2.1.1. Study on the types of farmhouses.

Fan [10] divided happy farmhouses into nine types according to different types of operations, including sightseeing farm, urban park, agricultural park, leisure farm and farmhouse. Ding & Sun [11] divided Leisure Agriculture into six types according to different management structures: leisure planting, leisure forestry, leisure animal husbandry, leisure fishery, leisure sideline and leisure ecological agriculture. The various functions can be divided into seven types: sightseeing farm, leisure farm, science and technology farm and environmental farm. According to Chen, three different classification standards can be seen [12]. The first is according to the types of independent management, farmers as the leading operators, company-led participation of farmers, companies and communities, farmers’ cooperation, leasing, village committee-led enterprises, individual operations, equity farming enterprises and multi-stakeholder cooperation. The second classification standard is defined according to the actual situation and the specific local human customs or natural resources. The third criterion is to classify the benefits brought by Happy Farmhouse in various aspects, such as economic and cultural.

From the point of view of management subject, taking Zhejiang province as an example, Shen and Yu [13], believe that the factors affecting the management mode of happy farmhouse in Zhejiang mainly include the government and the operators, which are reflected in the weak infrastructure of happy farmhouse and the low level of management, such as the strong modern flavor of farmhouse, and give the relevant suggestions. Li [14], points out that there are many factors affecting the operation mode of Neijiang Happy Farmhouse. Still, the critical one is the government support, followed by investors, Happy Farmhouse owners and service personnel. It is believed that as long as corresponding measures are taken under the active guidance of the government, Neijiang’s rural happy farmhouse will be developed better. Yao [15] based on the development status of happy farmhouse tourism in Hefei and the excellent happy farmhouse tourism models, proposes that the factors affecting the development of the local happy farmhouse tourism model are the government’s support, enterprise training and network publicity, and the development is carried out according to local conditions through the government’s strong support, increased publicity and good operation norms.

2.1.2. Development of Happy Farmhouse.

Guo [16], in a comparative study of leisure agriculture in Taiwan and the Chinese mainland, concludes that the mainland’s leisure agriculture has such limiting factors as similar products without characteristics, lagging research planning, incomplete infrastructure, extensive operation and management, single function, insufficient initial funds, and insufficient publicity. Bai [17] takes Guiyang city as an example and points out that Guizhou’s leisure agriculture is weak in brand promotion and industrial cluster scale expansion. Wang and Zhang [18] point out that Guizhou’s leisure agriculture still has problems such as serious product homogeneity, lack of cultural characteristics, weak brand awareness and extensive resource development methods. Xu [19] investigates the happy farmhouse in Guizhou, and finds that there are four problems in service quality and level to be improved, infrastructure construction to be strengthened, supervision and management mechanisms to be imperfect, and a lack of outstanding characteristics of the happy farmhouse brand.

Scholars have revealed the problems encountered in the development of rural tourism and also proposed optimization paths from multiple dimensions. For instance, Han et al. [20] describe specific characteristics of upgrading the consumption structure, and took Bengbu Farmhouse as an example to suggest cultivating customer groups, building brands, and characteristic positioning. Jin [21] points out that to meet Chinese residents’ tourism demand after consumption upgrades, it is necessary to start with infrastructure, a tourism complex, and a quality tourism service. Yuan [22] proposes an innovative path of happy farmhouse development mode: improve infrastructure construction, optimize industrial structure and activate the operation mechanism of happy farmhouse, to promote the sustainable development of rural tourism. Wang [23] points out that there are still problems, such as complex complaints, blindly following the trend, a lack of cultural connotation and low benefits in the operation of farm happy farmhouse, and puts forward corresponding countermeasures. Dong [24] analyses the current situation and problems of Chengdu’s rural tourism development and gives a development strategy.

2.1.3. Research on the management efficiency of Happy Farmhouse.

When the global tourism economy reaches a particular scale, the pursuit of tourism development quality and efficiency becomes a deeper problem that needs to be solved. Many researchers studied tourism efficiency and influencing factors at different scales from different angles. Marrocu and Paci [25] selected human capital, social capital, technological capital, and tourism flow as basic variables to study the total productivity of tourism factors and its spatial spillover effect in 199 regions of the European Union. Hadad et al. [26] evaluated the economic efficiency of tourism in 34 developed countries and 71 developing countries, and it shows that globalisation and accessibility are crucial to the efficiency of tourism in developing countries. At the same time, labor productivity has an essential impact on the overall efficiency of tourism globally. Predonu [27] took Romania as a case study to discuss the significance of tourism development and internal mechanisms to improve national economic and social efficiency. Soysal Kurt [28] used the CCR-DEA method to estimate the relative tourism efficiency of 29 European countries in 2013, taking tourism cost, number of employees and number of reception beds as input variables and tourism income, number of tourists and number of stay days as output variables. The results show that the number of countries where tourism is relatively effective and ineffective is 16 and 13, respectively. Gursoy and Goral [29] measured and compared the tourism efficiency index of 129 countries in the world based on the DEA super efficiency model. Corne and Peypoch [30] first used DEA to evaluate the tourism performance of 13 tourist destinations in France. Then, based on the tourism attraction theory, fuzzy set qualitative comparative analysis (fsQCA) was used to discuss the number of monuments, the number of museums, whether there are beaches and ski resorts, and their impact on the efficiency of tourist destinations. Radovanov et al. [31] investigated the tourism development efficiency of 27 EU member states and 5 Balkan countries from the perspective of sustainable development by using panel data from 2011 to 2017. The study shows that the average tourism efficiency of the 15 EU countries is the highest, the Balkan countries are higher, and the tourism efficiency of the later member states is the lowest. At the same time, it is found that the sustainability of tourism development, the share of GDP, the number of tourists and the income of inbound tourism have significant effects on tourism efficiency. Jiang et al. took the rural tourism in the outer suburbs of Beijing as the research object and used the DEA-Malmquist index model to conduct a comprehensive evaluation of the rural tourism efficiency, and found that the expenditure on agriculture, forestry and water affairs, employees, per capita GDP and the number of tourism resources had significant positive influences on the tourism efficiency. It also gives development suggestions from accelerating the integration of rural tourism resources, continuously increasing the financial investment to support agriculture, and strengthening the education and training of practitioners. Dong [32] measured and analysed the efficiency of the six different scenic areas in China and concluded that the efficiency of East China and Southwest China was better, while that of Northwest and Northeast China was worse. In addition, Liu et al. [33], Xu [34], Luo [35] and Yang [36] took Anhui Province, Jiangxi Province, Li minority areas in Hainan Province and Jilin Province as research objects, respectively, to explore the efficiency of rural tourism poverty alleviation in the above regions and put forward suggestions for optimizing development.

2.1.4. Review of research.

The existing research has made rich achievements in the types of farmhouse operation, development problems and operation efficiency. Compared with previous studies, this paper makes two improvements and explorations. First, regarding research scope, while previous efficiency assessments predominantly focused on macro- and meso-level analyses, this study shifts to a micro-level perspective by specifically targeting happy farmhouse operations. This approach fills a critical gap in the literature, which has largely overlooked efficiency evaluations at the firm level within China’s micro level.

Second, in methodological rigor, conventional Data Envelopment Analysis (DEA) models often neglect external environmental factors that significantly influence operational performance, leading to biased estimates. To address this limitation, we adopt the three-stage DEA framework proposed by Fried et al., which systematically disentangles the effects of environmental variables and statistical noise. Our analysis leverages panel data (2014–2023) from 50 happy farmhouses in Huzhou, Zhejiang Province, sourced from the Zhejiang Provincial Department of Culture and Tourism. This dataset enables a granular examination of comprehensive efficiency, pure technical efficiency, and scale efficiency across diverse operational models. Further analytical techniques—including slack value analysis and super-efficiency modelling—provide actionable insights for sectoral governance. Key contributions of this study include: firstly, Enhanced Measurement Accuracy. Integrating Stochastic Frontier Analysis (SFA) into the DEA framework isolates environmental confounders (e.g., policy support, regional GDP) and stochastic disturbances, yielding purified efficiency metrics that reflect actual managerial performance. Secondly, introducing the super-efficiency model enables horizontal comparison between business entities whose technical efficiency reaches the optimal level, to identify benchmark cases. Thirdly, dividing the happy farmhouses into three types: shareholder cooperative type, enterprise cluster type and individual family type, reveals how the governance structure regulates the relationship between economies of scale and technical efficiency.

3.2.1. Data envelopment analysis.

CCR-DEA is an efficiency evaluation model developed by Charnes [38] to evaluate the relative efficiency of each decision-making unit. Banker [39]and others removed the restriction that the return to scale should be fixed in the CCR model and proposed the BCC model. When using the DEA model to calculate efficiency, the first thing to consider is the choice of input-oriented and output-oriented DEA models. If the research object belongs to a profitable enterprise and cannot be controlled due to market demand volatility, but input items can be adjusted freely, input orientation should be adopted. Otherwise, output orientation is used. Therefore, the traditional input-oriented DEA model is adopted in this study.

Different from the CCR model, the BCC model assumes that variable returns to scale, that is, the increase of some inputs does not lead to a relative increase of output items. Therefore, the concept of distance function is used to derive a model that can measure pure technical efficiency, efficiency to scale and return to scale. Suppose the unit j (j = 1,..., n) uses i (i = 1,..., m) input quantity is x_ij, whose r (r = 1,... s) output is y_rj, h_j is the technical efficiency of the jth decision unit, τ_r and ω_i are the weights of the rth output item and the ith input item respectively, and μ_k is the return to scale index of the kth decision unit, then the linear programming model of the BCC model is expressed as:

(1)

For easy calculation and additional information in interpretation, the formula (1) is converted into the dual form as follows:

(2)

θ has no positive or negative restrictions, si- and sr + are the relaxation values of the input and output terms, respectively, and ε is a non-Archimedes constant.

If θ = 1, si- = sr+ = 0, it is strong efficiency. If θ = 1, si- ≠ 0 or sr+≠0, it is weak efficiency. If θ < 1, it is not efficiency.

Since there may be more than one relatively efficient DMU, Anderson [40] and others sorted efficient DMU based on the efficiency value of CCR model and proposed the concept of super efficiency. It is measured by removing a certain DMU with an efficiency value of 1 from the efficiency frontier, then constructing a new efficiency frontier based on the remaining DMU, and recalculating the distance from the eliminated DMU to the new efficiency frontier. Since the eliminated DMU is not enveloped by the new efficiency frontier, the calculated new efficiency value will be greater than 1, so that it is easier to sort the efficient DMU, and the efficiency value is the super efficiency value. Its basic mode is expressed as:

(3)

θ* is the super efficiency value, and s- and s + are the vectors composed of the relaxation values of the input and output terms, respectively.

3.2.2. Random frontier analysis.

In the second stage, environmental factors, statistical interference and management inefficiency are decomposed by the cost boundary function of the stochastic production frontier. Assuming that the ith input relaxation value of the jth decision making unit is sij and subject to k exogenous environmental variables z_j = [z_1j, z_2j,..., z_kj], according to the ideas of Battese [41] the relationship between input relaxation value and exogenous environmental variables can be set as follows using stochastic frontier method:

(4)

f_i (z_j; β_i) is the influence function of ecological variables on input relaxation value, β_i is the parameter vector of ecological factors to be estimated, residual term is combination error v_ij + u_ij, v_ij ~ N (0, ) is statistical error, u_ij ~ N(u_i, ) is management inefficiency, and they are assumed to be independent and uncorrelated. The parameters (β_i,, u_i, are estimated using the maximum likelihood method.

In order to adjust the variable value of the original input term of the DMU, the conditional estimator of uij is obtained by using formula (5), where Φ and φ are the distribution function and density function of the standard normal distribution, respectively.

(5)

Combining formula (4) and formula (5), statistical errors can be separated and conditional estimates of statistical factors can be obtained:

(6)

i=1, 2, …, m; j=1, 2, …, n.

Then the results of SFA are used to adjust the input items, and all DMUs are adjusted to the same environmental conditions. There are two ways to adjust: one is to increase the number of DMU input with good operating environment, so that its operating environment reaches the worst and its luck reaches the worst degree; The other way is to reduce the number of DMUs that deal with bad operating environment to the optimal operating environment and the best luck. This approach follows Fried et al. (2002), who recommend the first adjustment method to avoid potential negative input values that could complicate subsequent analysis [47] The formula is expressed as:

(7)

x_i*j is the adjusted value of the actual input x_ij. The first bracket represents adjusting all DMUs to the same operating environment, and the second bracket represents adjusting all DMUs to the same luck. When both environmental factors and statistical interference items are adjusted to be consistent, only the effect of management inefficiency remains.

4.1.1. Stage I: DEA efficiency under original input-output data.

In this study, the input-oriented BCC model and MaxDEA8 software were used to measure the relative operational efficiency of 50 happy farmhouses in Huzhou area of Zhejiang Province in 2023. Among them, only 14 happy farmhouses had effective DEA, and 36 had decision units that produced input relaxation value. (As shown in Table 2).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Efficiency analysis table of the first stage.

https://doi.org/10.1371/journal.pone.0339582.t002

4.1.2. Stage II: Environmental analysis and adjustment affecting input-output efficiency.

Frontier 4.1 software is used to analyses the SFA of the environmental factors affecting the operational efficiency of the Happy Farmhouse. That is, perform regression analysis with three environmental variables with logarithm as the independent variable and the relaxation of four inputs in the DMU measured in stage I is taken as the dependent variable with logarithm. The results are shown in Table 3.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. The second stage random frontier regression estimation results.

https://doi.org/10.1371/journal.pone.0339582.t003

According to the SFA regression results in Table 3, all three environmental variables demonstrate statistically significant effects on input slacks at the 5% level, confirming the validity of our variable selection. Specifically, the analysis reveals distinct directional impacts: land area exhibits a significant positive coefficient on the slack of high-quality practitioners (β = 0.304, t = 5.691), indicating that larger operational areas correlate with increased input redundancy. In contrast, both the level of economic development (β = −0.1080, t = −2.547) and policy innovation (β = −0.1809, t = −4.660) show significant negative coefficients on the same input slack, suggesting these factors contribute to reducing resource waste and improving input utilization efficiency. The highly significant γ values (approaching 1.0) and the conclusive LR one-sided error tests (all exceeding critical values) further validate that management inefficiency, rather than random noise, dominates the performance variations, thus confirming the necessity of our second-stage SFA analysis.

A positive regression coefficient indicates that increasing the environmental variable value will increase the input slack value (input redundancy), resulting in waste or insufficient output of each input variable. Conversely, a negative regression coefficient indicates that increasing the value of environmental variables is conducive to reducing the amount of input relaxation, that is, to reducing the waste of each input variable or increasing the output.

1. (1). The level of economic development. The regression coefficients of the economic development level of the four input relaxation variables are all negative, indicating that the economic development level measured by regional per capita GDP reduces the relaxation value of all input variables and improves the efficiency of fund use, which is consistent with the findings of Chen et al. [43] and Migue et al. [44] With the increase of per capita GDP, on the one hand, there will be greater consumption power for infrastructure services, such as transportation, public facilities, schools, and medical and health facilities, and the demand will increase with the increase of income. On the other hand, it can also provide sufficient financial support for infrastructure investment and construction, promote the supply of infrastructure, and drive the improvement of the input-output efficiency of infrastructure.
2. (2). Land area. The land area regression coefficients of the four input relaxation variables are all positive, indicating that the increase of land area has a reverse effect on the efficiency of these four investments, reflecting that the larger the farmland management area, the greater the investment input should be.
3. (3). Policy innovation. The policy innovation regression coefficients of the four input relaxation variables are all negative, indicating that the degree of regional policy innovation support reduces the relaxation value of input variables and improves the technical efficiency of the management performance of farmhouse. This is consistent with the findings of Hu et al. [45] and Wang and Hao [46]. The introduction of policy innovation can bring advanced management ideas and promote the application of new technologies [47], which is conducive to achieving higher output efficiency with less input cost.

The above results show that the three environmental variables have different directions of influence on the efficiency of happy farmhouse, which may lead to overestimation of the efficiency value of DMUs in better environment or luck, and underestimation of the efficiency value of DMUs in worse environment or luck. Therefore, the environmental impact value of relaxation value of each input term can be obtained by means of SFA regression equation through the analysis of the second stage. The mixed error term is obtained from formula (4), and then the random interference term is obtained by separating the management inefficiency value from formula (5) and formula (6), and then the environmental adjustment value and random adjustment value are further obtained. Finally, the adjustment value of the input term is obtained from formula (7), so that all DMUs are in the same environment.

4.1.3. Stage III: Empirical results of DEA after input adjustment.

According to the SFA empirical results of stage II, the MaxDEA8 software is used to bring the adjusted input value and the original output into the model again for efficiency measurement, and it was divided into three categories for statistical analysis, and the results were shown in Table 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Efficiency analysis table of the third stage.

https://doi.org/10.1371/journal.pone.0339582.t004

1. (1). From Table 4, there is a significant difference in the operational efficiency of the 50 happy farmhouses before and after the input adjustment. In terms of absolute numbers, the number of DMUs with constant returns to scale in phases I and III is 14 and 20, respectively, for a total increase of 6 units. The DMUs with increasing returns to scale were 19 and 22, for a total increase of three; The number of DMUs with decreasing returns to scale was 17 and 8 each, for a total reduction of 9; Among the 19 Not Efficiency DMUs, 5 of them are due to scale inefficiencies, while the remaining 14 Not Efficiency ones are due to both pure technical efficiency and scale inefficiencies. The average value of the overall technical efficiency of the decision-making units increased from 0. 7818–0. 8335, a total increase of 6. 61 per cent. The average value of pure technical efficiency increased from 0. 8910–0. 9294, for a total improvement of 4. 3109 per cent. The average value of scale efficiency increased from 0. 8648–0. 8915, for a total improvement of 3. 0864 per cent. After removing the environmental impact, the pure technical efficiency of the provinces, autonomous regions and municipalities has increased to some extent, while the scale efficiency has decreased significantly. The adjusted pure technical efficiency is significantly higher than the scale efficiency, indicating that the lower comprehensive technical efficiency is mainly caused by the scale inefficiency.
2. (2). From the perspective of the three different types of business entities, DMU1-DMU16, belong to the individual farmer type of business entities, and their overall efficiency is low, with a total of 10 Not Efficiency, 5 DEA weakly efficient, and only one DEA strong Efficiency. DMU17-DMU33 belong to the farmer participation type of business entities, while their overall efficiency values are relatively high, with the existence of 9 DEA strong Efficiency, 4 DEA weak Efficiency and 4 Not Efficiency happy farmhouses. DMU34-DMU50 belong to the company cluster type of business entities and have the highest overall efficiency values, with the presence of 10 DEA strong Efficiency, 2 DEA weak Efficiency, and 5 Not Efficiency. This suggests that through the company cluster model of the happy farmhouse, it has a stronger supply chain supply capacity, higher efficiency of business operation, and the clustered area can provide a better business environment and customer flow environment for the local happy farmhouses compared to non-clustered areas.
3. (3). In this study, the MaxDEA8 software was used to measure the super-efficiency using the CCR model. The analysis of the DEA super-efficiency model yielded specific scores for the 50 happy farmhouses (See Table 5), focusing on the 31 DMUs that were in a relatively efficient position, i.e., those whose original technical efficiency value was equal to 1. Six DMUs were found to have an efficiency value greater than 2. They are 039 (24. 9382912), 034 (16. 8759667), 022 (3. 987862), 040 (2. 983749), 021 (2. 764891), and 048 (2. 093847).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Results of performance evaluations.

https://doi.org/10.1371/journal.pone.0339582.t005

1. (4). In order to further find the direction of improving the operational efficiency of the happy farmhouse, this article uses PCA to reduce the dimensionality of input and output data from various decision-making units, and then through the way of quadrant diagrams presenting the spatial distribution of decision-making units, which can be divided into the four quadrants of ABCD. Quadrant A: high inputs, high outputs; Quadrant B: low inputs, high outputs; Quadrant C: low inputs, low outputs; Quadrant D: high inputs, low outputs. As shown in Fig 1. Most of the farmhouses in the sample cases are in the C phenomenon, i. e., low-input, low-output mode, which is also due to the business nature of farmhouses in this province. The farmhouse itself is a small business, the threshold of entry is relatively low, and the small yard in the farmer’s house can be opened for business after a little bit of organizing and cleaning.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Quadrant Analysis.

https://doi.org/10.1371/journal.pone.0339582.g001

4.2.1. Trends in the M-Productivity Index and its decomposition.

Table 6 reports the changes in the productivity index, technical progress and combined technical efficiency of happy farmhouses in the Huzhou region over the last 10 years. From 2014 to 2023, the overall average annual growth rate of TFP in the region was 3. 1%, but the fluctuation was significant, with the highest being 21. 5%, which occurred from 2022 to 2023; The lowest is −3. 7%, which occurred from 2019 to 2020. From 2014 to 2016, TFP showed a slow upward trend in stages, with an average annual growth rate of 2. 8%. From 2016 to 2019, TFP showed a clear upward trend, with an average annual growth rate of 8. 6%; From 2019 to 2021, TFP showed a downward trend, with an average annual growth rate of −3. 0%. From 2021 to 2023, TFP showed a clear upward trend. This result indicates that since 2014, under the extensive development model that emphasizes “scale expansion”, the investment in rural infrastructure construction in the Huzhou area has grown rapidly, and the number and location of rural leisure activities have quickly increased year-on-year. However, the M productivity index is less than 1, indicating that this growth is inefficient and lacks quality assurance; Since the proposal of rural revitalization in 2017, various policy measures favorable to the development of rural tourism have been introduced in Zhejiang Province, resulting in a significant increase in the number of rural tourism tourists. This has prompted investment in rural tourism in Zhejiang Province to shift towards a cluster development model, leading to rapid growth in operational efficiency. From 2016 to 2019, the M productivity index was significantly greater than 1, indicating that the growth of TFP during this period is guaranteed high quality. At the end of 2019, the epidemic’s impact significantly affected China’s entire tourism economy. There was no unexpected negative growth in the operation of rural tourism. However, as the epidemic subsided, the rise of short-distance rural tourism led to a rapid upward cycle in the operating performance of rural tourism.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Efficiency and Productivity Index of Happy Farmhouse in Huzhou.

https://doi.org/10.1371/journal.pone.0339582.t006

Further examination of the reasons for the change in TFP shows that the TFP improvement mainly stems from technological progress. Over the past 10 years, technological progress’s average annual growth rate has been 2.6%, contributing 83.9% to TFP and exhibiting the same trend as the M productivity index. With the implementation of the rural revitalization policy in 2017, investment and construction in rural infrastructure have been vigorously promoted, and the average annual growth rate of technological progress has significantly increased; After 2019, in addition to external factors such as the COVID-19, Zhejiang Province has basically achieved the transformation of new rural areas, completed the updating and replacement of basic hardware equipment in rural areas, gradually released institutional dividends. The speed of technological progress will decrease. Its contribution to TFP will also decrease accordingly.

4.2.2. TFP analysis of different business models.

Further dividing the management model of rural tourism into individual farmers, farmer participation and company clusters, the TFP changes of 50 rural tourism enterprises in the Huzhou area from 2014 to 2023 were calculated and compared, as shown in Table 7. Overall, the company cluster type has the highest mean value of TFP (1,058), and the individual farmer type has a lower TFP (0,998). The evolution of the above pattern differences can be roughly divided into two stages: In 2014–2016, the TFP of all three regions declined, and the difference was not noticeable; in 2016–2019, the TFP of each area was amid fluctuation and increase, and the regional difference was gradually expanding. These changes may be attributed to the large-scale and rapid investments made by local governments with the help of land finance funds, leading to the widespread duplication of infrastructure construction as well as the problems of focusing on inputs rather than management and care, and development rather than protection; after 2017, the construction of urbanization around the world has been transformed into high-quality development, and the gap between the efficiency of infrastructure investment in various regions has been enlarged alongside the general increase in efficiency.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. TFP analysis of different business models.

https://doi.org/10.1371/journal.pone.0339582.t007

Further decomposition of TFP shows that the average annual growth rates of technological progress for individual farmers, farmer participation, and company clusters are 0. 6%, 2. 8%, and 5. 3%, respectively, are higher than the average annual growth rates of their combined technological efficiencies of −0. 8%, −0. 1%, and 4%, respectively, suggest that the impetus for the growth of regional TFP comes mainly from technological progress. Among them, technological progress in company cluster types has the most obvious role in promoting TFP, which is because their higher level of economic development gives them a greater advantage in management innovation, technological progress, and the introduction of advanced equipment and talents in urban infrastructure investment; while the continuous deterioration of the scale efficiency of the individual farmer type has led to the largest annual decline in the comprehensive technical efficiency, which may be attributed to the weak financial resources of the government of the individual farmer type areas, coupled with the low level of marketization, the public finance budget is mainly used to maintain recurrent expenditure (basically “food finance”). The participation of social capital is low. This may be because the governments of individual farmer-type areas have weak financial resources. The public finance budget is mainly used to maintain recurrent expenditures (basically belonging to the “food finance”), coupled with the low level of marketization and the low degree of participation of social capital, and the funds for the construction of urban infrastructure have been seriously insufficient for many years, so that the improvement of the efficiency of the scale of investment has been weak.