• No results found

Instrumental Variable Approach

The 1999 Village Land Act changed de-jure rights for all Tanzanians, but as described above indi-vidual certification was an endogenous decision, making simultaneous causation the main threat to identification. One possibility to circumvent this is to use some kind of exogenous variation that is related to individual titling status. Specifically, I use the average CRO share per district and year as an instrumental variable for household’s certification status in Equation 1. The district share is chosen both because implementation of the land reform was a district-wide process (Fairley 2013) and to increase instrument sample size.

The proposed instrument is relevant. A change in the district level CRO share is indicative of the progress of land formalization and the individual ability to secure tenure. In fact, a household is only able to secure certification if spatially more aggregated requirements are satisfied, such as the estab-lishment of a district land registry. Fairley (2013) describes that the estabestab-lishment of a district land office is the first point of entry, as it constitutes the primary land registry, processes CROs and keeps copies of certificates. Table 8 summarizes the first stage of regressing the instrument on the house-hold’s CRO share. It is important to exclude a househouse-hold’s own certification status in constructing the instrument to ensure that the instrument is not correlated with the household-level error term.

Thus, I calculate the CRO share per year and district for each household by omitting this specific household from the calculation. Another technical aspect in this regard is that standard errors are clustered on the district level to avoid serial correlation from omitted variable bias within districts.

Furthermore, the group fixed effect is on the district rather than the household level. The coefficient of the First Stage shows a strong correlation. This is highly significant and the F -statistic with 16.36 well above the conventional threshold for valid instruments.

The instrument is also exogenous. As outlined in Section 1.2, the Village Land Act requires imple-mentation on the district level first before villages can be titled, before in turn individual titles can be issued. Surveying, registering and land use planning on the district level was largely project driven;

districts did not signal willingness to have their land surveyed but mainly international donors and NGOs decided on the schedule of implementation efforts as funding was a key constraint for realiza-tion (Pedersen 2010). The World Bank project as the biggest one of its kind included fifteen districts

that were specifically targeted to implement the land reform. They were chosen to capture the diver-sity of livelihoods among Tanzanian villages and to analyse the impact of the land reform in different contexts (Fairley 2013). Selection of districts in which implementation activities were supposed to take place were also of practical nature, based on availability of basic infrastructure (Schreiber 2017).

Overall, a variety of international institutions engaged in implementation activities, each of which covered a limited number of districts, leaving implementation of the land reform to appear scattered and random (Pedersen 2010). This variation in village registration and subsequent CRO issuance across districts and time provides a potential instrument to identify the causal effect of certification on agricultural activities (Galiani, Gertler & Schargrodsky 2005).

Table 8: Instrumental Variable - First Stage

CRO Share Leave-Out-Mean (LOM) 0.5651∗∗∗

(0.0665)

Household Size 0.0009

(0.0011)

Assets 0.0001

(0.0002)

Quality -0.0022

(0.0014)

Slope 0.0024

(0.0016)

Distance -0.0001

(0.0001)

Constant -0.1223

(0.0537)

District FE Yes

Time FE Yes

R2 0.08

F 16.36

Observations 9,300

* p < 0.10, ** p < 0.05, *** p < 0.01; robust and clustered standard errors in parentheses. Note:

Table shows the first stage of regressing individ-ual CRO share on LOM, which is the leave-out-mean of the average CRO share per district and year. Standard errors are clustered on the district level and the group fixed effect is the district.

A third question is whether the exclusion restriction holds. The instrument should not affect outcomes directly or indirectly other than through the endogenous regressor. Changes in district titling might directly affect agricultural behaviour, for instance by a changing prevalence of land disputes. On the one hand, formalization can resolve conflicting claims and create clarity for land owners. In turn, a farmer might increase investments through improved perceived tenure security but not because she let her own plots title, but everyone around her has certified land. On the other hand, official

demarcation may exacerbate salient land disputes and create more uncertainty. This might not hamper investments but induce visible investments primarily into trees or agricultural machinery.

Angrist (2014) questions the applicability of group average outcomes as a causal interpretation for individual outcomes, also if endogeneity concerns are weakened by omitting individual behaviour from group averages. Troubling is that households from the same district are similar in many ways, almost certainly including aspects of their behavior captured by a household’s CRO share. Angrist (2014) shows that with spillovers the exclusion restriction can be violated and Two-Stage Least-Squares (2SLS) estimates might exceed OLS results. Looking at the results of 2SLS in Table 9 suggests that this is indeed the case. The estimates should thus be interpreted with care. Nonetheless, the direction and significance of effects provides valuable insights in comparison with TWFE and Dynamic DiD.

Table 9: Instrumental Variable Estimates Maize Yield

per Acre (log)

Tree Seedlings per Acre

Investment

per Acre (log) Security Credit Value per Acre (log)

(1) (2) (3) (4) (5) (6)

CRO Share 0.8951 351.6 0.3852 0.2673∗∗∗ 0.1780∗∗ 3.050∗∗∗

(= LOM) (0.4701) (208.2) (1.266) (0.0945) (0.0876) (0.6212)

Household Size -0.0129 -1.603 0.0309 0.0068∗∗∗ -0.0002 -0.0299∗∗∗

(0.0075) (2.381) (0.0208) (0.0016) (0.0015) (0.0074)

Assets 0.0005 -0.4610 0.0203 0.0005 0.0004 0.0030

(0.0048) (1.818) (0.0107) (0.0005) (0.0004) (0.0040)

Quality 0.0033 0.9580 0.0273 0.0014 -0.0005 0.0369∗∗∗

(0.0082) (7.930) (0.0185) (0.0012) (0.0007) (0.0131)

Slope 0.0035 6.638 0.0202 -0.0001 -0.0009 0.0248∗∗∗

(0.0078) (8.766) (0.0194) (0.0012) (0.0007) (0.0080)

Distance 0.0010 4.483 0.0206∗∗∗ -0.0032∗∗∗ 0.0000 -0.0003

(0.0024) (3.817) (0.0070) (0.0007) (0.0004) (0.0026)

Constant 7.1813 54.23 6.982 1.284 -0.2884 15.24

(1.6699) (278.6) (3.891) (0.3013) (0.2013) (1.082)

District FE Yes Yes Yes Yes Yes Yes

Time FE Yes Yes Yes Yes Yes Yes

Adjusted R2 0.00 0.01 0.00 0.00 0.01 0.03

Observations 9,300 9,300 9,300 9,300 9,300 9,300

* p < 0.10, ** p < 0.05, *** p < 0.01; robust and clustered standard errors in parentheses. Note: Table shows Two-Stage Least-Squares estimates where individual CRO share is instrumented by the leave-out-mean CRO share per district and year, based on Equation 1. Standard errors are clustered on the district level and the group fixed effect is the district.

In line with the main results and preceding robustness checks, households with a CRO significantly increase maize productivity and tree plantation. As hypothesized, the magnitudes are considerably larger and do not permit a causal interpretation. Commensurate with the conclusions so far, per-ceived security and land values strongly benefit from titling. The instrumental variable approach also generates a significantly positive collateral channel, largely supporting the evidence from other methods. Most strikingly, in contrast to TWFE, but in line with the findings from Dynamic DiD,

instrumenting individual CRO share does not reveal a significant increase in investments. Although the overall effect is positive, it is far from significance and of lower magnitude than what the TWFE results in Table 3 suggest. Given the likely upward bias in the 2SLS estimates cements the argument that the investment response from securing tenure is not significant and most likely suffers from endogeneity.

Discussion & Conclusion

In contrast to many studies on land reforms and their effects on rural development in Sub-Saharan African contexts, this empirical investigation draws more benevolent conclusions. The analysis shows that households that obtain certification have higher agricultural productivity and plant more perma-nent trees. The latter points to a more diverse food production in particular, both findings suggest the emergence of income-generating opportunities in general. The observed gains in land produc-tivity from titling join a small number of similar findings for African economies: Zambia (Smith 2004), Madagascar (Jacoby & Minten 2007) and Ethiopia (Holden, Deininger & Ghebru 2009). The latter is the only other evidence I am aware of that also finds a positive link with tree cultivation in Sub-Saharan Africa. Looking at the channels through which the results can be explained, I find robust evidence of increased tenure security and higher land values. Both conclusions are interesting in their own right. Some have argued that informal tenure arrangements are working well and as a consequence formalizing land claims is insufficient to induce agricultural development. Although tenure is relatively secure at baseline, certification adds a pronounced component to this. Descriptive evidence suggests that this might be more important to more recent land holders with less established informal tenure. Increased plot values raise the income households can extract from selling or renting out land, as well as the collateralizable stock. Increased leasing out of land following certification points to higher turnover in land factor markets with potentially large efficiency gains. This might have pronounced welfare implications, particularly for women who might gain in intra-household decision-making via increased joint titling of plots.

Evidence on a collateral channel is overall positive but not convincingly comprehensive. Both the propensity to take up a loan and the amounts borrowed just miss standard levels of significance.

Disaggregating heterogeneous results reveals that credit access following certification significantly increases for male headed households and low-income households. Women do not seem to benefit from improved collateralization at all. From an affirmative perspective, low-income households can use loans in order to fulfill subsistence needs, pointing to an equity enhancing effect. Given that the equity effects from land titling are oftentimes unclear, this is an important finding. More generally, the results are noteworthy since most empirics fail to detect a positive collateral channel all together (Panman & Lozano-Gracia 2021; Galiani & Schargrodsky 2010; Kerekes & Williamson 2010; Do &

Iyer 2008; Jacoby & Minten 2007; Torero & Field 2005; Place & Migot-Adholla 1998). However, they are not without a flip side. Further refinement of the analysis reveals that land certification does not increase loan uptake from official sources, such as commercial banks, mortgage lenders or micro-finance institutions. It is rather informal lenders who are willing to issue more loans at higher amounts. Especially the World Bank project as the largest-scale endeavour in implementing the 1999 Village Land Act was costly and enduring, but of limited quantitative impact when it comes to

commercial lending. Considering that both the government’s and international donors’ objective was to facilitate entrepreneurial activity by improving formal credit access through collateral, the impact twenty years after the act came into effect appears less promising. In this regard, Tanzania adds a cautioning example of considering land reforms as cost-effective tools to rapidly improve institutions and financial markets. If smallholder lending remains prohibitive due to banks considering agricul-tural income as too risky or seizing collateral as too costly because of strong informal ties within villages, then a change in de-jure property rights might not translate into general equilibria effects of economic development. Indeed, the 1999 Village Land Act legally protects village governments’

resistance to selling land to outside investors (Schreiber 2017; Stein et al. 2016), making foreclosure difficult. Moreover, certification was oftentimes promoted by giving individuals greater access to credit. If this is not the case but the major benefit communicated, perceived advantages from having a CRO might fall behind costs of obtainment and dissemination could stall overall, despite other beneficial effects shown here.

The results also indicate that obtaining certification is associated with increased short-term invest-ments, but unlikely to be a causal effect in the absence of exogenous treatment. Two-way fixed effects suggest a strong positive effect of securing tenure on agricultural investments. This holds even if one accounts for potential self-selection bias. This robust link is cast into doubt by alternative econometric methods that more appropriately address endogeneity concerns. A novel event-study methodology based on staggered treatment adoption enables examining pre-adoption behaviour of treated households, thereby visualizing the parallel trends assumption. To the best of my knowledge this is the first time this procedure is used in this way. The investment response turns insignificant when creating an artificial sharp treatment point to study behaviour from an event perspective. Also aggregating group-time average treatment effects in calendar time does not yield significant estimates.

A shortcoming is that the limited sample size of treated households leads to imprecision in temporally refined group-time effects in the dataset employed here. Nevertheless, I believe that, apart from the obstacles in two-way fixed effects it aims to overcome, Dynamic Difference-in-Difference estimation offers many opportunities to better understand dynamic and heterogeneous treatment effects. Instru-menting individual household’s CRO share with leave-out-mean CRO shares per district and year reinforces the doubt on a significant investment effect. Although the exclusion restriction appears to be flawed, Two-Stage Least-Squares should produce an upward bias. The investment outcome is still insignificantly positive. Households do purchase more fertilizer, pesticides and to a lesser extent seeds, but it is suggested that they do so in anticipation of certification and it does not seem to be a main driver.

More generally, the results question the validity of employing Two-Way Fixed Effects or as a reduced form Difference-in-Difference methods in this kind of setting if certification is endogenous. Random-ization conditional on group and time fixed effects is a strong assumption, and claiming causality seems unreasonable. A step forward might be to obtain detailed information on the progress of systematic adjudication. This approach was implemented in Tanzania to lower costs by generating economies of scale, replacing the more demanding spot adjudication. Whereas for the latter individual plots were registered, the systematic approach surveys land parcels of the entire village. This also reduces individual titling costs and facilitates access to certification. Besides, the Tanzania National Panel

Survey offers comprehensive data to study a variety of research questions related to land reforms and offers the opportunity to exploit a longitudinal dimension, rare for African countries. Its national and temporal coverage makes it a unique data source for Sub-Saharan Africa and suitable to gain insights into the workings of a large African economy. Yet I believe that, at least in the case of property rights, this dataset has only sporadically been used. A number of other interesting outcomes can be tested in a rigorous manner, such as changing entrepreneurial activity or non-farming labour participation.

Given that the Tanzanian government aims to issue millions of certificates and to draw-up land use plans for thousands of villages, future waves will make this even more insightful.

The positive effects on land productivity, profitable tree cultivation and the operating channels are in line with findings for Ethiopia (Deininger, Ali & Alemu 2011; Holden, Deininger & Ghebru 2009).

Deininger, Ali & Alemu (2011) suggest that in Ethiopia decentralized and participatory implemen-tation with an emphasis on the provision of information, the issuance of certificates rather than titles, and a focus on gender equality helped avoid some of the problems of land titling in Africa.

These arguments seem to equally apply to the case of Tanzania and should be given attention in policy-making. In addition, it appears that a major distinctive feature of many Sub-Saharan African economies is their weakly developed formal lending market. Tanzania’s mortgage market is one of the smallest in East Africa; in 2017, only 6% of the population had access to formal credit (Jones et al. 2016; World Bank 2019b). In the absence of functioning financial markets land reforms are likely to continue to underperform and may not automatically translate into large-scale economic development as some might hope. Further strengthening of commercial financial markets appears of vital importance to enable formalization of land rights to unfold their full potential. This analysis shows that institutions matter for economic development, but that reforming them continues to be a long-term challenge. Although the microeconomic literature on property rights and their impact on agricultural development has a long history, this kind of research question will certainly remain a heavily debated one in the future.

Appendix

Table A1: Probit Model for Probability of Attrition based on Baseline Characteristics in 2008 Attrition

(1 = Yes, 0 = No)

Region -0.00824

(0.00515)

Loss of Land -0.318

(0.399)

Household Head - Female 0.105

(0.126)

Household Head - Age 0.00388

(0.00323) Household Head - Village Indigenous -0.0242

(0.120)

Household Size -0.104∗∗∗

(0.0335)

Assets 0.00421

(0.172)

Consumption (log) -0.151

(0.120)

SACCO -0.0322

(0.282)

Total Area -0.0124

(0.0297)

Number of Plots 0.0199

(0.0508)

Distance from Dwelling 0.00358

(0.0177)

Soil Quality 0.101

(0.0822)

Slope -0.0549

(0.0555)

Yield per Acre (log) 0.000903

(0.0284)

Year possessed 0.0000910

(0.000672)

Constant 0.513

(2.188)

Pseudo R2 0.1100

Observations 2,657

* p < 0.10, ** p < 0.05, *** p < 0.01; robust standard errors in paren-theses.

Table A2: Covariate Relations of Independent Variables

CRO Share Assets HH Size Distance Quality Slope CRO Share 1.0000

Assets 0.0007 1.0000

HH Size 0.0195 0.0875 1.0000

Distance -0.0150 -0.0096 -0.0419 1.0000

Quality 0.0071 -0.0126 -0.0749 0.2340 1.0000

Slope 0.0230 -0.0174 -0.0846 0.2783 0.7896 1.0000

Table A3: Total Area and Land Use

Total Area Maize Area Maize Yield per Acre (log) (conditional on Maize Area)

(1) (2) (3)

CRO Share -0.318 0.0370 0.621∗∗

(0.636) (0.413) (0.253)

Household Size 1.007∗∗∗ 0.343∗∗∗ -0.00325

(0.133) (0.0583) (0.0160)

Assets 0.000406 0.000524 0.000151

(0.000457) (0.000359) (0.000110)

Quality -0.0564∗∗ -0.000198 -0.00637

(0.0233) (0.0138) (0.0146)

Slope -0.0290 -0.00795 0.0130

(0.0186) (0.00827) (0.0135)

Distance -0.0164 -0.0197 -0.00125

(0.0145) (0.0129) (0.00369)

Constant 0.719 0.559 1.723

(0.819) (0.361) (0.102)

Household FE Yes Yes Yes

Time FE Yes Yes Yes

Adjusted R2 0.06 0.02 0.00

Observations 9,300 9,300 9,300

* p < 0.10, ** p < 0.05, *** p < 0.01; robust standard errors in parentheses. Note: Table shows two-way fixed effects estimates based on Equation 1. Total Area is the total size of a household’s landownings. Maize Area is the area of land devoted to maize cultivation.

Maize Yield per Acre conditional on Maize Area is maize yield per acre divided by land used for maize production in log terms.

Table A4: Alternative Outcomes

Cash Crop Trees

per Acre Renting Out Joint Title

(1) (2) (3)

CRO Share 126.0 0.0342∗∗ 0.0941∗∗∗

(71.3678) (0.0162) (0.0292)

Household Size -1.664 -0.0004 0.0051∗∗

(2.936) (0.0008) (0.0024)

Assets 2.027 -0.0009 0.0003

(1.391) (0.0006) (0.0010)

Quality 3.612 -0.0010 0.0031∗∗

(9.299) (0.0011) (0.0013)

Slope 8.924 0.0005 0.0004

(9.7579) (0.0007) (0.0018)

Distance 5.607 0.0001 -0.0009

(4.6633) (0.0001) (0.0005)

Constant 190.8 0.0316 0.2377

(24.26) (0.0049) (0.0148)

Household FE Yes Yes Yes

Time FE Yes Yes Yes

Adjusted R2 0.01 0.00 0.00

Observations 9,300 9,300 9,300

* p < 0.10, ** p < 0.05, *** p < 0.01; robust standard errors in parentheses.

Note: Table shows two-way fixed effects estimates based on Equation 1. Cash Crop Trees per Acre is the stock variable of planted permanent cash crop trees relative to the total area owned. Renting Out is the probability of a household leasing out at least one of owned plots (1 = yes, 0 = no). Joint Title is the probability that both wife and spouse are registered as plot owners on at least one of owned plots (1 = yes, 0 = no).

Table A5: Investment Components

Fertilizer per Acre (log)

Pesticides per Acre (log)

Seeds per Acre (log)

(1) (2) (3)

CRO Share 0.3359 0.5894∗∗∗ 0.2427

(0.1779) (0.1870) (0.2529)

Household Size -0.0272∗∗ 0.0203 0.0514∗∗∗

(0.0121) (0.0161) (0.0198)

Assets 0.0146 0.0117 0.0191∗∗

(0.0108) (0.0087) (0.0090)

Quality -0.0053 0.0012 0.0344

(0.0121) (0.0062) (0.0182)

Slope 0.0014 -0.0087 0.0163

(0.0085) (0.0063) (0.0188)

Distance 0.0033 -0.0015 0.0153∗∗∗

(0.0046) (0.0028) (0.0059)

Constant 1.4292 1.0226 3.2930

(0.0762) (0.1001) (0.1248)

Household FE Yes Yes Yes

Time FE Yes Yes Yes

Adjusted R2 0.00 0.00 0.01

Observations 9,300 9,300 9,300

* p < 0.10, ** p < 0.05, *** p < 0.01; robust standard errors in parentheses. Note:

Table shows two-way fixed effects estimates based on Equation 1. Fertilizer per Acre is the value of purchased fertilizer relative to total area owned in log terms (in TSZ). Analogous definitions apply for the value of purchased pesticides or herbicides and the value of purchased seeds.

Figure A1: Summary Statistics - Credit Usage

Figure A2: Transition Probabilities of Household CRO Share

Table A6: Main Results & Channels without Households that acquired Land Maize Yield

per Acre (log)

Tree Seedlings per Acre

Investment

per Acre (log) Security Credit Value per Acre (log)

(1) (2) (3) (4) (5) (6)

CRO Share 0.5147∗∗ 61.0274 0.7538∗∗∗ 0.0571∗∗∗ 0.0034 0.4386∗∗∗

(0.2612) (32.0364) (0.2862) (0.0168) (0.0196) (0.1079)

Household Size -0.0337 0.6252 0.0336 0.0101∗∗∗ 0.0007 -0.0302∗∗∗

(0.0217) (2.5803) (0.0223) (0.0019) (0.0017) (0.0082)

Assets 0.0003 -0.6315 0.0210 0.0005 0.0005 0.0057

(0.0119) (1.6640) (0.0111) (0.0004) (0.0004) (0.0032)

Quality -0.0013 4.7643 0.0365 0.0008 -0.0013 0.0445∗∗∗

(0.0253) (10.8366) (0.0288) (0.0015) (0.0008) (0.0079)

Slope 0.0105 7.3769 0.0277 0.0004 -0.0002 0.0253∗∗∗

(0.0129) (10.4536) (0.0199) (0.0011) (0.0006) (0.0048)

Distance -0.0020 3.5058 0.0110 -0.0032∗∗∗ 0.0002 -0.0005

(0.0062) (3.5939) (0.0062) (0.0007) (0.0004) (0.0029)

Constant 2.1651 85.1410 4.3999 0.8075 0.0743 12.8087

(0.1402) (20.5819) (0.1414) (0.0122) (0.0105) (0.0525)

Household FE Yes Yes Yes Yes Yes Yes

Time FE Yes Yes Yes Yes Yes Yes

Adjusted R2 0.00 0.02 0.01 0.02 0.00 0.03

Observations 8,295 8,295 8,295 8,295 8,295 8,295

* p < 0.10, ** p < 0.05, *** p < 0.01; robust standard errors in parentheses. Note: Table shows two-way fixed effects estimates based on Equation 1. The sample is restricted to households that did not acquire land.

Table A7: Main Results & Channels without Zanzibar Maize Yield

per Acre (log)

Tree Seedlings per Acre

Investment

per Acre (log) Security Credit Value per Acre (log)

(1) (2) (3) (4) (5) (6)

CRO Share 0.7540∗∗∗ 50.4404 0.9351∗∗∗ 0.0593∗∗∗ 0.0352 0.4195∗∗∗

(0.2680) (27.6071) (0.2675) (0.0143) (0.0223) (0.0872) Household Size -0.0147 -2.1528 0.0279 0.0081∗∗∗ 0.0001 -0.0307∗∗∗

(0.0191) (2.1660) (0.0212) (0.0016) (0.0015) (0.0062)

Assets 0.0028 -0.8663 0.0243 0.0006 0.0005 0.0067

(0.0131) (1.6252) (0.0126) (0.0005) (0.0004) (0.0038)

Quality -0.0055 7.7531 0.0268 0.0013 -0.0008 0.0322

(0.0189) (6.9235) (0.0319) (0.0019) (0.0011) (0.0166)

Slope 0.0139 -3.0749 0.0425 0.0002 -0.0003 0.0284∗∗

(0.0234) (6.3726) (0.0368) (0.0022) (0.0013) (0.0120)

Distance -0.0016 3.8804 0.0131∗∗ -0.0030∗∗∗ -0.0000 0.0004

(0.0050) (2.9525) (0.0063) (0.0006) (0.0004) (0.0022)

Constant 2.2943 88.5595 4.7695 0.8232 0.0877 12.6795

(0.1210) (16.2486) (0.1321) (0.0101) (0.0091) (0.0395)

Household FE Yes Yes Yes Yes Yes Yes

Time FE Yes Yes Yes Yes Yes Yes

Adjusted R2 0.00 0.01 0.01 0.02 0.00 0.04

Observations 8,386 8,386 8,386 8,386 8,386 8,386

* p < 0.10, ** p < 0.05, *** p < 0.01; robust standard errors in parentheses. Note: Table shows two-way fixed effects estimates based on Equation 1. The sample is restricted to households living in mainland Tanzania, omitting observations from the island of Zanzibar.

Figure A3: Group-Time Average Treatment Effects - Calendar Time

A: ATT per Year - Maize Yield per Acre (log) B: ATT per Year - Tree Seedlings per Acre

−0.5 0.0 0.5

2010 2012 2014 2020

post 1 Average Effect by Time Period

−100 0 100 200

2010 2012 2014 2020

post 1 Average Effect by Time Period

C: ATT per Year - Investments per Acre (log) D: ATT per Year - Security

−1 0 1

2010 2012 2014 2020

post 1 Average Effect by Time Period

−0.1 0.0 0.1 0.2

2010 2012 2014 2020

post 1 Average Effect by Time Period

E: ATT per Year - Credit F: ATT per Year - Value per Acre (log)

−0.15

−0.10

−0.05 0.00 0.05 0.10

2010 2012 2014 2020

post 1 Average Effect by Time Period

0.0 0.5 1.0 1.5

2010 2012 2014 2020

post 1 Average Effect by Time Period

Note: Figure plots how group-time average treatment effects on the treated vary across treatment groups. Dots represent means, lines simultaneous confidence bands.

References

Acemoglu, D., & Johnson, S. (2005). Unbundling institutions. Journal of Political Economy, 113(5), 949–995. doi.org/10.1086/432166

Acemoglu, D., Johnson, S., & Robinson, J. (2004). Institutions as the Fundamental Cause of Long-Run Growth. NBER Working Paper Series, 2004.

Retrieved from onlinelibrary.wiley.com/doi/10.1002/cbdv.200490137/abstract Aikaeli, J., & Markussen, T. (2017). The effects of land titling in Tanzania.

Ali, D. A., Collin, M., Deininger, K., Dercon, S., Sandefur, J., & Zeitlin, A. (2016). Small price incentives increase women’s access to land titles in Tanzania. Journal of Development Economics, 123, 107–122. doi.org/10.1016/j.jdeveco.2016.06.001

Ali, D. A., Collin, M., Deininger, K., Dercon, S., Sandefur, J., & Zeitlin, A. (2014). The Price of Empowerment: Experimental Evidence on Land Titling in Tanzania. SSRN Electronic Journal, (June). doi.org/10.2139/ssrn.2458152

Angrist, J. D. (2014). The perils of peer effects. Labour Economics, 30, 98–108.

doi.org/10.1016/j.labeco.2014.05.008

Athey, S., & Imbens, G. W. (2021). Design-based analysis in Difference-In-Differences settings with staggered adoption. Journal of Econometrics, (xxxx). doi.org/10.1016/j.jeconom.2020.10.012 Atwood, D. A. (1990). Land registration in Africa: The impact on agricultural production. World Development, 18(5), 659–671. doi.org/10.1016/0305-750X(90)90016-Q

Baker, A., Larcker, D. F., & Wang, C. C. Y. (2021). How Much Should We Trust Staggered Difference-In-Differences Estimates? SSRN Electronic Journal, (March). doi.org/10.2139/ssrn.3794018

Bandiera, O. (2007). Land tenure, investment incentives, and the choice of techniques: Evidence from Nicaragua. World Bank Economic Review, 21(3), 487–508. doi.org/10.1093/wber/lhm005 Banerjee, A. V., Gertler, P. J., & Ghatak, M. (2002). Empowerment and efficiency: Tenancy re-form in West Bengal. Journal of Political Economy, 110(2), 239–280. doi.org/10.1086/338744 Bellemare, M. F. (2013). The productivity impacts of formal and informal land rights: Evidence from Madagascar. Land Economics, 89(2), 272–290. doi.org/10.3368/le.89.2.272

Bertrand, M., Duflo, E., & Mullainathan, S. (2004). How much should we trust Difference-in-difference. (February), 249–275.

Besley, T. (1995). Property Rights and Investment Incentives : Theory and Evidence from Ghana.

Journal of Political Economy, 103(5), 903–937.

Biddulph, R. (2018). The 1999 Tanzania land acts as a community lands approach: A review of research into their implementation. Land Use Policy, 79(August), 48–56.

doi.org/10.1016/j.landusepol.2018.07.036

Binswanger, H. P., Deininger, K., & Feder, G. (1995). Power, Distortions, Revolt and Reform in Agricultural Land Relations. In Handbook of Development Economics: Vol. III.

Boone, C. (2019). Legal Empowerment of the Poor through Property Rights Reform: Tensions and Trade-offs of Land Registration and Titling in Sub-Saharan Africa. Journal of Development Studies, 55(3), 384–400. doi.org/10.1080/00220388.2018.1451633

Brasselle, A. S., Gaspart, F., & Platteau, J. P. (2002). Land tenure security and investment in-centives: Puzzling evidence from Burkina Faso. Journal of Development Economics, 67(2), 373–418.

doi.org/10.1016/S0304-3878(01)00190-0

Callaway, B., Sant’Anna, P. H. C. (2020). Difference-in-Differences with multiple time periods.

GERELATEERDE DOCUMENTEN