Micro, small, and medium enterprises (MSMEs) and poverty reduction: empirical evidence from Indonesia

ABSTRACT

This study aims to analyze the effect of MSMEs on poverty reduction, both directly and indirectly, through labor absorption from 1997 to 2018. The poverty measurements used in this study comprise Head Count Index (P0), Poverty Gap Index (P1), and Poverty Severity Index (P2), while the MSMEs are classified based on their scales, which are small-medium enterprises (SMEs) and micro-small enterprises (MSEs). Control variables, such as economic growth, openness, government expenditure, and investment, are also covered in this model. In general, the results show that MSMEs statistically affect poverty reduction in Indonesia both directly and indirectly. Nevertheless, different business scales offer various implications for poverty reduction. SMEs play a bigger role in alleviating poverty than MSEs as they reduce not only the percentage of poor people but also the Poverty Gap and Severity Index. Furthermore, of the four control variables, only economic growth has a significant effect on poverty reduction, both direct and indirect. Hence, policymakers should support the market certainty of SMEs products to sustain the production cycle.

KEYWORDS:

• Poverty reduction
• micro
• small
• and medium enterprises
• quantitative approach
• labor absorption
• economic development

1. Introduction

Indonesia is one of the countries committed to achieving the United Nations Sustainable Development Goals (SDGs) by 2030. Among the 17 SDGs, the first one, about poverty eradication, is the main focus of governments and stakeholders. Data on poverty published by the Indonesian Central Bureau of Statistics (2019) still show high percentages of poverty that decline very slowly from year to year. During the last two decades (2001–2018), on average, the percentage of poor people decreased by 0.5%, and its number reduced by 720 million annually. Also, on average, the Poverty Gap Index and Poverty Severity Index have remained high which account for 2.78% and 0.77% respectively.

Under such conditions, one of the most strategic efforts to solve the problems of poverty in Indonesia is the empowerment of micro, small, and medium enterprises (MSMEs). In recent years, MSMEs in Indonesia have been experiencing rapid improvements that have impacted the national economy. The annual rate of labor absorbed increased by 2.15% during each of the last seven years of the study (2012–2018). Similarly, the contribution of MSMEs to Gross Domestic Product (GDP) showed a significant average annual improvement of 54% (Ministry of Cooperatives and SMEs 2019). Although the MSMEs have grown, the progress appears slow, such that they have not fully contributed to the reduction of the number and percentage of poor people. MSMEs still face various problems and barriers when it comes to influencing their product competitiveness. According to Manzoor et al. (2019), the development of MSMEs can absorb more labor, improve income, and push economic growth to further reduce the number of poor people and improve their socioeconomic condition (Geremewe 2018).

Some earlier empirical studies have demonstrated a strong correlation between the role of MSMEs and the reduction of poverty, both in developed and developing countries, such as West Virginia, U.S. (Gebremariam and Gebremedhin 2004), Pakistan (Ali, Rashid, and Khan 2014), metropolitan Ibadan, southwestern Nigeria (Adeyemi and Lanrewaju 2014), and most recently, in the eight countries of the South Asian Association for Regional Cooperation (SAARC) (Manzoor et al. 2019). However, other empirical studies have described different results, including Harvie (2003) and Beck, Demirguc-Kunt, and Levine (2005), who sampled 45 countries; the Indonesian Ministry of National Development Planning Agency (2014) and Bank Indonesia (2016) investigations in that country; and Geremewe (2018) in the U.S., who performed their research through literature surveys. They indicate that the problems of poverty faced by different countries are complex in numerous sectors, with varying MSE and SME characteristics in each, as in the number of barriers and challenges inhibiting their development.

Theoretically, the relationship between MSMEs and poverty reduction can be analyzed via economic growth trends and the absorption of labor. The relationship suggests that the growth of MSMEs positively impacts the income of MSMEs actors so that it contributes to the reduction of poor people (Adeyemi and Lanrewaju 2014). Meanwhile, the indirect relationship means MSMEs can contribute to the reduction of poverty through job openings and economic growth.

The connection between labor and poverty is quite strong. This can be understood at the macro and micro levels. At the macro level, poverty occurs due to a low level of labor productivity in producing output. This implies the real wage of labor gains. At the micro level, it is caused by low productivity of a household member who is working, with low human capital in the household along with a large burden of household dependents. Under this condition, labor absorption would hardly solve the poverty problem. Conversely, if labor could produce a high level of work productivity, they would presumably see a raise in their real wage, which in turn could contribute to reducing poverty. Hence, the key that strengthens the relationship between labor and poverty lies in their work productivity. This is in line with the study conducted by Singh (1999) who found that the relationship between employment and poverty reduction could be attained on three conditions: (i) the overall growth rate of labor must be able to absorb new workers with high levels of productivity, (ii) job creation must produce an equitable job distribution between the poor and non-poor, and (iii) the jobs created must have a wage standard or at least a livable, satisfactory wage. The labor economist Rizwanul Islam (2004) asserts that when a high level of economic growth leads to increased production capacity and productivity, the poor have an opportunity to be absorbed into the various productive sectors that can generate higher incomes. Through this process, each laborer absorbed into these sectors would positively affect the poverty alleviation.

In most developing countries, including Indonesia, micro-enterprises are more dominant than small and medium enterprises in terms of numbers of units, labor, and output. According to data from the Indonesian Central Bureau of Statistics (BPS) in 2018, approximately 98.68% of the total MSME units in Indonesia come from micro-enterprise. 89.04% of the total MSME jobs derive from micro-enterprises, and the output constitutes 37.77% of the total MSME output. This enterprise does not require much skill and capital, nor does it have a wage standard, so it is open for anyone. In Indonesia, the majority of micro-enterprise owners are poor. Meanwhile, in small and medium enterprises, the poor only tend to be laborers rather than owners. Thus, the involvement of the poor in the business either as owner or laborer plays an essential role in driving the economy nationally and internationally.

This also could contribute to achieving future SDGs. In Indonesia, the MSMEs empowerment programs have become one of the most important strategies for alleviating poverty. These programs include production capacity expansion and human capital capability improvement through training, capital assistance, and technology support. In the medium- and long-terms, these empowerment programs are expected to lead a change in the status of the business scales from micro to small, small to medium, and even medium to larger entities. These changes imply revamping of the labor market structure, new job creation with high levels of productivity, and other job creation in MSME sectors that directly affect the poor. These could improve the socioeconomic conditions of society as a whole.

Earlier studies found a significant influence of MSMEs on the number of job vacancies (Herman 2012; Hussain, Bhuiyan, and Said 2017; Rotar, Pamić, and Bojnec 2019) and economic growth (Beck, Demirguc-Kunt, and Levine 2005; Manzoor et al. 2019). As Manzoor et al. (2019) observe, SMEs are economy growing machines, and there is a positive correlation between the growth of SMEs and a rising economy. The mechanism of economic growth can further reduce poverty. Therefore, the growth of MSMEs influences the reduction of poverty through the economic growth path.

Nevertheless, earlier studies correlating indirect paths via the quantitative approach are rarely undertaken in Indonesia. Manzoor et al. (2019) estimated the effect of MSME growth on the reduction of poverty using a quantitative approach but did not observe its indirect effect. MSMEs are the drivers of increased demand for labor and economic growth that are expected to further improve the socioeconomic prospects of the MSEs actors (Hussain, Bhuiyan, and Said 2017).

This study aims to analyze the effects of MSMEs on the reduction of poverty in Indonesia through labor absorption. The emphasis on direct and indirect mechanisms is expected to provide recommendations for more appropriate MSME anti-poverty policies for Indonesia or other developing countries.

2. Literature survey

Poverty eradication has become one of the most important goals of SDGs. More than 160 countries, including Indonesia, have signed and committed to the alleviation of poverty by 2030. Those countries seem to need appropriate strategies to help them achieve the goal due to the uncertain global economy. According to Mukras (2003); Harvie (2003); and Hussain, Bhuiyan, and Said (2017), despite this uncertain condition, many countries are still able to survive because they are reinforced by their successful performances of their micro, small, and medium enterprises. From that, MSMEs have become one of the important development strategies and play a big role in the mitigation of poverty.

Many empirical studies have analyzed the correlation between MSMEs and the reduction of poverty. Some of them argue that they have a strong connection like Harvie (2003); Mukras (2003); Beck, Demirguc-Kunt, and Levine (2005); Koshy and Prasad (2007); Asikhia (2010); Adeyemi and Lanrewaju (2014); Hussain, Bhuiyan, and Said (2017); Ali, Rashid, and Khan (2014); Geremewe (2018); and Manzoor et al. (2019), while others, like Katua (2014); Abdullahi and Sulaiman (2015); and Bank Indonesia (2016) reject the argument (Research methodology, especially the approach of an analysis model, is one of the factors leading to differing empirical research findings.

Harvie (2003) analyzed the contribution of MSMEs on poverty alleviation through a descriptive approach and concluded that the growth of MSEs has improved the social and economic condition of poor people in East Asia. This study is supported by Mukras (2003) in Bostwana, Gebremariam and Gebremedhin (2004) in West Virginia, and Koshy and Prasad (2007) in numerous countries, including Swaziland, Malawi, Kenya, Tanzania and Ethiopia, Bangladesh, Sri Lanka, and the Caribbean. Other qualitative studies like the one carried out by Asikhia (2010) found that the market and the capability of MSMEs are two main factors supporting MSME workers in raising their profits that contributes to the reduction of poverty in Nigeria. What is more, Adeyemi and Lanrewaju (2014) assert that the income of MSEs workers in the Ibadan Metropolis in southwestern Nigeria experienced a profit increase of 39%.

Some empirical studies, on the other hand, found that MSMEs have not improved the social and economic conditions of poor people because of the several problems and barriers they face (Katua 2014; Abdullahi and Sulaiman 2015; Bank Indonesia 2016; Geremewe 2018). There are many factors inhibiting the MSMEs solution to poverty, including lack of access to financial services (Ferdousi 2015), inadequate infrastructure, limited market, insufficient training of business management (Geremewe 2018), political instability (Katua 2014), lack of information about business (Abdullahi and Sulaiman 2015), and low competitiveness (Bank Indonesia 2016).

Debates on the findings of empirical studies are also found in quantitative research. Using a log-linear autoregressive model, Ali, Rashid, and Khan (2014) found that the output of small-scale industry has negative effect on the poverty level in Pakistan. His study recommended the importance of simplify procedures to access finances and reduce the cost of credit. This empirical finding is supported by a study carried out by Manzoor et al. (2019) through analysis with a data panel model in SAARC. Furthermore, Kowo, Adenuga, and Sabitu (2019), using analysis of variance, coefficient of correlation, and regression analysis, found that the development of SMEs influences poverty in Nigeria. However, it is contradicted by previous studies conducted by Beck, Demirguc-Kunt, and Levine (2005), which found that MSMEs did not affect the reduction of poverty in the 45 countries where they gathered research samples. Vijayakumar (2013), found that the effect of SMEs on the reduction of poverty in Sri Lanka was not statistically significant. In the recent study, Gamo and Gollagari (2020) who combine the method approach using quantitative and qualitative data in Ethiopia assert that MSMEs could contribute to reduce poverty if they are supported by local government institution.

SMEs can contribute to the reduction of poverty as long as their growth can create vacancies for labor. Herman (2012) did the quantitative analysis and found that SMEs contribute to the absorption of more labor than bigger companies in Romania. This study is in line with the one conducted by Agyapong (2010) on the cases of women working for MSEs in Ghana. Moreover, Kowo, Adenuga, and Sabitu (2019) point out that MSEs influence the reduction of poverty through the creation of job vacancies in Nigeria. Rotar, Pamić, and Bojnec (2019) also found that there is a significant correlation between MSEs and labor absorption in the service industry. Gebremariam and Gebremedhin (2004); Hamdar, Najjar, and Karameh (2017); Manzoor et al. (2019) found that the growth of SMEs contributes to the reduction of poverty through the economic growth and unemployment reduction. Darma et al. (2020) investigated brown sugar as one of MSE products in Indonesia and found that the use of appropriate technology could absorb more labor and increase the income of rural communities.

3. Material and methods

3.1. Data and choice of variable

This study uses secondary data in the form of time series from the period 1997–2018. They were obtained from statistical year book of Indonesia issued by the Indonesian Central Bureau of Statistics (BPS), and the Ministry of Cooperatives and SMEs of the Republic of Indonesia. Secondary statistics include poverty data, MSME output, MSME labor, GDP, economic growth, openness, and investment. Poverty is measured by three indicators, namely Headcount Index, Poverty Gap Index, and Poverty Severity Index.

The poverty line, or poverty threshold, is the average cost of living for ‘basic needs.’ That necessarily includes food (including water), shelter, and clothing. P0 measures the percentage of the total population living below that poverty threshold; P1 measures the gap between the threshold and the average income of people living below the threshold; P2 measures the squares of the poverty gap relative to the poverty line which take into consideration the disparity of average expenditure among poor people (United Nations 2017). These index numbers were calculated by the Central Bureau of Statistics, based on the National Social Economic Survey (SUSENAS), conducted every year by the government of Indonesia. Based on the definitions of these poverty indices, the analysis unit of this study is individuals, not households.

The measurement of MSMEs itself is divided into two scales, namely SME and MSE scales. MSEs and SMEs are measured from the growth of MSEs and SMEs value respectively, while labor absorption in MSEs (L_MSEs) and labor absorption in SMEs (L_SMEs) are measured from the growth of labor in both categories, respectively. Other economic variables used as control variables (CV) are economic openness (Op), which is measured from a percentage of the export and import to the GDP, private investment (I), which is the percentage of domestic and foreign investment to GDP, government expenditure (Gov), which is measured from a percentage of government expenditure to GDP, and economic growth (Gr), which is the real growth of GDP at a constant price.

3.2. Analysis model

This study develops an analysis model that is different from earlier empirical studies. This analysis model is divided into two categories, which show the direct effect of MSMEs on the reduction of poverty and the indirect effect through labor absorption.

3.2.1. Estimation model of direct effect

The functional model to measure direct influence can be seen in equation (1) as follows: (1) $Po{v}_i=F\left(MSME{s}_j,L_j,C{V}_b\right)$(1) where $i$ is the indicator to measure the poverty, including P0, P1, and P2. j is the business scales that are divided into two categories, namely MSEs and SMEs. This study partially calculates MSEs and SMEs into the estimation model. Lj is the growth of labor in MSEs and the growth of labor in SMEs. b is the type of variable control in the model, which are economic openness, private investment, government expenditure, and economic growth. The functional equation (1) is formulated as below: (2) $Po{v}_i={\propto }_0+{\propto }_{1}MSME{s}_j+{\propto }_2{L}_j+{\propto }_3\sum _{b=1}^{4}CV+\mu$(2) By explicitly inserting each variable control, we arrived at the equation (3). (3) $Po{v}_i={\propto }_0+{\propto }_{1}MSME{s}_j+{\propto }_2{L}_j+{\propto }_{3}Gr+{\propto }_{4}Gov+{\propto }_{5}Op+{\propto }_{6}I+{\mu }_1$(3) Based on the classification of poverty indicators and business scales, the equation (3) forms six equations: three equations to analyze the effects of SMEs on three poverty measurements and three equations to analyze the effects of MSES on three indicators of poverty measurement (Appendices).

3.2.2. Estimation model of indirect effect

Furthermore, based on the literature review, it was found that the improvement of MSME input has an effect on the reduction of poverty when supported by an increase of labor absorbed into the business. Therefore, this study estimated the indirect effect of MSMEs on poverty reduction as a result of labor absorption. In addition, other CV like economic growth, economic openness, government expenditure, and investments, also affect poverty as long as they can push the improvement of labor absorption in MSMEs sectors.

Therefore, the structural equation to measure the indirect MSME effect on the reduction of poverty through the absorption of labor in MSME is as follow: (4) $L_j=F\left(MSME{s}_j,C{V}_b\right)$(4) Equation (4) explains that besides variables of micro-small and medium enterprises, economic growth, government expenditure, private investment, and economic openness also affect the absorption of labor in MSME sectors that can further reduce poverty. The functional equation is transformed into the formula below: (5) $L_j={\text{β}}_0+{\text{β}}_{1}MSME{s}_j+{\text{β}}_{2}Gr+{\text{β}}_{3}Gov+{\text{β}}_{4}Op+{\text{β}}_{5}I+u_2$(5) The equation (5) separately forms two equations showing the effect of MSEs and four CV on the absorption of labor in MSEs, and the effect of SMEs and four variable controls on the absorption of labor in SMEs. The equation (5) is substituted into equation (3), which formulates the equation to measure total effect (direct plus indirect) of MSMEs, Gr, Gov, Op, and I on each of the three poverty indicators. (6) $\begin{array}{r}Po{v}_i={\propto }_0+{\propto }_{1}MSME{s}_j+{\propto }_2\left({\text{β}}_0+{\text{β}}_{1}MSME{s}_j+{\text{β}}_{2}Gr+{\text{β}}_{3}Gov+{\text{β}}_{4}Op+{\text{β}}_{5}I+u_2\right)\end{array}+{\propto }_{3}Gr+{\propto }_{4}Gov+{\propto }_{5}Op+{\propto }_{6}I+{\mu }_1$(6) After that, the equation (6) is simplified into the formula below: (7) $\begin{array}{r}Po{v}_i={\mathrm{\partial }}_0+{\mathrm{\partial }}_{1}MSME{s}_j+{\mathrm{\partial }}_{2}Gr+{\mathrm{\partial }}_{3}Gov\\ +{\mathrm{\partial }}_{4}Op+{\mathrm{\partial }}_{5}I+\in \end{array}$(7) ${\mathrm{\partial }}_0$ is ${\propto }_0+{\propto }_2{\text{β}}_0$ is the total intercept, ${\mathrm{\partial }}_1$ is the total effect of MSMEs consisting of: ${\propto }_1$ which is the coefficient of the direct effect of MSMEs, and ${\propto }_2{\text{β}}_1$ referring to the coefficient of the indirect effect of MSMEs through labor absorption. ${\mathrm{\partial }}_2$ is the total effect of economic growth consisting of: ${\propto }_3$ that is the coefficient of direct effect and ${\propto }_2{\text{β}}_2$ that is the coefficient of indirect effect. ${\mathrm{\partial }}_3$ is the total effect of government expenditure consisting of: ${\propto }_4$ for the coefficient of direct effect, ${\propto }_2{\text{β}}_3$ for the indirect effect. ${\mathrm{\partial }}_4$ is the total value of economic openness consisting of: ${\propto }_5$ that is the coefficient of direct effect, ${\propto }_2{\text{β}}_4$ for the indirect effect. ${\mathrm{\partial }}_5$ is the total value of the effect of investment consisting of: ${\propto }_6$ as the coefficient of the direct effect, ${\propto }_2{\text{β}}_5$ for the indirect effect. The equation (7) is used to estimate the total effect of MSEs and SMEs on each of three separate poverty measurement indicators (P0, PI, and P2).

In the poverty structural equation model, labor affects poverty, but in the labor structural equation model, poverty has no effect on labor. Thus, the applied structural equation model does not cause the simultaneous problem as there is no interdependence among endogenous variables. This equation model is known as a causal model, where each equation exhibits unilateral causal dependence (Gujarati and Porter 2009). It means that each equation could apply the OLS method. Moreover, in the poverty structural equation model, there is no correlation between labor and disturbance terms so that labor acts as a predetermined variable. This study also carries out several econometric tests, namely the multicollinearity test, heteroskedasticity test, autocorrelation test, and endogeneity test.

4. Results

In general, based on several econometric tests that have been carried out, the results show that the developed equation models satisfy the assumption of classical linear regression models. The results of our models are divided into direct and indirect effects.

4.1. Estimation result of direct effect

Table 1 presents the data about the estimation result of the direct effect of SMEs and MSEs on the reduction of poverty according to the analysis model developed previously. According to the estimation result, SMEs have a significant negative direct effect on three poverty reduction indicators. It indicates that the increase in growth of SME output would lower the percentage of the poor (P0), the gap between the average expenditure of the poor and the poverty line (P1), and the disparity in average expenditure among the poor (P3). Meanwhile, MSEs have a significant effect on only two indicators, Headcount Index (P0) and Poverty Gap Index (P1). Moreover, labor in SMEs is statistically significant on P0 and P2 at a 10% level of significance. On the contrary, labor in MSEs is insignificant on all poverty indicators.

Table 1. Estimation of direct effect of SMEs and MSEs on poverty reduction.

Poverty indicators independent variables	The effect of SMEs and other control variables on poverty reduction			The effect of MSEs and other control variables on poverty reduction
	P0	P1	P2	P0	P1	P2
MSEs				−4,035***	−0,366**	−0,102
L_MSEs				0,123	−0,018	−0,008
SMEs	−1,453**	−0,471***	−0,152***
L_SMEs	−0,025*	−0,004	−0,002*
Gr	−0,240**	−0,012	0,0006	−0,354**	−0,023	−0,004
Gov	−0,420***	−0,079***	−0,021***	0,075	−0,076***	−0,023***
Op	−0,043	−0,0002	−0,006	−0,051	0,009	0,003
I	0,106	0,048	0,022*	−0,154	0,007	0,005
R^2	0,94	0,94	0,92	0,93	0,92	0,89
F-Statistic	41,246	40,48	29,591	35,39	30,53	21,86
Prob (F-Stat)	0,0000	0,0000	0,0000	0,0000	0,0000	0,0000

*Significant at 10%; **significant at 5%; ***significant at 1%.

As for the CV, for the SMEs estimation, it can be clearly seen that the control variable of economic growth has only significant effect on P0, while government expenditure has a significant effect on all indicators of poverty reduction with P value less than 1%. On the other hand, for the MSEs estimation, economic growth has a significant negative effect on P0 with P value less than 5%, whereas the control variable of government expenditure is significant on two indicators, P1 and P2 at 1% level of significance.

Based on the estimation results, the R-squared value for all estimation results is more than 0.90, which means a more than 90% variation in changes in SMEs, MSEs, economic growth, government expenditure, openness, and investment can explain the variation in the percentage change in the poor population in Indonesia. F-statistic shows all estimation results are significant. Although the values of R-squared and F-statistic are high for all estimations, there is no multicollinearity problem found. According to Gujarati and Porter (2009), the terms of rule of correlation matrix is 0.75 and VIT is 10%. This study found that the correlation coefficient values among independent variables are under 0.75% and the coefficient values for VIT are below 10%, thereby, it can be concluded that all equations are free of multicollinearity symptoms.

Based on the data in Table 2, both SMEs and MSEs are statistically significant and have positive correlation with the labor absorption. It means that the greater the growth of SMEs and MSEs output, the more labor that could be absorbed. The coefficient shows the numbers of 1.483 and 0.212 for SMEs and MSEs, respectively. It indicates that each 1% increase in SME output will increase labor absorption in SMEs by 1.483%, and each 1% increase in MSE output will increase the amount of labor absorbed in MSEs by 0.212%.

Table 2. Estimation of effect of SMEs and MSEs on labor absorption.

Variables	Labor absorption in SMEs	Labor absorption in MSEs
SMEs	1,483**
MSEs		0,212***
Gr	0,100**	0,008***
Gov	0,153	0,005
Op	−0,066***	−0,006
I	0,176***	0,001
R^2	0,85	0,98
F-Statistic	18,41	176,13
Prob (F-Stat)	0,0000	0,0000

*Significant at 10%; **significant at 5%; ***significant at 1%. ns = not significant.

4.2. Estimation result of indirect effect

Turning our attention to Table 3, it can be seen that SMEs have a significant indirect negative effect on two poverty indicators through labor absorption, namely P0 and P2. On the other hand, MSEs are insignificant on all poverty indicators. What is more, among the CV, only government expenditure is insignificant on poverty reduction, while economic growth, openness, and investment have significant negative indirect effect on P0 and P2.

Table 3. Indirect effect of SMEs and MSEs on poverty reduction through labor absorption.

Equations variables	Indirect effect of SMEs on poverty reduction through labor absorption in SMEs			Indirect effect of MSEs on poverty reduction through labor absorption in MSEs
	P0	P1	P2	P0	P1	P2
MSEs				0,0258	0,00381	0,0017
SMEs	−0,037**	−0,0059	−0,003**
Gr	−0,0025**	−0,0004	−0,0002*	0,0010	−0,0001	−0,0001
Gov	−0,0038	−0,0006	−0,0003	0,0006	−0,0001	0,0000
Op	−0,0017***	−0,0003	−0,0001*	−0,0007	0,0001	0,0000
I	−0,0044*	−0,0007	−0,0004*	0,0001	−0,0001	0,0000

*Significant at 10%; **significant at 5%; ***significant at 1%.

The coefficient value of the indirect effect obtained from multiplying the coefficient value of labor in equations (A1–A6) and the coefficient of each independent variable in equations (A7–A8) (these equations can be found in appendices).

Lastly, Table 4 presents the total effect estimation result. It shows that SMEs have a significant negative effect on poverty, mainly on P0 and P2. For the CV, the only variable with a significant effect is the economic growth for P0 with P value less than 1%.

Table 4. Total effect of SMEs and MSEs on poverty reduction.

Equations variables	Equation (7) for total effect of SMEs on poverty reduction			Equation (7) for total effect of MSEs on poverty reduction
	P0	P1	P2	P0	P1	P2
MSEs				−4,009	−0,369	−0,1037
SMEs	−1,490**	−0,476	−0,155**
Gr	−0,242***	−0,012	−0,0008	−0,353**	−0,0231	−0,0041
Gov	−0,424	−0,079	−0,0213	0,075	−0,0761	−0,0230
Op	−0,045	−0,0031	−0,0061	−0,052	0,0091	0,0030
I	0,102	0,0555	0,0216	−0,154	−0,0070	0,0050

Note: The coefficient of each variable is obtained from summing up coefficients in Tables 1 and 3.

5. Discussion

5.1. The analysis of direct effect of MSMEs on poverty reduction

Based on the estimation result of the direct effect of SMEs and MSEs on poverty reduction shown in Table 1, it can be seen that during the observation, the increase of MSEs output can generate income for MSE actors that further contribute to the reduction of poverty. The increase of MSEs actors’ income increased expenses, thus minimizing the poverty gap. That result is in line with the real condition of Indonesia, where most MSME actors come from micro-small enterprise scales. Therefore, the more developed the scale of a micro-small enterprise, the stronger the business competitiveness that further improves the income of the actors. This finding is in line with Ali, Rashid, and Khan (2014) in Pakistan.

The interesting finding of this study was that the MSE variables did not significantly reduce the poverty based on the measurement of the Poverty Severity Index. It means that the additional value gained by MSEs has not been fully able to minimize the disparity of average expenditure per capita between poor people. Capacities and capabilities of MSEs’ actors and the types of business they run were quite varied, which influences the income they get. Business actors who are more creative and able to access the international market will receive a higher income compared to others who have limited access to the market. Based on a study conducted by Adeyemi and Lanrewaju (2014), there is a difference in the additional value between actors of MSEs and SMEs. As a rule, actors of medium enterprises have higher input than actors of micro and small enterprises. The differences in growth affect the income they earn. Moreover, micro and small enterprises have a limited range of markets while medium enterprises have wider markets. Studies carried out by Geremewe (2018) through a literature survey found that the role of MSEs significantly impacts the reduction of poverty though data gathered from surveys show that MSEs face many limitations, including access to markets, financial support, and training.

This study confirms the findings above that the influence of SMEs on poverty reduction was significant and showed negative signs on three poverty indicators. The negative signs indicate that the increase of small-medium enterprise outputs can reduce the number of people under the poverty line, minimizing the gap between the average expenses per capita between poor people.

Based on the estimation, it can be concluded that MSMEs – both MSEs and SMEs –statistically, directly and indirectly, influence the reduction of poverty in Indonesia. This finding confirms some earlier studies like Ali, Rashid, and Khan (2014) in Pakistan, Adeyemi and Lanrewaju (2014) in Ibadan Metropolis, southwestern Nigeria, and the latest one, Manzoor et al. (2019).

The findings of this study suggest an idea for policymakers that an effective strategy to solve poverty problems in Indonesia could to push MSMEs actors to improve the volume and values of their business outputs through innovation and use of various technologies so that they can access a wider market. Policies to alleviate poverty have been implemented by the government through social empowerment programs. Programs promoting MSME empowerment should be improved to solve problems and challenges faced by MSME actors. However, there is another aspect that must be considered so that the government policies can be more accurately targeted, that is to pay attention to the characteristics of MSMEs, including human resources, kinds of business, and type of management (Hussain, Bhuiyan, and Said 2017). In general, micro-enterprises are managed by families and mostly by women (Harvie 2003).

Other variables that contribute to reducing poverty are the absorption of labor in the MSMEs sector. Theoretically, the more employment created or absorbed in the sectors of MSMEs, the fewer people under the poverty line. As each laborer receives a salary, there is a good chance for them to shift status from poor to not-poor, and for the average monthly expenses of poor people to approach the poverty line or at least minimize the disparity of expenses between them.

Based on the estimation, the variable of labor absorption on MSEs and SMEs scales can bring different effects on each poverty reduction. Variables of MSEs did not significantly influence all indicators of poverty reduction. It did indicate that MSE labor was only able to fulfill their daily needs, such that their earnings were still unable to push them above the poverty line. Indeed, the Poverty Gap Index grows wider, while the imbalance of average expenses still looms. Laborers for a small business are generally recruited from within the family and are usually not paid with money (Harvie 2003).

In general, the micro and small enterprises are found in informal sectors, both in rural and urban areas. These enterprises could absorb much labor but with low levels of productivity. Furthermore, these enterprises usually have an inelastic demand, no wage standard, and lack of market access, which affects the sustainability of the business and the income generated. As a consequence, labor would have difficulty fulfilling their basic needs.

On the contrary, as the payroll system applied in small and medium business scales is systematic and distributed well to all workers, the business contributes more to the reduction of poverty. There are several factors that strengthen the relationship between labor in SMEs and poverty reduction. Those are (i) Capacity and ability of the business owner or the laborer themselves because the ability of human resources strongly affects the competitiveness of the business and the income generated; (ii) Financial access support; (iii) Technology support; and (iv) Market access. If all of these factors can be fulfilled, then the labor absorbed will receive sufficient income so that they can fulfill their basic needs and escape their cycle-of-poverty status. This study found that, based on the statistical data, the SME variables show significant effect and negative coefficient for poverty indicators P0 and P2. However, the increase of labor absorbed in the SME sector has not moved the average expenditures of the poor approaching the poverty line.

Economic growth and some other CV like government expenditures, economic openness, and investment play a big role in alleviating the poverty in developed and developing countries. Those variables are substituted into the estimation model together with the MSE and SME variables. The roles of those variables have been supported by some theories and empirical findings in certain developed and developing countries.

However, the current study shows different findings by using a wider measurement of poverty. Economic growth variables that are regressed together with MSEs and SMEs variables significantly reduce the percentage of poor people. The more the economy grows, the more people rise above the poverty line. Some earlier studies by Nursini and Tawakkal (2019) and Manzoor et al. (2019) supported this study. When the poverty measurement is expanded by including the Poverty Gap Index and Poverty Severity Index, the economic growth variable will draw a negative but not statistically significant sign. This shows that the movement of economic activities is still dominated by particular community groups, which in this case are the middle and upper economic level societies. As a result, the average expenses per capita by most poor people do not demonstrate an increase along with economic acceleration. Nor do the effects of economic acceleration have a real effect on reducing inequality in per capita spending among the poor. Economic growth does not cause the income earned by most of the poor population to increase significantly. This is not unexpected because the characteristics of the poor are quite varied in the sense that while some of them still do not have routine work, some others already have. As a result, the disparity did not experience significant changes.

The interesting finding of this study is the government expenditure. So far, government policies to alleviate poverty are quite abundant, as with financial support for MSMEs. That means that the impact of policies that are supported by budgeting can solve the problems for three poverty indicators. The increase in the nation’s expenditure, including the budget allocated for poverty alleviation, not only reduces the poverty gap level but also reduces the gap of the average poor person’s expenditures. In general, this study is in line with Mehmood and Sadiq (2010) in Pakistan, Taruno (2019) in Indonesia, and Oriavwote and Ukawe (2018) in Nigeria.

However, the Indonesian economy that is progressively opening to other countries, as indicated by the improvement of export and import trading, has not significantly affected poverty reduction. One of the factors causing the insignificance of economic openness is that exported goods are still dominated by medium- and large-scale industries, while the number of products from MSMEs remains relatively small. This condition occurs due to the low competitiveness of goods produced by MSMEs at a global level (Bank Indonesia 2016) and their limited access to the market. Likewise, imported goods are generally still supported by medium- and large-scale businesses. However, it indicates that economic openness can reduce the number of poor people through the promotion of MSME products into the global market. This study is not in line with the cases of SAARC countries (Manzoor et al. 2019).

The variable of private investment does not affect poverty reduction on all poverty measurement indicators. This study is not in line with the one conducted by Dhamija and Singh (2018) in India. Two factors are leading this finding. (i) Most investments in Indonesia are directed to large-scale businesses so that they do not directly involve poor people. As a result, large-scale investment has not fully created a multiplier effect on employment, especially for poor people. Employment mostly absorbs labor with a high educational backgrounds or strong skills, while poor people mostly lack these capabilities. (ii) The investment is not yet evenly distributed and is still limited to regions with large numbers of poor people, like Papua and East Nusa Tenggara (Nusa Tenggara Timur, or NTT, the southernmost province of Indonesia).

5.2. The analysis of indirect effect of MSMEs on poverty reduction

The estimation of the direct effect of MSMEs on the reduction of poverty has proved that they play an important role in poverty reduction. On the other hand, MSMEs contribute to poverty reduction only if they can push labor absorption in the MSMEs sector and simultaneously increase the income of labor in that sector. As a result, MSMEs only indirectly affect poverty reduction through the labor absorption in MSMEs.

Based on the estimation of the labor absorption equation, MSEs and SMEs have a significant effect on labor absorption. The data indicate that the higher the MSE and SME output, the more labor each business absorbs. However, compared to the coefficient between MSEs and SMEs, it seems that SME variables are more elastic than MSEs. This is quite justifiable as SMEs tend to have products with high competitiveness, so vendors can access wider markets. Through this condition, the chance to absorb labor is greater. However, micro-enterprises do not require high technology, so that to some extent, they create a larger demand for labor. Herman (2012) found that SMEs contribute more to the absorption of labor in Romania compared to bigger businesses.

Some theories explain that economic growth, economic openness, government expenditure, and investments, in the aggregate, play a major role in the absorption of labor. This study analyzed the effects of those variables on the disaggregate absorption of labor in MSEs and SMEs scales. The estimation shows multiple effects of those variables. Economic openness and investment have a significant effect on the absorption of labor in SME sectors.

Interestingly, the variable of economic openness has a significant effect on the absorption of labor in SMEs but not on MSEs scale. This indicates that the more open the nation’s markets are to the world, the lower the labor absorption on MSEs and SMEs. One of the factors causing this condition is that Indonesian exports are still dominated by big industries. Job seekers tend to choose big companies or industries that promise higher and more sustainable income. Markets for MSME products are limited to local and regional levels. Very few are found on the international scale. This study echoes the finding of Weldeslassie et al. (2019) in Ethiopia who also found that most of MSME products are solely marketed in the local and regional markets.

On the other hand, imported goods are raw and capital materials which are aimed at encouraging export-oriented industries. For now, however, it is only the large industries that are increasingly developing output. As a consequence, MSME labor may choose to resign and relocate to larger-scale companies. Although the data show that the number of laborers in MSMEs sectors is greater than in big companies, the effect of economic openness tends to affect the absorption of labor in big companies. Therefore, to bring a more positive effect on the MSMEs, the productivity and competitiveness of this industry must be upgraded so that their market will not rely solely on the domestic level but also at the global level.

Investments hold an important role in the development of MSMEs, including labor absorption. The flow of private investment from foreign countries, or FDI, can create technology transfers to specific national companies and have an impact on economic development, including the absorption of labor (Tülüce and Doğan 2014). This study found that investment affects the absorption of labor in medium-scale business but not in MSEs. The reason is that most products in medium-scale businesses have achieved some measure of market certainty and clarity in local, regional, and international levels. By contrast, most products from micro-enterprises are sold only in local markets and are not sustainable, so that greater private investment does not positively affect the absorption of labor in small-scale businesses.

Overall, it can be concluded that the growth of MSMEs has a positive effect on the absorption of labor in the MSME sector. However, it should be underlined that the increase of labor numbers in the MSMEs sector does not always lead to the reduction of poverty. That depends on the income laborers earn in the sector.

Estimation results show the difference in the effect between MSEs and SMEs on poverty reduction through labor absorption. SME output growth is statistically significant with regard to poverty reduction, especially in the percentage of the poor population indicators and the Poverty Gap Index. The findings of this study explain that the growth of SME output indirectly reduces the percentage of people under the poverty line, as well as reduces the gap in average per capita expenditure among them. This fact shows that the new laborers in medium-scale businesses derive income that significantly reduces the number of poor people, and that older laborers will receive pay hikes that can also minimize the disparity of income among poor people. However, the income of labor in SMEs is still far below the poverty line.

In the end, the indirect effect of MSE output on poverty reduction was not significant on all poverty indicators. Although the growth of MSEs has a significant effect on the absorption of laborers, the increase of labor numbers in micro-enterprises do not significantly reduce poverty. It means that income earned by labor in MSEs has not passed the poverty line, causing the percentage of poor people to remain the same. Similarly, the income of labor in SMEs also has not yet reduced the Poverty Gap and Severity Index, meaning that the average monthly expenditure of poor people has not changed and that they are still far below the poverty line and the disparity of average expenditures of poor people has not improved. This is quite reasonable, as in general, micro-enterprises are still very small, home-based businesses, so that laborers are not usually paid according to productivity. One factor that influences this condition is the market uncertainty of the MSE output, so that the income earned by the owners and laborer is also uncertain, implying the stagnancy of poverty. Low skills and lack of education about the market are the main factors behind the inability of MSEs to alleviate poverty in Nigeria (Asikhia 2010). To push labor absorption in the MSE sector, requires the creation of a market for MSEs products that can sustain and contribute to poverty reduction.

The CV of economic growth indirectly affects poverty reduction, mainly in the P0 and P2 sectors, through labor absorption in SMEs. Meanwhile, the indirect effects of government expenditures, economic openness, and investment in poverty reduction through labor absorption were also not significant on all poverty measurement indicators. Economic openness and investment are statistically significant with regard to the absorption of labor but this effect does not significantly influence poverty. This finding indicates that the creation of labor in MSEs and SMEs result from economic openness and increases in investment have not been able to accelerate the income earned by poor people working in MSEs or SMEs.

The total effect of the improvement of MSMEs and the four CV on the reduction of poverty related to the three poverty measurement indicators can be seen in Table 4. In general, it can be concluded that the total effect (the sum of the direct and indirect effect coefficients) of MSEs and SMEs on poverty reduction is quite significant. It indicates that MSMEs sectors play a large role in overcoming the problems of poverty, both directly and indirectly, through labor absorption. Similarly, the control variable also plays a large role in reducing poverty through its influence on the absorption of labor in MSEs and SMEs. The implications are that economic growth, government expenditure, economic openings, and investment have to be directed to the creation of new jobs in MSME sectors by considering market certainty and guarantees for MSME products. Wider market access that is not only on the local/regional/national scale but also on the global scale will increase the income of labor and MSMEs actors, and will not just reduce the percentage of the poor but also reduce their depth and severity of poverty.

6. Conclusion

In sum, this study found that the direct effect of SMEs and MSEs are significant on all indicators of poverty reduction. This indicates that the development of MSMEs can reduce the percentage of the poor, the distance between the average expenditure of the poor and the poverty line, and the disparity in average expenditures by the poor. Meanwhile, based on the estimation of indirect effects, it was found that MSEs and SMEs had different effects on poverty reduction. The increase in SME output was significant in reducing the percentage of poor people and the discrepancy between the average expenditures by poor people through labor absorption in SMEs, but not significant in bringing the average expenditures closer to the poverty line. On the other hand, the increase of MSE output did not significantly impact the indicators of poverty reduction through the absorption of labors in MSE scales. Therefore, it can be concluded that, in general, the growth of MSME output was still dominated by owners while the laborers, in general, have yet to receive compensation that could solve their poverty problems.

The estimation of CV showed a different result. In general, government expenditures and economic growth directly showed a significant effect on poverty reduction, while economic openness and investment did not reduce poverty. On the other hand, the indirect effect of economic growth, government expenditure, economic openness, and investment on the three poverty reduction indicators was not significantly indicated by labor absorption in MSEs, but it was significant through the labor absorption in SMEs, except for the variable of government expenditure.

Based on this finding, therefore, the policy recommendations to alleviate the poverty problems are: (i) there should be support of market certainty for MSMEs products, especially for MSEs to sustain the production cycle through digitalization; (ii) products of MSMEs should be promoted so that economic openness can reduce poverty; (iii) investment should be also available in MSMEs sectors, not only in large-scale businesses and industry.

Acknowledgements

The author gratefully thanks to lecturers in my faculty for giving advice and motivation during the accomplishment of this article. The data and views presented in the article are fully the responsibility of the author.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Appendices

Appendix 1

Estimation equation of the direct effect of MSEs on poverty reduction with separately (A1) $P0={\propto }_{10}+{\propto }_{11}MSEs+{\propto }_{12}L\mathrm{\_}MSEs+{\propto }_{13}Gr+{\propto }_{14}Gov+{\propto }_{15}Op+{\propto }_{16}I+{\mu }_{11}$(A1) (A2) $P1={\propto }_{20}+{\propto }_{21}MSEs+{\propto }_{22}L\mathrm{\_}MSEs+{\propto }_{23}Gr+{\propto }_{24}Gov+{\propto }_{25}Op+{\propto }_{26}I+{\mu }_{21}$(A2) (A3) $P2={\propto }_{30}+{\propto }_{31}MSEs+{\propto }_{32}L\mathrm{\_}MSEs+{\propto }_{33}Gr+{\propto }_{34}Gov+{\propto }_{35}Op+{\propto }_{36}I+{\mu }_{31}$(A3)

Estimation equation of the direct effect of MSEs on poverty reduction with separately (A4) $P0={\propto }_{40}+{\propto }_{41}SMEs+{\propto }_{42}L\mathrm{\_}SMEs+{\propto }_{43}Gr+{\propto }_{44}Gov+{\propto }_{45}Op+{\propto }_{46}I+{\mu }_{41}$(A4) (A5) $P1={\propto }_{50}+{\propto }_{51}SMEs+{\propto }_{52}L\mathrm{\_}SMEs+{\propto }_{53}Gr+{\propto }_{54}Gov+{\propto }_{55}Op+{\propto }_{56}I+{\mu }_{51}$(A5) (A6) $P2={\propto }_{60}+{\propto }_{61}SMEs+{\propto }_{62}L\mathrm{\_}SMEs+{\propto }_{63}Gr+{\propto }_{64}Gov+{\propto }_{65}Op+{\propto }_{66}I+{\mu }_{61}$(A6) Estimation equation of the effect of MSEs and SMEs on Labor absorption (A7) $L\mathrm{\_}MSEs={\text{β}}_{10}+{\text{β}}_{11}MSEs+{\text{β}}_{12}Gr+{\text{β}}_{13}Gov+{\text{β}}_{14}Op+{\text{β}}_{15}I+u_{21}$(A7) (A8) $L\mathrm{\_}SMEs={\text{β}}_{20}+{\text{β}}_{21}SMEs+{\text{β}}_{22}Gr+{\text{β}}_{23}Gov+{\text{β}}_{24}Op+{\text{β}}_{25}I+u_{22}$(A8) Estimation equation of the indirect effect of MSEs on poverty reduction through labor absorption (A9) $\begin{array}{rl}P0& ={\propto }_{10}+{\propto }_{11}MSEs+{\propto }_{12}({\text{β}}_{10}+{\text{β}}_{11}MSEs+{\text{β}}_{12}Gr\\ & +{\text{β}}_{13}Gov+{\text{β}}_{14}Op+{\text{β}}_{15}I+u_{21})\\ & +{\propto }_{13}Gr+{\propto }_{14}Gov+{\propto }_{15}Op+{\propto }_{16}I+{\mu }_{11}\end{array}$(A9) (A10) $\begin{array}{rl}P1& ={\propto }_{20}+{\propto }_{21}MSEs+{\propto }_{22}({\text{β}}_{10}+{\text{β}}_{11}MSEs+{\text{β}}_{12}Gr\\ & +{\text{β}}_{13}Gov+{\text{β}}_{14}Op+{\text{β}}_{15}I+u_{21})\\ & +{\propto }_{23}Gr+{\propto }_{24}Gov+{\propto }_{25}Op+{\propto }_{26}I+{\mu }_{21}\end{array}$(A10) (A11) $\begin{array}{rl}P2& ={\propto }_{30}+{\propto }_{31}MSEs+{\propto }_{31}({\text{β}}_{10}+{\text{β}}_{11}MSEs+{\text{β}}_{12}Gr\\ & +{\text{β}}_{13}Gov+{\text{β}}_{14}Op+{\text{β}}_{15}I+u_{21})+{\propto }_{32}Gr\\ & +{\propto }_{4}Gov+{\propto }_{5}Op+{\propto }_{6}I+{\mu }_{31}\end{array}$(A11) Estimation equation of the indirect effect of SMEs on poverty reduction through labor absorption (A12) $\begin{array}{rl}P0& ={\propto }_{40}+{\propto }_{41}SMEs+{\propto }_{42}({\text{β}}_{20}+{\text{β}}_{21}SME{s}_t+{\text{β}}_{22}G{r}_t\\ & +{\text{β}}_{23}Go{v}_t+{\text{β}}_{24}O{p}_t+{\text{β}}_{25}{I}_t+u_{22})\\ & +{\propto }_{43}Gr+{\propto }_{44}Gov+{\propto }_{45}Op+{\propto }_{46}I+{\mu }_{41}\end{array}$(A12) (A13) $\begin{array}{rl}P1& ={\propto }_{50}+{\propto }_{51}SMEs+{\propto }_{52}({\text{β}}_{20}+{\text{β}}_{21}SMEs+{\text{β}}_{22}Gr\\ & +{\text{β}}_{23}Gov+{\text{β}}_{24}Op+{\text{β}}_{25}I+u_{22})\\ & +{\propto }_{53}Gr+{\propto }_{54}Gov+{\propto }_{55}Op+{\propto }_{56}I+{\mu }_{51}\end{array}$(A13) (A14) $\begin{array}{rl}P2& ={\propto }_{60}+{\propto }_{61}SMEs+{\propto }_{62}({\text{β}}_{20}+{\text{β}}_{21}SMEs+{\text{β}}_{22}Gr\\ & +{\text{β}}_{23}Gov+{\text{β}}_{24}Op+{\text{β}}_{25}I+u_{22})\\ & +{\propto }_{63}Gr+{\propto }_{64}Gov+{\propto }_{65}Op+{\propto }_{66}I+{\mu }_{61}\end{array}$(A14)

Appendix 2

Multicollinearity testing

Table A2.1. Correlation matrix among variables for all models.

	GOV	GR	I	L_MSEs	OP	MSEs	L_SMEs	SMEs
GOV	1.000000	−0.664281	0.485473	−0.059743	0.662011	−0.660241	−0.159207	−0.641177
GR		1.000000	−0.651157	0.148565	−0.354977	0.241819	−0.055977	0.222868
I			1.000000	0.202691	0.706179	−0.340480	0.185030	−0.228322
L_MSEs				1.000000	0.275396	−0.395830	−0.035135	−0,270526
OP					1.000000	−0.625172	−0.084174	−0.565173
MSEs						1.000000	−0,366724	0,754056
L_SMEs							1.000000	−0.084174
SMEs								1.000000

Table A2.2. Variance Inflating Testing (VIT).

Variables	The effect of SMEs on poverty reduction	The effect of MSEs on poverty reduction	The effect of SMEs on L_SMEs	The effect of MSEs on L_MSEs
	VIT	VIT	VIT	VIT
SMEs	3.086452		7.522295
L_SMEs	1.712240
MSEs		6.086873		4.279657
L_MSEs		1.303344
Gr	2.618303	2.745674	3.588444	3.223294
Gov	3.622411	4.625732	5.049682	5.370518
Op	3.973896	3.249241	3.832329	3.279709
I	6.891882	4.055836	2.863709	3.055291

Appendix 3

Table A3.1. Heteroskedasticity and autocorrelation test for all equation models.

Dependent variables	Direct effect of SMEs and 4 control variables
	Heteroskedasticity test		Autocorrelation test
	Breusch-Pagan-Goldfrey	Conclusion	Breusch_Goldfrey serial correlation LM	Conclusion
P0	Prob Obs*R-squared: 0.11^ns	No Heteroskedasticity	Prob Chi-squared = 0.11 ^ns	No autocorrelation
P1	Prob Obs*R-squared: 0.15^ns	No Heteroskedasticity	Prob Chi-squared = 0.11 ^ns	No autocorrelation
P2	Prob Obs*R-squared: 0.07*	Heteroskedasticity	Prob Chi-squared = 0.42 ^ns	No autocorrelation
Dependent variables	Direct effect of MSEs and 4 control variables
	Heteroskedasticity test		Autocorrelation test
	Breusch-Pagan-Goldfrey	Conclusion	Breusch_Goldfrey serial correlation LM test	Conclusion
P0	Prob Obs*R-squared: 0.03**	Heteroskedasticity	Prob Chi-squared: 0.35^ns	No autocorrelation
P1	Prob Obs*R-squared: 0.46^ns	No heteroskedasticity	Prob Chi-squared: 0.49^ns	No autocorrelation
P2	Prob Obs*R-squared: 0.42^ns	No heteroskedasticity	Prob Chi-squared: 0.47^ns	No autocorrelation
Dependent variables	The effect of MSE and SMEs on L_MSEs and L_SMEs, respectively
	Heteroskedasticity test		Autocorrelation test
	Breusch-Pagan-Goldfrey	Conclusion	Breusch_Goldfrey serial correlation LM test	Conclusion
L_SMEs	Prob Obs*R-squared: 0.25^ns	No heteroskedasticity	Prob Chi-squared: 0.55^ns	No autocorrelation
L_MSEs	Prob Obs*R-squared: 0.13^ns	No heteroskedasticity	Prob Chi-squared: 0.24^ns	No autocorrelation

Note: ns = not significant; * = 10% level of significance; and ** = 5% level of significance.

Table A3.2. Endogeneity test of L_SMEs and L_MSEs in the poverty structural equation model.

Equation	Endogeneity test of L_MSEs in the poverty structural equation model
	Hausman test	Conclusion	Exogenous test	Conclusion
	Prob residual of L_SMEs		Prob estimated of L_SMEs
P0	0.66^ns	H0 is not rejected (no simultaneous): Exogenous	0.48^ns	H0 is not rejected (Exogenous)
P1	0.6^ns	H0 is not rejected (no simultaneous): Exogenous	0.97^ns	H0 is not rejected (Exogenous)
P2	0.49^ns	H0 is not rejected (no simultaneous): Exogenous	0.75^ns	H0 is not rejected (Exogenous)
Endogeneity test of L_MSEs in the poverty structural equation model
Equation	Hausman test	Conclusion	Exogenous test	Conclusion
	Prob residual of L_MSEs		Prob estimated of L_MSEs
P0	0.01**	H0 rejected (simultaneous): Endogenous	0.85^ns	H0 is not rejected (Exogenous)
P1	0.26^ns	H0 is not rejected (no simultaneous): Exogenous	0.39^ns	H0 is not rejected (Exogenous)
P2	0.3^ns	H0 is not rejected (no simultaneous): Exogenous	0.57^ns	H0 is not rejected (Exogenous)

Note: ns = not significant, ** = a 5% level of significance.