Estimated Results of Impacts on Productivity

The rate of households with land use right certificates is 1.2 kg/m 2 , and households without certificates is 0.84 kg/m 2 . The average age of the household head is 52 years old, of which 72.14% are male and 75.61% are Kinh ethnic group. The number of household heads who have just finished junior high school is 74%.

Estimated results

After the Hausman test to confirm the estimation results, the selected results are the panel data model with fixed effects presented in Table 4.17. The coefficient of the redbook variable is positive and statistically significant at the 5% level. The results indicate that households producing on land with land use rights certificates have higher output than households without land use rights certificates by about 0.71 kg. This finding again provides further insight into the land law adjustments in Vietnam over the past decades. Without secure land use rights, farming households are less likely or not to invest in production activities like households with secured land use rights, leading to lower output (Abdulai et al., 2011). Similarly, without land use rights, household owners do not have enough incentives for investment returns, such households may operate inefficiently (Otsuka and Hayami, 1988). Therefore, paying attention to granting land use rights certificates to farmers is one of the factors promoting productivity.

Similar to the models assessing the impact on technical efficiency and allocative efficiency, in this model, as farm size increases, rice output increases. Other input factors such as household labor, seed costs, fertilizers, pesticides and machinery for production are all positively correlated with output, in which the impact of household labor on output is the largest. Machinery costs have a smaller impact than the remaining costs. This can be explained by the fact that rice production in Vietnam is mainly on small-scale fields, using household labor with manual cultivation experience "first water, second fertilizer, third diligence, fourth seed". Therefore, labor and fertilizer factors are still the main factors affecting output efficiency. This result is consistent with previous studies on rice production efficiency in Vietnam by Kompas (2004), Khai et al. (2011), Hoang Linh (2012).

Table 4.17: Results of estimates of impacts on productivity

Variable name

Regression estimation with fixed effects
Coefficient	Standard error	P_value
redbook_1	0.70730	0.29659	0.017

Maybe you are interested!

area_1

0.15057	0.05025	0.003
area_2	0.31745	0.07183	0.000
ln_labor	0.21651	0.03161	0.000
ln_spf	0.13035	0.03300	0.000
ln_machine	0.00810	0.00376	0.032
educn_1	- 0.01690	0.04043	0.676
educn_2	- 0.01635	0.05313	0.758
agent_1	- 0.05939	0.03720	0.111
agent_2	- 0.05869	0.05266	0.265
female	- 0.10314	0.07206	0.153
terrible	0.07249	0.04148	0.081
ln_income_per	0.64231	0.04769	0.000
pci	0.00351	0.00717	0.624
redbook*pci	- 0.01112	0.00496	0.025
year_2016	0.04592	0.03540	0.195
year_2018	0.11812	0.06153	0.055
_cons	0.73667	0.57411	0.200
n = 1,388
Prob > F = 0.000

Source: author's calculation from Stata 16 software.

In this study, the variables of education, age and gender of the household head were not statistically significant. The Kinh ethnic group grows rice with higher yields than other ethnic groups. When the average income from rice production of the household increases, rice productivity will increase. It can be said that increased income is a significant motivation for farmers to focus on production to increase yields.

In the model, the pci variable is not statistically significant, but the coefficient of the interaction variable redbook*pci is negative and statistically significant. This shows that provinces with better access to land have a significantly reduced difference in output between the two groups of farmers with and without land use right certificates. Thus, when timely information about land is provided, the concerns about being

Land acquisition is eliminated which will make farmers produce with higher productivity.

4.4.2. Assessing the impact of land use rights on agricultural TFP using panel data models and quantile regression

With total factor productivity estimated, to explain the impacts on productivity the thesis considers factors related to agricultural production activities. The model has the form:

𝑙𝑛𝑇𝐹𝑃 𝑖𝑡 = 𝛾 0 + 𝛾 1 𝑟𝑒𝑑𝑏𝑜𝑜𝑘 𝑖𝑡 + 𝛾 2 𝑒𝑑𝑢𝑐𝑛 𝑖𝑡 + 𝛾 3 𝑎𝑔𝑒𝑛 𝑖𝑡

+𝛾 4 𝑓𝑒𝑚𝑎𝑙𝑒 𝑖𝑡 + 𝛾 5 𝑘𝑖𝑛ℎ 𝑖𝑡 + 𝛾 6 𝑤𝑒𝑎𝑡ℎ𝑒𝑟 𝑖𝑡 + 𝛾 7 𝑙𝑛_𝑖𝑛𝑐𝑜𝑚𝑒_𝑝𝑒𝑟 𝑖𝑡 (4.14)

+𝛾 8 𝑝𝑐𝑖 𝑖𝑡 + 𝛾 9 𝑟𝑒𝑑𝑏𝑜𝑜𝑘. 𝑝𝑐𝑖 𝑖𝑡 + 𝛾 10 𝑦𝑒𝑎𝑟 𝑖𝑡 + 𝑒 𝑖𝑡

Where the explanatory variables are defined as in the models in section 4.2.1.

Descriptive statistics for the variables are presented in Table 4.18.

Table 4.18: Statistics of factors affecting TFP

Variable

Unit	Number of observations	Medium	Standard error
lnTFP		2,961	4,35450	0.45846
pci		2,961	5,74434	0.94552
ln_income_per	thousand dong/person	2,952	9,43583	0.81531
Variable	Define	Number of observations	Frequency	Frequency (%)
redbook	1, if the land has been granted land use rights	2,952	2,085	70.63
redbook	0, if the land does not have a land use right certificate	2,952	867	29.37
area	0, if the cultivated area is less than 1100m2	2961	1.001	33.81
	1, if the area is from 1100m 2 to 22000m 2		980	33.10
	2, if the cultivation area is over 2200m2		980	33.10
education	0, if the head of household completed primary school	2,961	1,176	39.72
	1, if the head of household has completed junior high school		890	30.06
	2, if the head of household has completed junior high school		895	30.23
agent	0, if the head of household is under 40 years old	2,961	974	32.89
agent	1, if the head of household is over 40 and under 55 years old	2,961	1,210	40.86

2, if the head of household is over 55 years old		777	26.24
female	0, if the head of household is male	2,961	2,366	79.91
female	1, if the head of household is female	2,961	595	20.09
terrible	1, if the head of household is Kinh ethnic group	2,952	1,821	61.69
terrible	0, if the head of household is of another ethnicity	2,952	1.131	38.31
year	2012	2,961	1.121	37.86
	2016		962	32.49
	2018		878	29.65

Source: Author's calculation from research data

The regression results in Tables 4.19 and 4.20 show that, in both the fixed effects model and the quantile regression model, households with land use rights certificates have higher total factor productivity than households without land use rights certificates. Specifically, in the fixed effects model, the coefficient of the redbook variable is 0.28 with a significance level of 5%, which shows that households with land use rights have higher productivity than households without land use rights by about 0.28%. In most quantiles, the regression coefficient of the redbook variable is positive, in which at high quantiles (q70, q80), the impact of land use rights on factor productivity is the highest, with a difference of 0.32%. This result clearly reflects the role of land use rights on agricultural productivity efficiency.

Table 4.19. Fixed effects estimates for total factor productivity

Variable name

Regression estimation with fixed effects
Coefficient	Standard error	P_value
redbook_1	0.28062	0.12523	0.025
area_1	0.07631	0.04366	0.081
area_2	0.10726	0.07073	0.130
educn_1	- 0.00936	0.03933	0.812
educn_2	0.06094	0.03561	0.087
agent_1	- 0.05756	0.03575	0.108
agent_2	- 0.08460	0.04884	0.083
Female	0.03347	0.06867	0.626

Terrible

- 0.06414	0.09444	0.497
weather_1	0.04355	0.01916	0.023
ln_income_per	0.04841	0.01727	0.005
pci	0.02064	0.02105	0.327
redbook*pci	- 0.04241	0.02162	0.050
year_2016	0.60546	0.02075	0.000
year_2018	0.07954	0.02423	0.001
_cons	3.50127	0.21618	0.000

Source: author's calculation from Stata 16 software.

The positive effect of farm size on allocative efficiency is shown in both models. Specifically, larger farms achieve higher productivity through the use of newer production technologies than smaller farms or are better able to combine available inputs. This result is supported by Heltberg (1998) in Pakistan, who found that due to the more convenient use of fertilizers and agricultural machinery, farm size is positively correlated with agricultural productivity efficiency. Kawasaki (2010) using data from Japan, and more recently Sheng et al. (2015) in Australia, Singh et al. (2018) in India also come to similar conclusions.

Table 4.20: Quantile regression estimates for total factor productivity

Variable name

Quantile regression estimates

q40

q50

q60

q70

q80

redbook_1

0.24694*

(0.14919)

0.28020***

(0.10081)

0.31453***

(0.10951)

0.32242***

(0.12027)

0.32422***

(0.12305)

area_1

0.08130

(0.05033)

0.07637**

(0.03399)

0.07129*

(0.03694)

0.07012*

(0.04058)

0.06985*

(0.04152)

area_2

0.13559*

(0.08129)

0.10761**

(0.05502)

0.07873

(0.05969)

0.07210

(0.06551)

0.07058

(0.06702)

educn_1

-0.01764

(0.04719)

- 0.00946

(0.03188)

- 0.00102

(0.03464)

0.00092

(0.03804)

0.00136

(0.03892)

educn_2

0.05803

(0.04270)

0.06091**

(0.02883)

0.06387**

(0.03134)

0.06456*

(0.03443)

0.06471*

(0.03522)

agent_1

-0.06903

(0.04267)

- 0.05770**

(0.02885)

- 0.04602

(0.03133)

- 0.04333

(0.03440)

-0.04272

(0.03519)

agent_2

-0.09641*

(0.05775)

- 0.08475**

(0.03902)

- 0.07272*

(0.04239)

- 0.06996

(0.04656)

-0.06933

(0.04763)

female

0.03402

(0.08040)

0.03348

(0.05428)

0.03293

(0.05901)

0.03280

(0.06483)

0.03277

(0.06632)

terrible

-0.04101

(0.09894)

- 0.06385

(0.06686)

- 0.08742

(0.07263)

- 0.09283

(0.07976)

-0.9407

(0.08161)

weather

0.03858*

(0.02266)

0.04349***

(0.01531)

0.04855***

(0.01663)

0.04972***

(0.01827)

0.04998***

(0.01869)

ln_income_per

0.04736**

(0.02127)

0.04840***

(0.01436)

0.04947***

(0.01561)

0.04971***

(0.01715)

0.04977***

0.01755

pci

0.01886

(0.02507)

0.02062

(0.01693)

0.02243

(0.01840)

0.02285

(0.02022)

0.02294

(0.02068)

redbook*pci

- 0.03939

(0.02588)

- 0.04237**

(0.01748)

- 0.04544**

(0.01900)

- 0.04615**

(0.02087)

-0.04631**

(0.02135)

year_2016

0.61241***

(0.02503)

0.60555***

(0.01692)

0.59846***

(0.01838)

0.59684***

(0.02018)

0.59646***

(0.02064)

year_2018

0.08571***

(0.02824)

0.07962***

(0.01908)

0.07333***

(0.02073)

0.07189***

(0.02276)

0.07156***

(0.02329)

Note: Symbols: *, **, ***, represent statistical significance levels of 10%, 5% and 1%, respectively.

Source: author's calculation from Stata 16 software.

Similar to previous research results, the regression coefficients are positive and significant for the educational level of the household head, specifically, household heads with more years of schooling will have higher factor productivity. This shows that the level of education is a necessary factor for increasing production efficiency for farmers. Farmers with secondary education or higher produce with higher or better productivity than those with no education or only primary education. This result is consistent with most studies on agricultural production efficiency such as Abedullah and Mushtaq (2007), Khai and Yabe (2011), Koirala et al. (2014). However, the results also show that farmers' education is still low, mainly below high school level.

The age of the household head has a negative impact on factor productivity, with the oldest household heads cultivating with lower factor productivity than the youngest household heads. This result has also been confirmed by the study of Onumah et al. (2010) or Shaheen et al. (2011) both of whom argued that younger farmers are more advanced and willing to adopt new technologies and thus have higher productivity efficiency than other farmers.

In the study results, households with higher average income from agricultural production have higher factor productivity. In addition, policies and institutions are also an important factor determining the rate of technology diffusion in agricultural production. Because the agricultural structure is widely dispersed, it requires that before new technologies can really affect productivity, they must be applied to a large number of households. In this study, local land management policies also have a positive impact on productivity, specifically, in provinces with better land access index, the productivity gap between households with certificates and households without certificates is smaller. Ethnicity and gender of the household head do not affect the aggregate productivity efficiency in this study.

Chapter Summary

Chapter 4 is the main content of the thesis, this chapter focuses on estimating the agricultural production efficiency of farming households including technical efficiency, allocative efficiency, economic efficiency and total factor productivity. Next is the quantitative analysis of the impact of land use rights on different aspects of production efficiency in four aspects: (i) Assessing the impact of land use rights on technical efficiency and allocative efficiency using a panel data model with general estimation methods and non-parametric regression. (ii) Assessing the impact of land use rights on output efficiency using a panel data model with fixed effects. (iii) Assessing the impact of land use rights on agricultural TFP using a panel data model and quantile regression. To diversify efficiency measures, the thesis uses different dependent variables and different evaluation methods to ensure the stability of the impact of land use rights on production efficiency. The estimated results from the models are summarized as follows:

(1). The level of efficiency in production is still quite low, especially the allocative efficiency and economic efficiency, in which the average technical, allocative and economic efficiency levels are 73.8%, 46.8% and 35.3% respectively. This shows that households still have many opportunities to improve productivity with available natural conditions, techniques and resources. In addition, the difference in efficiency between households is still quite large, showing that the production level is not uniform, with technical efficiency ranging from 11% to 95%; allocative efficiency ranging from 17% to 61%, households with low production efficiency still have to make great efforts to catch up with the technical level and ability to allocate production resources compared to the most efficient farming households. This result is also consistent with previous studies in Vietnam (Khai et al., 2008), South Africa (Londiwe et al., 2014), China (Tang et al., 2015) or Ethiopia (Musa, 2015).

(2). All models show the important role of agricultural land use rights on production efficiency, including: technical efficiency, allocative efficiency, output and total factor productivity. When having a land use right certificate, farming households with technical efficiency are about 11% higher than those without, with allocative efficiency of 3%, with output of 0.71 kg and with factor productivity of about 0.28. The consensus in these results strongly affirms the role of land use rights on the production efficiency of farming households. This result is also supported by many other similar studies in Nicaragua (Abdulai et al., 2001), Ethiopia (Ahmed et al., 2002), Bangladesh (Rahman et al., 2009), India (Manjunatha et al., 2013) or China (Zhou et al., 2019).

Comment