1 Introduction

Sub-Saharan Africa (SSA) has continued to rely heavily on human power in agriculture, and the use of farm machinery remains the lowest in the world (Diao et al. 2020). There have been some attempts to promote agricultural mechanization in SSA, notably during the 1960s and 1970s when several countries in the region, including Tanzania, launched state-sponsored mechanization projects (Pingali 2007).Footnote 1 Although these projects contributed to an increase in the number of potentially available tractors, due to poor management, most failed to realize the objective of widespread tractor use. As a result, the farming system in the region remained far from intensive, and sustainable demand for mechanized tillage was not generated (Pingali et al. 1987).

During the 1970s and 1980s, agricultural experts and development agencies shifted their attention to supporting the development of draft animal-powered implements and low-power mechanical equipment, which were seen as appropriate labor-saving technologies for smallholder farmers in SSA (Binswanger and McIntire 1987; Binswanger and Pingali 1989; Ruthenberg 1980). However, there was no widespread adoption of these technologies. By the 1990s, the number of tractors, as well as interest in investment in draft animal traction, had declined significantly in the SSA. This was in contrast to other developing regions, such as Southeast Asia and South Asia, which saw a significant increase in mechanization over the same period (Aryal et al. 2019; Belton et al. 2021; Paudel et al. 2019).

In the past three decades, however, two important trends have emerged related to mechanization in SSA. First, there has been renewed interest in agricultural mechanization in SSA, driven by wage growth, high food demand due to urbanization, and growth in the rural non-farm economy as discussed in Chap. 7. In particular, the use of tractors to perform power-intensive tillage activities is becoming common in some parts of SSA (Diao et al. 2020). In some areas, custom tractor hire services have emerged, where tractor owners, mostly large- and medium-scale farmers, provide tillage services to smallholder farmers (Daum et al. 2021). Some governments consider custom tractor hire services as an essential tool in the mechanization of small farms and have stepped up their efforts to reduce bottlenecks in the dissemination of tractors and spare parts (Adu-Baffour et al. 2019; Diao et al. 2014; Takeshima et al. 2013).

Second, while tractor use shows an upward trend, the number of draft animals in SSA has been declining since the late 1990s (Baudron et al. 2015; Kirui and von Braun 2018). Although draft animals are concentrated in a few areas where the prevalence of the tsetse fly (a vector of trypanosomiasis) is low, they have been playing a significant role in tillage and transport (Sims and Kienzle 2017).Footnote 2 The decline in draft animals is mainly caused by frequent epidemics of animal diseases and severe drought, which has reduced the availability of feed grasses (Mrema et al. 2008). Other factors include the shrinking of communal grazing areas, which leads to increases in the cost of keeping draft animals (Mrema et al. 2020; Tegebu et al. 2012). Recently, there has been a call to gradually transition from draft animals to mechanized tillage (Kormawa et al. 2018).

Despite these trends, the question of how tractors can be beneficial to the majority of smallholder farmers in SSA remains. Generally, tractors have been known for facilitating the expansion of cultivation areas by converting the fallow land to cultivable land (extensification) and improving labor productivity by reducing the labor required for crop production per unit area (Pingali 2007; Pingali et al. 1987). Although some positive effects of tractor use on farm extensification have been reported in recent years in SSA (Adu-Baffour et al. 2019; Jayne et al. 2019), the effects tend to vary with the farming systems (Takeshima et al. 2013). One strand of empirical evidence also suggests that the mechanization of land preparation can induce an increased application of labor-intensive, yield-enhancing agricultural technology (intensification), resulting in higher land productivity (Baudron et al. 2019; Mano et al. 2020; Nin-Pratt and McBride 2014; Takeshima and Liu 2020). However, some studies did not find distinguishable effects of tractor use on land productivity (Benin 2015; Houssou and Chapoto 2015). It is also notable that studies comparing different types of tools, such as large four-wheel tractors (4WT), small two-wheel tractors (2WT), and draft animal power (DA), remain scarce. Given the diverse agroecological conditions and farming systems in SSA, rigorous analyses of the effects of tractor use are crucial in identifying and implementing appropriate mechanization policies (Kormawa et al. 2018).

In this chapter, we use two-year panel data collected from rice farmers in Tanzania, one of the major rice producers in SSA, to discuss the effects of tractor use on the intensification and extensification of rice farming. Tanzania was among the SSA countries with comparably high use of tractors, as we will discuss in Sect. 9.2. Farmers can access different types of tractors, mainly through custom tractor hire services provided by private machinery operators. This practice is becoming common, especially in maize and rice farming areas. This allows us to examine the differential effects of 2WTs and 4WTs, DAs, and handheld tools (HT) such as hand hoes. We find that the use of 2WTs for land preparation significantly increases the adoption of transplanting in rows, improved modern varieties, and chemical fertilizers, resulting in an increase in paddy yield compared to the use of DAs. The positive effects of the 2WTs on rice productivity might be due to their effectiveness in puddling, thereby increasing the plant's ability to absorb nutrients from the soil (Sharma and de Datta 1985). On the other hand, we found little evidence that 4WT use contributes to the intensification or extensification of rice farming. Our findings contribute to the growing mechanization literature in SSA by showing that two types of tractors might play different roles in rice cultivation.

The rest of the chapter is organized as follows. Section 9.2 offers general information about rice cultivation and tractor use in Tanzania, whereas Sect. 9.3 presents details about the study site and data collection method. Section 9.4 presents descriptive analyses, and Sect. 9.5 explains estimation methods. We discuss our estimation results in Sect. 9.6 and offer some conclusions in Sect. 9.7.

2 Rice Production and Tractor Use in Tanzania

2.1 Rice Production

Rice production in Tanzania has increased from 1.76 million tons of milled rice equivalent in 2010 to 3.03 million tons in 2020, overtaking Madagascar to become the second largest rice producer in SSA after Nigeria (FAO 2022). The increase is partly in response to the growing demand for rice, especially in urban areas, where consumers prefer rice to other traditional staples. Increasing rice production in Tanzania is considered important because the country has a high per capita rice consumption. There is a strong preference for domestically produced rice over imported rice, which is mostly of low quality (Lazaro et al. 2017). Furthermore, rice produced in Tanzania is exported to neighboring countries, including Rwanda, Uganda, Kenya, Zambia, Malawi, and the Democratic Republic of Congo (Sekiya et al. 2020).

Rice in Tanzania is produced in three agroecological zones: the Lake Zone located in the northwestern part of the country, the Eastern Zone, and the Southern Highlands zone. About 70% of the total production in the country is undertaken in five administrative regions, namely Morogoro and Pwani in the Eastern Zone, Mbeya in the Southern Highlands Zone, and Tabora and Shinyanga in the Lake Zone. About 70% of the land suitable for rice cultivation is situated in rainfed lowlands ecosystems, with the remainder either in the rainfed uplands or irrigated ecosystems. Production techniques tend to vary across the country as farmers adapt to the agroecological environment around them (Sekiya et al. 2020).

2.2 The Trend of Tractor Use in Tanzania

Tractors were first introduced to Tanzania in the 1940s by the colonial government as part of economic recovery programs following World War II (Pingali et al. 1987). The statistics from FAO in Fig. 9.1 show that, by the time Tanzania became independent in 1961, there were 16,550 operational tractors.

Fig. 9.1
A vertical bar graph shows the number of tractors in use versus years. The plot dips in 1990 and increases by 2000.

(Source FAO 2021)

Changes in number of tractors in use in TanzaniaFootnote

FAO data on tractors in use in Tanzania are available up to 2002. They are based on data reported by the country in FAO questionnaires, official government reports as well as FAO estimates.

Full size image

When the government implemented a socialist and self-reliance policy, the number of tractors declined significantly from 17,000 units in 1970 to about 8,000 units in 1985 (FAO 2022).Footnote 4 Under the policy, private enterprises, including large-scale private farms, were nationalized, and millions of rural residents were relocated to communal settlements (Owens 2014). Incentives to increase production in these farms were low mainly due to centralized price control and poor management of communal production activities. By the early 1980s, it was becoming clear that most state-sponsored projects, including mechanization, were costly and unsustainable, and the government was unable to support them due to budget deficits. In 1986, Tanzania accepted an offer to transform its economic policy toward economic liberalization under the Structural Adjustment Program (SAP).

Although tractor usage remained low in the early years after joining the SAP, policies adopted during that time laid the foundation for subsequent growth in tractorization rates. Among others, the liberalization of the financial sector in 1991, Village Land Act of 1990, and new investment policies were crucial in attracting private investment in agriculture and contributed to the increase in the number of tractors after the mid-1990s.

Since then, there has been an upward trend in tractorization in Tanzania, partly due to conducive conditions for investing in agricultural sector and the emergence of manufacturers of affordable tractors—particularly 2WTs. The 2WTs were first introduced in Tanzania in the early 2000s as part of a policy aiming to encourage the use of appropriate mechanization (Agyei-Holmes 2016). Since then, the number of 2WTs grew rapidly, making Tanzania one of the countries with the largest 2WT fleets in SSA.Footnote 5 4WTs have also been increasing, but at a slower pace compared to 2WTs (Mrema et al. 2020). Tractors operating in Tanzania are mostly imported from Europe and Asia by private entrepreneurs and well-off farmers, but the process of importing only a few tractors at a time is considered to be costly. For new tractors, the government has intervened periodically by importing them in bulk and then selling them to private owners at subsidized prices, while second-hand tractors are traded in private markets without subsidies.

The use of 4WT for tillage is common in maize and rice farming system, while 2WTs are commonly used in rice farming as they are not sufficiently powerful to use on upland fields, where the soil is harder than in paddy fields. Smallholder farmers can hire a tractor (4WT or 2WT) from private service providers, and contracts are made based on the market-determined piece rate per unit of land. Although studies on the supply side of mechanization in Tanzania remain rare, our field observations tell us that tractor hire service providers move from one area to another as cultivation seasons of rice and other crops, such as sunflowers and cowpeas, tend to differ across the locations.

3 Study Site and Data

To examine the impact of machinery on rice farming, we use a part of a Rice Extensive Survey conducted in three administrative regions, namely Morogoro (Eastern Zone), Mbeya (Southern Highland Zone), and Shinyanga (Lake Zone). In each region, we selected two rice-growing districts, Kilombero and Mvomero in Morogoro Region, Kahama and Shinyanga Rural in Shinyanga Region, and Mbarali and Kyela in Mbeya Region. We used information from the 2002–2003 agricultural census to determine the number of villages to be covered in each district and randomly selected a total of 76 villages in all six districts. In each village, we randomly selected 10 rice-growing households, making a total sample size at the baseline survey of 760 households.

During the interview, we observed that farmers grow rice on multiple plots. We asked each household to identify the rice production plot that they considered to be their most important (hereafter referred to as a sample plot) so that we could collect information on one plot per household. We gathered detailed information about technology adoption, production costs, and rice productivity in the sample plot. In addition to the household-level survey, interviews were conducted with village leaders in all 76 villages, in which information about rice cultivation, mechanization, and access to public services and markets was collected. We revisited and interviewed the same households in 2018. For this second round of surveys, we interviewed a randomly selected replacement household if the household interviewed in 2009 could not be traced.

Although we intended to analyze the impact of mechanization using the total sample, we excluded observations from Kahama, Shinyanga Rural, and Kyela districts after finding that about 90% of surveyed farmers used draft animals to prepare their rice plots, and the use of tractors in these districts was low. One of the reasons for this is that most households are agro-pastoralists, who grow rice and other crops during the main rain season and graze their livestock herds in plains near the settlements. Therefore, we use data only from Kilombero, Mvomero, and Mbarali. After dropping outliers, our data becomes unbalanced two-year household-level panel data with a total of 662 observations.

Since we could not reinterview a significant number of our original sample households in the endline survey in 2018, our results may suffer from attrition bias. To examine this, we estimate the attrition probit model using the 2009 observations, where the dependent variable is the dummy variable taking one if the household is attrited in the endline, and the independent variables are baseline household characteristics. We find that there are no household characteristics that significantly influence the probability of being attritted, suggesting that the attrition could be considered random to some extent. We also find that farmers in Kilombero and Mvomero districts were significantly less likely to be attritted than those in Mbarali district. Given that our sample includes replacement households, however, handling attrition is not simple. Although we admit that attrition is not fully controlled, we mitigate this problem by adding district-fixed effects.

In our study sites, there are two cultivation seasons; the main season from October to June and the dry season from July to September. During the main season, farmers grow rice in the rainfed and irrigated lowland plots and other crops such as maize in the upland plots. During the dry season, lowland rainfed plots are usually left to fallow due to water scarcity, and irrigated plots are used to grow rice, vegetables, and other crops depending on water availability. As only a few plots, if any, are used for rice cultivation during the dry season, our analysis focuses on rice cultivated in the sample plot during the main season.

Farmers in our study sites use 4WTs, 2WTs, DAs, or HTs to perform land preparation activities, including plowing, harrowing, and puddling.Footnote 6 The choice of implements is determined by several factors, including accessibility, soil characteristics, and farmers’ socioeconomic characteristics. 4WTs, for example, are often used for plowing and harrowing in rainfed lowland plots before they are flooded. They are also preferred for plowing and harrowing in plots with heavy clay soil since their engines are powerful. 2WTs are commonly used for puddling in standing water conditions in irrigated or rainfed rice plots surrounded by bunds. 4WTs fitted with special implements can also be used to perform puddling in standing water conditions, although in many areas, this is rare due to bogging or sinking in muddy fields. Thus, 2WTs, which are much lighter, have the advantage in puddling.

As some rice farmers in our sample utilized more than one implement for preparing their plots, we generate four mutually exclusive dummy variables—namely, 4WTs, 2WTs, DAs, and HTs—to simplify our analyses. Out of 662 observations in our sample, 22 farmers used 4WTs along with 2WTs, 16 used 4WTs with DAs, and 49 used 4WTs with HTs. After close examination, we categorize these farmers as 4WT users. Even when we categorize the 22 farmers who used 4WTs and 2WTs as 2WT users, our estimation results remain largely the same. Similarly, the 35 farmers who used 2WTs with DAs and 49 with HTs are categorized under 2WTs, and 73 farmers who used DAs with HTs are categorized under DAs. Note, however, that the results of our main analyses are largely the same, even if we allow for the use of multiple instruments by one household.

Tractors (4WTs and 2WTs) are usually hired from private operators, where agreements between the two parties depend on the size and condition of the plot. Payment is made before or immediately after the work is completed. The hiring of DAs for plowing also follows such an agreement, although most farmers use their own cattle. During the study period, there were no cases of farmers in our sample who used their own 4WTs, and just 13% of those who use 2WTs (about 2% of the active sample of 662 observations) own them.

4 Descriptive Analysis

Table 9.1 shows changes in farm appliances used by farmers in our sample to prepare rice plots between 2009 and 2018 (Panel A) and other variables related to tractor access and related village-level variables. The data show that, in general, the use of tractors for land preparation has increased over time, while the use of DAs and HTs has decreased. The percentage of farmers who used 4WTs to prepare their rice plots increased from 34.2% in 2009 to 46.4% in 2018, while users of 2WTs increased from 7.6 to 22.3% during the same period. The increase is in line with data presented by Mrema et al. (2020), showing that 2WTs have contributed substantially to the trend toward increased tractor use in Tanzania. Conversely, the use of DAs decreased somewhat, declining from 18.5% to 14.2%, and the use of HTs fell from about 39.7% to 17.17%.

Table 9.1 Accessibility and use of tractor hire services and related variables
Full size table

Panel B reports a significant increase in the number of tractors owned by villagers, particularly 2WTs. In 2009, the average number of 4WTs and 2WTs per village was merely 2.3 units and 0.7 units, respectively. In 2018, the average number of 4WTs per village increased slightly to 3.4 units, while the number of 2WTs increased to 8.2 units. The increase in the number of tractors and the decline in machinery hire fees suggest that tractor hire services are becoming more accessible and affordable. The fees for hiring DAs also decreased to a large extent, but their use also declined. Despite the fact that DAs are cheaper than machinery, farmers may prefer machinery because they are more time efficient than DAs. Panel B also shows that the village-level population density also increased from 142 to 165 persons per square kilometer between 2009 and 2018, suggesting increased availability of labor and demand for food in rural areas.

To fully understand how farmers transit from one source of power to another, we conducted a further descriptive investigation. To make it easier to identify the transition path, we use 236 households with balanced panel household-level observations. Table 9.2 shows how farmers changed the implements they used for land preparation between 2009 and 2018. Bolded numbers show farmers who used the same implements between the two periods. Out of the farmers who used DAs in 2009, only about 36% remained in the same category in 2018, while others shifted to 2WTs (35.7%), 4WTs (19.1%), and HTs (9.5%). Similarly, 36% of HTs users in 2009 remained in the same category, while others shifted to 4WTs (33%), 2WTs (16.5%,) and DAs (14.3%). These trends indicate a shift away from labor-intensive and time-consuming implements to tractors.

Table 9.2 Power substitutions over time from 2009 to 2018
Full size table

In Table 9.3, we compare rice cultivation based on farm implements used to prepare the rice plots. We stratify our sample on whether the sample rice plot was irrigated (Panel A) or rainfed (Panel B). As indicators for extensification, we focus on the total area under rice cultivation at the household level and the size of the area under rice cultivation within the sample plot. Farmers sometimes cultivate only a part of the plot and leave the remaining part under fallow for various reasons, including labor shortages, the use of labor-intensive cultivation methods, and lack of sufficient water. Therefore, the actual area under rice cultivation within a plot can be smaller than the size of the plot. In such environments, the use of tractors is considered to help farmers prepare a larger area of each plot within a short time. Regarding the area under rice cultivation at the household level, it is important to note that we collected data on machinery use only in the sample plot, and we do not have similar data for other rice plots. Therefore, when the area is measured at the household level, our machinery use variables are prone to measurement error.

Table 9.3 Cultivated area, technology adoption, and paddy yield by farm implements used for land preparation
Full size table

As indicators for intensification, we focus on the adoption of yield-enhancing technologies and paddy yield. The key technologies include transplanting in rows and the use of modern rice varieties (MVs), chemical fertilizer, and insecticides and herbicides.Footnote 7 The modern variety widely available in Tanzania is TXD 306, commonly known as SARO 5. This variety has high yield potential, particularly in irrigated ecosystems, and has some characteristics that are preferred by consumers, such as the aroma inherited from traditional parental varieties (Nakano et al. 2018; Sekiya et al. 2020).

We conduct a t-test for comparisons between 4WTs (Column 1), 2WTs (Column 2), and HTs (Column 3), against the reference category, DAs (Column 4). In order to compare the effectiveness between machinery and draft animal power, which can be substitutable without substantially affecting paddy yield —as discussed in Chap. 7—we keep the DAs as a base category. Regarding the irrigated lowlands, we find that the areas cultivated at the household level and within the sample plot by 4WT users do not significantly differ from DA users. 4WT users, however, have significantly higher adoption rates of the modern rice variety, and they apply more chemical fertilizers, insecticides, and herbicides than DA users. Users of 4WTs achieve a paddy yield of 4.0 tons per hectare, which is higher than the 3.3 tons per hectare of DA users. Similar to 4WT users, farmers who use 2WTs also do not cultivate significantly larger areas than DA users, but they have significantly higher adoption rates of all key rice technologies than DA users. Moreover, the users of 2WTs achieve an average paddy yield as high as 5.1 tons per hectare, which is significantly higher than all other farmers. Regarding the HT users, we find that they have high adoption rates of key technologies and cultivate relatively small areas at both the household level and within the sample plot, achieving an average paddy yield of about 3.8 tons per hectare.

In the rainfed lowlands, there is no significant difference between 4WT and DA users regarding cultivated areas, adoption rates of MVs, and chemical fertilizer application. Although users of 4WTs have a lower adoption rate of transplanting in rows and application of insecticides and herbicides than DA users, we find no significant difference in paddy yield between the two. This finding supports the hypothesis of Alsan (2015) that the unavailability of DAs is a major constraint on the intensification of farming in many areas in SSA. On the other hand, we find that farmers who use 2WTs have a higher adoption rate of transplanting in rows, the use of MVs, and chemical fertilizer use than DA users, achieving a high paddy yield of 4.3 tons per hectare under rainfed conditions. This is remarkable because this yield is comparable to the average yield in tropical Asia as well as in India (see Chap. 1). We also find that HT users cultivate small areas and have low adoption rates of MVs and application of insecticides and herbicides, but there is no significant difference in paddy yield between HT users and DA users.

Our descriptive analyses suggest that, although the use of tractors, in general, may not demonstrate a clear advantage over DAs in the extensification process, the use of 2WTs is likely to be more beneficial than DAs in enhancing technology adoption and paddy yield.

5 Estimation Methods

To examine the effects of mechanization of rice cultivation, we employ the fixed effects (FE) model specified as follows for the household \(i\) in year \(t\):

$${y}_{it}=\alpha + {\beta }_{1}{HT}_{it}+{\beta }_{2}{2WT}_{it}+{\beta }_{3}{4WT}_{it}+{\theta }_{1}{T}_{t}+{{\varvec{X}}}_{it}\eta +{c}_{i}+{u}_{it,}$$
(9.1)

where \({y}_{it}\) denotes the aggregate area under rice cultivation at the household level, the area under rice cultivation within the sample plot, technology adoption rates (including transplanting in rows, modern varieties, chemical fertilizers, insecticides, and herbicides), and paddy yield. The main explanatory variables of interest are \({HT}_{it}\), \({2WT}_{it}\) and \({4WT}_{it}\) which are binary variables that respectively indicate whether the household used HTs, 2WTs, or 4WTs in year \(t\). As discussed above, we define our machinery use variable to be mutually exclusive and we collected the data on one sample plot for each household. Thus, although we measure yield, technology adoption, and machinery use at the plot level (i.e., in the sample plot), we put subscript i (household) for these variables.

We control for several time-varying plot-, household-, and village-level characteristics (denoted by a vector \({{\varvec{X}}}_{it}\)), including the number of working-age adults, years of schooling of household head, female-headed household (dummy), age of household head, total landholdings (ha), the value of non-farm household assets (million TShs), amount of credit received by the household (‘00,000 TShs), size of the sample plot (ha), dummy variables indicating whether the sample plot is irrigated, has clay soil, or has bunds, as well as a time dummy variable \({T}_{t}\), which takes the value of 1 if year is 2018. \({\beta }_{1}\), \({\beta }_{2}\), \({\beta }_{3}\),\({\theta }_{1}\), and \(\eta \) are parameters to be estimated, while \({c}_{i}\) and \({u}_{it}\) respectively denote unobserved time-invariant household characteristics and the error term. The key parameters of interest are \({\beta }_{1}\), \({\beta }_{2}\), and \({\beta }_{3}\), respectively, denoting the effects of using HTs, 2WTs, and 4WTs relative to the reference category of DA.

For a robustness check, we employ the correlated random effects approach (CRE). The CRE, which is based on the random effects estimator, adjusts for time-invariant unobserved heterogeneity by including averages of time-varying household-level explanatory variables (referred to as MC devices) as additional regressors in the random effect model (Wooldridge 2019). We use similar sets of dependent and independent variables as in the FE model and include district-fixed effects and the MC devices instead of household FE.

6 Estimation Results and Discussion

Table 9.4 presents the estimation results for the impact of mechanization. Panel A shows the results of FE estimation followed by that of the CRE model in Panel B. In each estimation, we report the robust standard errors clustered at the village level in parentheses. To shorten the presentation of our findings, we report only the estimation results of key coefficients and exclude those of the control variables. We obtain similar results for both FE and CRE estimations, suggesting the robustness of our results. In addition, although our analytical framework assumes that each farmer’s choice of mechanization is mutually exclusive, some farmers use multiple implements as mentioned earlier. We therefore conducted further analysis to examine whether the estimation results would change if we allowed the choice of multiple types of mechanization in one household. For this purpose, we estimate the FE and CRE models using the dummy variables that take one as long as a farmer uses each instrument. Although the results are not shown, the estimates are consistent with the results presented in this chapter.

Table 9.4 Impacts of tractorization on cultivated area, technology adoption, and paddy yield
Full size table

Results of CRE estimation show that 4WT users cultivate a significantly larger area than DA users. 4WT users are significantly more likely to use MVs but less likely to adopt transplanting in rows. These significant effects, however, are not observed in FE estimation. We do not find any significant effects of HT or 2WT use on the areas of cultivation at the household level or within the sample plot relative to DA use. These results suggest that neither 2WTs nor 4WTs have a strong advantage in the extensification of rice cultivation compared to DAs. However, regarding the adoption of yield-enhancing technologies, our estimation results of both FE and CRE modes show that the use of 2WTs is associated with an increased adoption rate of transplanting in rows, modern varieties, and the application of chemical fertilizer. Specifically, the use of 2WTs significantly increases the adoption of transplanting in rows by 21 percentage points, modern varieties by 31 percentage points, and the application of chemical fertilizer by about 39 kg per hectare relative to the use of DAs. The use of 2WTs is also positively associated with an increase in paddy yield of about 1.1 tons per hectare, possibly due to the high adoption rates of yield-enhancing technologies. In addition, the use of HTs is associated with an increase in modern varieties and a decrease in the use of insecticides and herbicides. The use of HTs also increases the paddy yield by 1.0 tons per hectare compared to the use of DAs. In contrast, we do not find any significant effect of using 4WTs on rice technology variables.

Although the exact mechanism by which the use of 2WTs leads to input intensification and paddy yield remains unclear, it may be linked to the effectiveness of 2WTs in puddling in muddy paddy fields. As we discussed earlier, although any implement can be used to perform this activity, 2WTs are considered the most effective for puddling partly because they are lightweight and easily maneuverable in small paddy plots. Effective puddling by 2WTs could increase nutrient uptake by plants, and thus, increase the performance of yield-enhancing technology (Sharma and de Datta 1985). This agronomic observation is consistent with our findings that almost all 2WT users used them for puddling. It must also be emphasized that 2WTs are as efficient as 4WTs—and probably more efficient than DAs—in plowing and harrowing if the soil is not very hard. In sum, our analysis suggests that 2WTs are the appropriate technology for rice farming intensification in many areas in Tanzania. This is consistent with the rapidly increasing adoption of 2WTs reported in Table 9.1.

7 Conclusion

In this chapter, we examined the effects of mechanization on rice production and productivity using two-year panel data collected in Tanzania, one of the major rice producers in SSA. Specifically, we investigated whether mechanization of land preparation activities using 2WTs or 4WTs results in the expansion of the cultivated area and increasing use of yield-enhancing technologies compared to Das. Conducting this study in Tanzania, where DAs are widely used in rice cultivation, allows us to compare all four types of implements used to prepare the rice plots, including HTs. We estimated the effects of using HTs, 2WTs, and 4WTs on rice cultivation practices and yield using FE and CRE estimation methods.

Overall, we found that the adoption of 2WTs contributes to the adoption of transplanting in rows, modern rice varieties, and chemical fertilizer application, resulting in high paddy yield compared to the use of DAs. On the other hand, we find that the use of 4WTs does not have significantly different effects on rice cultivation compared with the use of DAs. Results of our analyses suggest that the effects of tractorization on rice intensification may differ depending on the type of tractors used. Our results are partly consistent with the case of Côte d'Ivoire reported in Chap. 8, which demonstrates the positive relationship between the use of 2WTs and intensification.

Our estimation results showing that 2WT tractors play a significant role in the intensification of rice farming in SSA are consistent with observations in some Asian countries, including Bangladesh, India, and Nepal (Aryal et al. 2019; Belton et al. 2021; Paudel et al. 2019). Regarding the 4WTs, our estimation results indicate that they do not have a significant advantage over DAs concerning extensification, adoption of rice technologies, and paddy yield. The heavier weight and lower maneuverability of large machines in small muddy patches of paddy fields are a disadvantage for use in rice production. This is partly consistent with Pingali (2007), who argued that 4WTs may be beneficial over draft animals, but only if they contribute to a significant reduction in the hours of labor use required for land preparation.

Our results provide supportive evidence for the recent trend toward the promotion of small-scale mechanization, especially for the intensification of rice farming in SSA. Note, however, that our results should not be interpreted as “small is beautiful.” The results may be location-specific because of the different functions that large- and small-scale machinery (i.e. 4WTs and 2WTs) play in rice cultivation and because of soil conditions (Daum et al. 2022). For example, 4WTs are used for plowing and a pair of oxen are used for puddling in Mwea Irrigation Scheme in Kenya, where the soil is vertisol, which is particularly hard. Further investigation is needed to identify the conditions under which 2WTs are particularly useful for intensifying rice farming in SSA.