balance and overlap that are needed for reliable matching estimation in the next stage.

3.2.2 Covariate Balance Test After Matching

Propensity score matching is based on the assumption that, given the estimated propensity score, the treated and control enterprises are essentially the same in terms of observable characteristics. The balance test of covariates is one of the basic means to check empirically whether the above assumption is true, and it is therefore necessary to conduct a test before any treatment effect estimates can be considered reliable.
The basic idea of the balance test is simple. If the matching is successful, then the distribution of all covariates should be approximately the same in the matched and non-matched samples of rattan weaving exporters. The difference in the standardised bias of the covariate is used to determine whether the two groups are balanced; this is calculated as the average variance of the first group minus that of the second group. Generally, the absolute value of the standardised bias after matching should be less than 5%, and it is usually required that the absolute value of the standardised bias be less than 10% before matching.
Before matching, the covariates in this study—firm age, workforce size, average wage, capital intensity, artisan skill ratio and access to credit—are not balanced among exporters and non-exporters; the standardised biases vary between 18.4% and 42.7%. The reason for this is that, due to selection bias, the simple difference in means will be affected. Thus, the export enterprises of the rattan weaving industry in the east of Guangdong Province are generally larger, older and have higher skill ratios. After nearest-neighbour matching with a caliper of 0.05, all of the absolute standardised biases of the covariates are below 5%, and the largest standardised bias is 3.8% for the artisan skill ratio. After matching, the average standardised bias was reduced from 27.6% to 2.1%, a reduction of 92.4%, and the median bias was also reduced from 24.9% to 1.9%. The reduction in bias is above the 80% threshold that is generally considered a good result.
The likelihood-ratio test of the joint significance of all covariates also serves as a supplementary way of evidence. Before the match, the test is very likely to reject the null hypothesis that all covariates have no explanatory power for export participation; that is to say, a strong selection into exporting has occurred. After matching, the value of the test statistic is not at a low level for any typical significance level and therefore cannot be rejected; thus, it does not indicate any systematic differences in observable characteristics between the two groups after matching. The estimated propensity scores are grouped by kernel density plot and show that after matching, the two distributions have almost the same shape and support, with no visible gaps or mass points that would indicate a common-support problem.
It should be noted that after the first nearest-neighbor matching, there is a small imbalance in the access-to-credit variable, with a standard bias of 11.2 percent. The kernel-matching method with a bandwidth of 0.06 is then used to replace the matching algorithm, and then all the covariates satisfy the 5 percent requirement. The above results show that the matching method can remove most of the observed selection bias and provide a good basis for accurately estimating the factor allocation effects in the next section.

3.2.3 Average Treatment Effect on the Treated (ATT) Calculation

The central quantity of interest in this paper is the Average Treatment Effect on the Treated (ATT), which is the expected difference in the results of factor allocation for rattan weaving workshops that have participated in exports and those that have not, under the condition that they participated in exports. Formally, ATT is the conditional expectation of Y1-Y0 given D=1, where Y1 is the potential outcome for export participation and Y0 is the hypothetical outcome without export participation. Since we cannot observe the counterfactual, we construct it from non-exporting workshops with similar observable characteristics using the propensity score matching framework of the previous section; that is to say, we can separate the causal effect of export participation from the selection effect caused by firm size, owner education and location conditions.
The ATT of the three outcome variables of skilled-labor share (the proportion of workers with formal weaving certificates or more than five years of experience), raw material intensity (rattan input expenditure per unit of output) and capital-to-labor ratio was estimated. The nearest-neighbour matching with one match, kernel matching with a bandwidth of 0.06, and radius matching with a caliper of 0.05 were all used, and the standard errors were obtained by 500 bootstrap samples to account for the uncertainty of the estimation of the propensity score itself.
The results of all three algorithms show similar results. Nearest-neighbour matching gives an ATT of 0.084 for skilled-labor share with a bootstrap standard error of 0.021 and t-statistic of 4.00; kernel matching gives an ATT of 0.079 (s.e. 0.019, t=4.16); radius matching gives an ATT of 0.081 (s.e. 0.020, t=4.05). All three estimates are significant at the 1% level, and they are relatively small (7.9 to 8.4 percentage points), so they are robust to the choice of matching technology. Export participation will increase the skilled-labor share of participating stalls by about 8 percentage points compared to the counterfactual, and the average skilled-labor share of non-exporters is roughly 31 per cent, which is relatively large. This is also in line with the mechanism that export buyers have higher quality and design requirements for goods, so workshops are required to keep and hire more experienced weavers.
The ATT value for raw material intensity varies from 0.062 (kernel) to 0.071 (nearest-neighbor), and its t-statistic is 2.85 to 3.20, which is significant at the 1% level. Therefore, the export stalls use roughly 6 to 7 percent more rattan per unit of production, that is to say, there are higher requirements for grade selection and lower defect tolerance in export orders. For the capital-labor ratio, the ATT is 0.055 under kernel matching (t=2.41) and 0.060 under radius matching (t=2.58), both are significant at the 5% level, meaning that participation will slightly increase the investment in equipment and tools per worker, but the effect is weaker than that of labour structure.
Robustness checks also support the above results. First of all, the top and bottom 1% of the output value of observations have little impact on the ATT; that is to say, the change is less than 0.006. Second, if we use a different set of covariates that excludes household wealth information and adds distance to the market, the ATTs are within one standard error of the baseline. Third, a placebo test is carried out to assign false treatment timing; that is, stalls will be treated as exporters one period before the actual entry time; thus, small and insignificant results (t < 1.2) are obtained, and there is no problem with the model.
There are some hidden biases left behind that cannot be completely ruled out; that is to say, unobservable entrepreneurial ability can also have a significant impact on the choice of export entry and factor employment. The consistency of estimates among algorithms, the stability under sample exclusion and covariate changes, and the failure of the placebo test jointly suggest that export participation has improved the allocation of factors at rattan weaving stalls in eastern Guangdong; mainly, it has enhanced the composition of labour skills, and there are also some secondary effects on material quality and capital deepening.

3.3 Heterogeneity Analysis of Export Effects on Factor Allocation

3.3.1 Heterogeneity by Stall Scale

In order to study whether the impact of exports on factor allocation is affected by the size of stalls, the sample will be divided into two parts based on the number of regular workers. Micro stalls have fewer than five workers and small stalls have more than five. In the rattan weaving clusters in the eastern part of Guangdong, it is generally believed that there are fewer than five employees; thus, the threshold is set there. Propensity score matching will be re-estimated for each subsample, and the same covariates will be used as in the baseline model, such as owner's age and education level, operating years, credit access, cooperative membership, distance to the nearest collection center, etc., to improve the comparability of treatment and control stalls among different scale groups.
The results of the matching diagnostic test show that the balance is also achieved in the two subsamples. After nearest-neighbor matching with replacement, the standardized bias of all covariates is less than 10 per cent in absolute value, and the pseudo-R² of the probit first stage has decreased significantly compared to before matching; thus, the systematic difference in observable characteristics between exporters and non-exporters has been eliminated. Common support has been satisfied for both groups; however, the region of overlap is relatively small due to the small share of small stalls in the sample.
According to the comparative ATT results, all three factors have significant scale differences. In small stalls, the number of laborers engaged in exports increased by an average of 0.42 people; however, at the 5% significance level, more weavers have been recruited and some of the division of labour in weaving and finishing work has been improved. Per unit of output, material expenses have risen by 18.6%, and in response to higher-quality rattan and processed binding materials required by overseas buyers, costs have increased. Fixed capital investment has risen by 31.4 percent and is significant at the 1 percent level, so steam-treatment equipment, drying racks and quality-inspection tools have been purchased. Micro stalls have no significant effect on labour and materials allocation, and although their capital response is positive (6.8 percent), it is small in magnitude and only weakly significant at the 10 percent level. Only the labour intensity of micro stalls has been significantly adjusted; that is to say, export orders have increased the average weekly working hours of family members by about 9%, and this type of extensive-margin substitution does not require much cash.
There are two possible reasons for this disparity. First, there is a fixed cost indivisibility: the fixed cost of equipment and quality certification investment is relatively high, and as the scale of production expands, the fixed cost is spread over more products, so the per-unit capital deepening barrier is lower. Second, credit access: there are 47 percent of small stalls that can obtain formal or cooperative loans, compared to only 19 percent of micro stalls; in the export group, the number of stalls with credit access is also relatively high, so they can buy more equipment because of the lack of financial constraints that would otherwise limit investment. Micro stalls do not have scale economies or external funds; thus, when there is an opportunity for foreign trade, only family labour can be increased.
Therefore, export-promotion support needs to be scaled differently. Small stalls can use their own investment funds to conduct trade-facilitation and buyer-matching services, but micro stalls need to rely on concessional credit and share equipment facilities such as those operated cooperatively to reduce fixed costs and finance barriers that limit labour-intensive adjustments.

3.3.2 Heterogeneity by Operator Human Capital

In order to study whether the export effect on factor allocation is different due to the quality of the operator, the matched samples are divided into two groups according to the operator's human capital. Following the usual way to measure operator human capital in firm-level studies, it is operationalised by three observable aspects: the level of formal education (whether the operator has completed senior secondary school or above), the length of accumulated weaving experience (whether the operator has worked in rattan weaving for more than ten years), and participation in vocational skills training during the past three years. Operators who meet at least two of the three criteria are in the high-human-capital group, and the rest are in the low-human-capital group. For each sub-group, the propensity score model is re-estimated, and the nearest-neighbour matching with a common support is carried out again to maintain covariate balance before comparing the average treatment effect on the treated.
The estimation results show a clear gradient in the export effect between the two groups. For the low-human-capital group, the ATT of design investment is 0.042, which is not significant; the ATT of technology adoption is 0.067, which is only significant at the 10% level; and the ATT of the proportion of skilled workers is 0.038, which is not significant. The corresponding ATTs for the high-human-capital group are 0.186, 0.213, and 0.159, respectively, all at the 5% level. The matching quality diagnosis shows that the Standardised Bias Reductions are all above 70% and the Pseudo R-squared values after matching are less than 0.05; thus, the subgroup comparisons are internally valid.
The results of the analysis are in line with the absorptive capacity theory. Operators with a high level of education and many years of work experience are more likely to comprehend the requirements of overseas buyers, translate market feedback into product design changes, and assess the profits of employing semi-mechanised equipment or new finishing technology. The high-human-capital operators also have better recruitment of skilled weavers, selection, and retention; therefore, their reputation capital and technical ability will reduce the information asymmetry in rural labour contracting. Low-human-capital operators have low-level production technology and only learn passively; thus, they have not absorbed advanced technology and experience after exporting, and their factor allocation is similar to that before export, so they can only take on simple processing tasks with limited learning.
Thus, the mechanism of learning by exporting is largely dependent on the quality of the operator. Exporting can provide knowledge about foreign standards, packaging, etc., but to turn this knowledge into a new factor, one needs to be aware of it, record it, and implement it; thus, it is precisely at this point that human capital comes into play.
The data also show that there is an interaction with the size of the stall. Operator human capital has a strong positive effect on design and technology investment in large and medium-sized stalls; export orders are sufficient to cover the fixed costs of experimentation and improvement. Small stalls, even though they have a large number of highly skilled people, will be impacted by low finance and small orders, and the advantages brought by human capital will be weakened. Therefore, in order to improve the quality of the operator and increase the number of orders, we should adopt a combined strategy of continuous upgrading and ordering together.

3.4 Analysis of Factor Allocation Improvement Pathways

The empirical results of propensity score matching analysis show that there are differences in the factors of production allocation among the four input dimensions. The labour response is most significant, and the exporting stall will increase the amount of skilled labour in a statistically significant manner compared with the non-exporting one, which shows that the quality of export orders requires high-skilled labour. Capital allocation responds moderately, mainly upgrading equipment rather than expanding fixed assets; the weak and not significant effect of raw material input indicates that rattan procurement remains the same for either the domestic market or export market. Design and skill input cannot be measured directly; therefore, they are used to carry out a robustness test. Collectively, the above results show that we should improve according to the responsiveness and seriousness of each factor's constraints in a certain order.
The first way to address this issue is labour. The lack of family-based labour supply cannot meet the requirements of overseas buyers' consistency and finish standards; thus, the focus is to change family members from untrained workers to trained skilled weavers. A cooperative-based training plan will be organised at the township level and adjusted according to the technical requirements of export orders; cooperatives can pool their demand for training, share the cost of trainers, and have their skill levels certified by intermediaries. Such programmes are in line with the existing vocational training support provided by county-level policies and do not need new system construction.
The second path is about capital. Small stalls have limited working capital, and export orders often require upfront purchases of raw materials to obtain payment. Industry association credit rules, in which the members' stalls can access revolving funds backed by association guarantees, will close the gap; trade-order financing, using confirmed purchase contracts as collateral, will also be extended. Both mechanisms use existing commercial relationships, so no new regulatory system is needed, and they are simple to operate.
The third path is to use materials. Purchasing rattan together through a cooperatively-operated purchasing pool can help reduce the cost of inputs and raise bargaining power with upstream suppliers, which will address the risk of profit fluctuations in the descriptive statistics. Since raw materials have the lowest intrinsic response to export participation, this path requires intentional coordination and is not spontaneous.
The fourth path is for skills and design. Operator training in finishing, along with design collaboration organised by export intermediaries, will enhance the value-added content of the product and lower the rate of rejection.
We need to distinguish between the various modes based on the size of the stall and the operator's human capital. For large stalls with educated operators, capital and design are the best choices; for small stalls with limited formal education, the first is labour training and collective procurement to lower entry barriers. It is feasible to use the first three, as they do not require institutional changes but can only be achieved gradually through the application of existing rules and the strengthening of industry cooperation.

Chapter 4 Conclusion (~600 words)

The last part is the end of this paper. In the first part, a group of people in eastern Guangdong have been unable to trade freely, according to empirical data. We need to use Propensity Score Matching to guide the allocation of factors in practice [8]. The issue that was being studied was why some regions with favourable geographic proximity to export centres and a long history of crafts are still experiencing slow growth in the export volume and value of their products in recent years. Based on the above two types of enterprises, the PSM method is used to study the effect of factor allocation on export performance, and then to control for selection bias due to company size, ownership type, labour composition, etc.
According to the main results, it can be concluded that enterprises with good distribution of labour, raw material procurement, working capital and technology are better at exporting, and the results are consistent under various ways of matching. Sensitivity analysis also showed that the hidden bias needs to be large enough to change the results, and so the internal validity of the conclusion is relatively good. There are three particular reasons for this. Reason 1 is that rational distribution of labour to match experienced weavers with standardised training can reduce the defect rate and increase compliance with the requirements in destination countries [4]. Second, a reasonable reduction in the purchase of rattan raw materials can help to reduce production costs and shorten the time for production, thus better responding to changes in orders. Third, a small amount of consistent investment in semi-mechanised processing equipment will increase the productivity of single workers without losing the handcraft quality that is an important competitive advantage for this region's industry.
Based on the empirical data, this study can also be used to show that quasi-experimental methods generally used in labour economics can be adopted for small-scale, cluster-based traditional industries that cannot be randomly intervened in. Based on this, there will be some new research on the handicraft export sector in other regions, but to do so, enough covariate data needs to be collected through enterprise surveys and customs records. The practical significance is also great. Based on the evidence, an increase in quantity due to an unbalanced expansion of the labour force without corresponding capital or process upgrading will lead to diminishing returns, so enterprise managers should first diagnose their internal situation before expanding exports. Industry associations and local trade promotion agencies can use the matching framework to find enterprises that are most likely to benefit from targeted support projects such as help with quality certification, export credit, and development of e-commerce channels; thus improving the accuracy of resource allocation.
There are some limitations in the study, so I will continue to expand on them. The sample is from one industrial cluster in the Eastern part of Guangdong, and although it provides a good context, the generalisation to other areas of rattan weaving centres is limited. Cross-sectional data cannot reflect the dynamic adjustment process; therefore, it is necessary to use panel data for enterprises that are in different export cycles to estimate the lagged effect of allocation. There may also be other unobserved factors, such as entrepreneurial capabilities and informal network ties, that affect both the allocation choice and export results; therefore, more sophisticated instruments or qualitative triangulation are needed to address this issue.
In short, the stagnation of rattan weaving exports in eastern Guangdong is not due to market saturation, but rather to the insufficient allocation of factors at the enterprise level. Based on the above reasons, the ineffective allocation of factors will be addressed through evidence-based, supportive interventions; thus, the export development of this traditional handicraft industry will be revived.

References

[1]白俊红, 王星媛, 卞元超. 互联网发展对要素配置扭曲的影响[J]. 数量经济技术经济研究, 2022, 39(11): 71-90.
[2]Tuan Le-Minh, Thu Nguyen-Bui Anh. Beyond Tourist Visits: Indirect Tourism Impacts and Pro-Poor Livelihoods in Bao La Bamboo-Rattan Weaving Village, Vietnam[J]. Tourism Cases, 2026.
[3]Guanchun Liu, Yuanyuan Liu, Chengsi Zhang. Factor allocation, economic growth and unbalanced regional development in China[J]. World Economy, 2017.
[4]Peng Han, Zehua Pan. Research on the Protection and Revitalization of Traditional Fujian Rattan Weaving Utensils from the Perspective of Folk Culture[J]. Economic society and humanities., 2024.
[5]Evi Yuniarti. The Value chain Analysis at rattan Industry Cluster in south Lampung[J]. Jurnal Ilmiah Esai, 2018.
[6]Thomas J. Holmes, Matthew F. Mitchell. A theory of factor allocation and plant size[J]. The RAND Journal of Economics, 2008.
[7]Ulrik Beck, Benedikte Bjerge, Marcel Fafchamps A. The Role of Social Ties in Factor Allocation[J]. The World Bank Economic Review, 2018.
[8]Jianbin Li, Xueyuan Cai, Zhixin Liu. Allocating Capacity with Demand Competition: Fixed Factor Allocation*[J]. Decision Sciences, 2016.

同类型范文继续对照

查看全部分类范文 →

这篇表达不错?生成你自己的

输入题目,生成同类型论文初稿,再对照范文逐段打磨到可提交。

生成我的初稿 回范文库继续看 →
← 返回分类:英文论文