• Non ci sono risultati.

Data description

Nel documento UNIVERSITA’ DEGLI STUDI DI PARMA (pagine 62-66)

Abstract

4. Data description

60

𝐸(𝑦𝑖𝑡1(0)|𝑦𝑖𝑡3 = 1) = 𝑥𝑖𝑡1𝛽0+ 𝑧̅𝑖𝜌0+ 𝜉0ℎ̂𝑖𝑡3 (12)

Following Heckman et al. (2001), Di Falco and Veronesi (2013), Abdulai and Huffman (2014) and Imbens and Wooldridge (2009), we may compute the average treatment effect on treated firms (ATET).

In other words, we may assess the impact of WCST adoption decision on land productivity for those farms that receive the treatment as the difference between the expected outcomes in both regimes for the treated farmers. Combining equations (11) and (12), we obtain:

𝐴𝑇𝐸𝑇 = 𝐸 (𝑦𝑖𝑡1(1)|𝑦𝑖𝑡3= 1) − 𝐸 (𝑦𝑖𝑡1(0)|𝑦𝑖𝑡3= 1) = 𝑥𝑖𝑡1(𝛽1− 𝛽0) + 𝑧𝑖(𝜌1− 𝜌0) + ℎ̂𝑖𝑡3(𝜉1− 𝜉0) (13)

which represents the effect of an innovating behaviour induced by climatic variability and other observable characteristics on agricultural yield per hectare that actually choose to innovate. It is worth to note that if selection is based on comparative advantage, then innovating strategy may give higher benefits in terms of land productivity (Abdulai and Huffman, 2014).

61

terms represents 162.82%. This suggests that adopting WCSTs for irrigation may play a significant role in increasing land productivity for the Italian farmers.

As regards the independent variables of the model, we distinguish among production inputs, further inputs, farms’ characteristics, farmers’ characteristics, other incomes and financial and accounting characteristics. We also add the macro-areas in which Italy is sub-divided.

The production inputs that affect the most the agricultural production activity are the total level of working hours spent within the farm (Working hours), the total machine power available in the farm (Machine power) and the market value of the farms’ land (Land value) all expressed in logarithmic term.

As underlined by Timmins (2006), land value should be considered as an endogenous variable8. Thus, we develop two different models. In Model A we overcome the endogeneity of land value by introducing a valid set of instrument variables. Following Timmins (2006), we consider three non-climate attributes, as instruments for the land value variable. Specifically, we introduce the average altitude of the farm fields, as a proxy for land location, the mixed soil texture for capture soil quality, and the external water source as a measure of access to irrigation water availability from consortium, river and natural and artificial lakes. In Model B, instead, we assume that land value is exogenous, and thus, we estimate land productivity considering land value as an input in the outcome equation.

We also consider further exogenous inputs, as the annual costs of energy, electricity and water and the amount spent on insurance to cover from production risks. As control variables, we introduce farms’

characteristics such as farm specialization in producing crops exclusively of high value and family-run management of a farm. The High value crop dummy variable takes the value of 1 if a farm cultivates olives, fruits, vegetables and grapes and 0 otherwise. This allows us to consider the technical-economic orientation of a farm while a family-run management suggests a small farm size. We comprise famers’

characteristics such as the age of the head of the farm and other two dummies indicating if the farm’s head is female, or the farmer holds at least a secondary school education. Further, we control for farmers’

other incomes by considering EU funds and no-EU funds as well as external activities and for financial characteristics such as return on investments (ROI) and leverage (indicating the dimension of external financial resources over the resource generated internally). All the monetary variables are deflated and converted to 2000 euros before logged. Moreover, macro regional and year dummies are included in order to consider the geographic heterogeneity and some exogenous macroeconomic shocks.

WCST adopters which show higher levels of land productivity, present a higher figure of working hours suggesting more working-intense activities and less capital and land value compared to the non-adopting farmers. Moreover, they bear an higher insurance cost as well as higher energy, electricity and water costs. In addition, adopting farms are specialized in high value crops and are not family-run. Farmers who choose to adopt WCSTs are younger, male, more educated compared to non-WCST adopters. The adopters have fewer funds both from EU and national level and are fonder of their own activities without searching for incomes coming from other activities.

8 Timmins (2006) refers to endogeneity of land value within a Ricardian model, but it can be extended to other models as ours in which land value is considered as an explanatory variable. The author argues that land value can be influenced by many unobservable determinants and only in part by climate conditions. Moreover, unobserved determinants of land value may differ with land use and its range of available alternatives, in which the wider alternative land uses are the more severe the estimation bias is.

62

This first mean comparison underlines how the differences between the adopter and the non-adopter in terms of land productivity is relevant but is not enough to explain the adoption of WCST decision across the sample famers. Since the process of WCST adoption could depend on farmers unobserved heterogeneity we should account for the self-selection issue based on the ease of adopting WCSTs by farmers who find these technologies more useful than those who do not adopt by applying a panel ESRM.

63 Table 1. Descriptive statistics

Variable

Adopters (n=8227)

Non-Adopters

(n=35849) Diff.

mean std mean std

Outcome variable Land productivity (€/ha) 20,933.65 81,326.14 7,965.13 62,657.28 12,968.5***

Instruments for Land value when is

considered as an endogenous variable

External water source (d) 5.45 16.35 7.22 28.85 -1.766***

Mixed soil texture (%) 44.73 32.32 58.34 59.84 -13.61***

Altitude avg. (m) 174.05 194.98 302.24 286.11 -128.2***

Production inputs

Working hours (h) 5,744.46 9,650.80 4,021.25 4,218.35 1,723.2***

Machine power (Kw) 157.25 168.87 188.63 202.12 -31.38***

Land value (€) 270,638.20 550,722.60 292,077.80 772,073.40 -21,439.6*

Further inputs

Energy, electricity and

water costs (€) 4,919.01 20,417.87 3,632.73 11,416.46 1,286.3***

Insurance (€) 2,594.30 10,675.00 1,445.71 5,105.74 1,148.6***

Farms’ characteristic High value crop (d) 0.78 0.41 0.32 0.47 0.464***

Family run (d) 0.74 0.44 0.89 0.32 -0.146***

Farmers’

characteristics

Age (years) 53.51 13.24 55.10 13.68 -1.586***

Female head (d) 0.20 0.40 0.21 0.41 -0.0126*

High education (d) 0.34 0.47 0.29 0.45 0.0454***

Other incomes

EU Funds (€) 8,244.64 22,764.86 13,866.22 41,361.35 -5,621.6***

No EU Funds (€) 5,173.33 11,456.52 6,204.37 9,797.57 -1,031.0***

External activities (d) 0.24 0.43 0.25 0.44 -0.0112*

Financial and accounting characteristics

ROI (no) 346.20 3,356.09 186.71 1,932.01 159.5***

Leverage (no) 1.35 5.37 1.26 12.19 0.0882

Macro-areas

North-west (d) 0.13 0.34 0.25 0.43 -0.114***

North-east (d) 0.23 0.42 0.22 0.42 0.01*

Centre (d) 0.16 0.37 0.24 0.43 -0.077***

South (d) 0.33 0.47 0.20 0.40 0.130***

Islands (d) 0.14 0.35 0.09 0.29 0.05***

Note: * p<0.05, ** p<0.01, *** p<0.001;in the units of measurement (d) stays for dummy variable and (no) stays for unit less variable (e.g. index).

64 Table 2: Climatic variability descriptive statistics

Variable Description

Micro-irrigation=1

(n=8228)

Micro-irrigation=0

(n=35855)

Diff.

mean std mean std Climate

variables (instruments for the selection indicator:

Micro-irrigation)

AIJFM Winter Aridity

Index 1.041 0.325 1.156 0.281 0.114***

AIAMJ Spring Aridity

Index 0.442 0.319 0.491 0.252 0.0492***

AIJAS Summer Aridity

Index 0.346 0.365 0.368 0.299 0.0223***

AIOND Autumn Aridity

Index 1.376 0.708 1.556 0.596 0.180***

Note: The Aridity Index is the ratio between P and ET0 and it is calculated considering the moving average of the last 5 years in mm*day-1. If 𝐴𝐼 ≥ 0.65 indicates humid areas, 𝐴𝐼 < 0.65 indicates arid areas. * p<0.05, ** p<0.01, *** p<0.001

In the selection equation as dependent variable a variable indicating the adoption of WCST technologies in each year has been used. This dummy variable assume the value of 1 if a farmer irrigates using drip, micro-sprinklers and sub-irrigation system and 0 otherwise. In the CRE probit model, as climate variables, seasonal AIs are introduced. These represent the exclusion restrictions of the selection indicator to account for the endogeneity of the famer’s choice of adopting WCSTs (Murtazashvili and Wooldridge, 2016). In Table 2, we present the descriptive statistics of climate variability distinguishing for adopters and non-adopters, while in Table A2 of Appendix A, we report the descriptive statistics for all the sample. In spring and summer period, the mean values show that Italy suffers for dryness since the period should be classified as semi-arid, while in winter and autumn season, the AIs measure a degree of humidity in line with the climatic zone. This difference is confirmed even when we distinguish between adopters and non-adopters. In the selection model, all the exogenous explanatory variables used in the outcome equation are also added to the exclusion restrictions as described in Murtazashvili and Wooldridge (2016).

Nel documento UNIVERSITA’ DEGLI STUDI DI PARMA (pagine 62-66)

Documenti correlati