C.3 Exploratory Data Analysis
5.4 ROC curves for out-of-sample predictions
(a) Informa on communica on
technology
(b) Crop diversifica on (c)Coopera ves (d)Irriga on
(e) Livestock introduc on
(f) Self-help groups (g)Stone pitched contour bund
(h)Moisture conserva on pit
(i) Farming system shi (j) Crop loans
1
Note: The graphs have been computed using the R package pROC by Robin et al. (2011).
find the opposite relationship for the use of self-help groups and stone-pitched contour bunds as adaptation strategies. Here, male-headed households are almost 66% (78%) less likely to use self-help groups (stone pitched contour bunds) than female-headed households. Dolisca et al.
(2006) find a similar result – that women’s participation in social activities and their attendance of meetings are more likely. One might argue that the gender effect we measure here is at least partially superimposed by a wealth effect. Male-headed households are likely to be wealthier, and the odds ratios we observe are picking up this relationship. However, we explicitly control for various aspects of wealth, which – apart from other omitted variables – should at least partially reduce this wealth effect.
TABLE5.3: Regression results – Odds ratios
Information and Crop Irrigation Cooperatives Livestock Self-help Stone-pitched Moisture
communication diversification introduction group contour conservation
technology bunds pits
Age 0.902 0.975 0.934 0.771∗∗ 0.905 0.997 1.087 1.276
Age squared 1.001 1.001 1.001 1.002∗∗ 1.001 1.000 0.999 0.998
Education1 0.909 1.942 0.686 3.025∗∗ 0.692 1.281 1.104 0.978
Education2 3.604 4.131 1.774 9.771∗∗∗ 0.212 0.399 0.000 7.761
Gender 7.155∗∗∗ 1.651 1.497 1.665 0.734 0.339∗ 0.220∗ 0.541
Household size 0.855 1.038 0.910 1.222 1.443∗ 1.349∗ 1.215 1.508
Farm Income 1.989 2.972 1.587 1.978 1.143 0.942 0.840 0.921
Household assets 1.751∗∗∗ 1.848∗∗∗ 1.520∗∗ 1.061 0.542∗∗ 0.797 1.928∗∗ 1.185
Livestock 2.109 1.259 1.273 3.585∗∗∗ 85.446∗∗∗ 2.009 0.719 0.700
Poverty status 2.709∗ 1.318 1.827∗ 0.570 0.642 0.338∗∗ 1.986 8.081∗∗
Farm size 0.822 1.154 1.564 3.188∗∗ 2.055 0.851 0.874 0.905
Well ownership 0.178 1.340 3.360 0.223 0.594 0.951 0.424 1.142
Extension service 2.400 1.512 10.323∗∗ 1.984 0.509 6.594∗∗ 2.112 2.796
Rainy days 5.255∗∗∗ 0.463 0.673 1.778 0.189∗∗ 1.872 0.342 0.221
Soil erosion 0.300 1.700 2.303∗ 0.933 1.570 3.172∗∗ 1.914 1.513
Temp rise 0.130∗∗ 0.827 1.459 0.684 0.763 2.284 0.912 1.163
Water depletion 1.236 1.987 2.655 0.599 0.594 1.615 3.526 7.236
Akkiyampadam 0.491 4.120∗∗ 2.731 17.481∗∗∗ 0.039∗∗∗ 16.144∗∗∗ 0.314 0.030∗∗
Eswaramangalam 0.031∗∗∗ 8.295∗∗∗ 3.918∗∗ 3.103 0.052∗∗∗ 1.459 0.011∗∗∗ 0.000
Constant 10.005 0.002∗ 0.058 26.470 138.065 0.091 0.000 0.000
Note: *p<0.1; **p<0.05; ***p<0.01, significance is indicated by bold values.
5.3.2.5 Household size
An odds ratio greater than one for the household size indicates a positive relationship between the household size and the probability of adaptation to climate change. We find that increasing the household size by one person leads to a 44% increase in the odds of using livestock intro-duction and a 35% increase in the use of self-help groups. This is in line with our expectations that the household size (and thus the labor endowment) is positively related to the adaptation process.
5.3.2.6 Various aspects of wealth
We use income, livestock ownership, household assets and poverty status to measure various aspects of wealth. Although we find no significant relationship between the farm’s income as the only source of a household’s income, the latter three measures of wealth seem to be of great importance for the use of adaptation strategies. We find a significantly positive relationship be-tween cooperatives and livestock ownership. Thus, a household is more than 250% more likely to engage in cooperatives if it owns livestock. Furthermore, the odds of introducing further livestock as an adaptation strategy are 85 times greater for a household that already owns live-stock. This might be due to the fact that these households already have prior experience with the challenges of animal husbandry.
Moreover, we find a mostly positive relationship between household assets and various adaptation strategies. Thus, the odds of using information and communication technology, crop diversification, irrigation or stone-pitched contour bunds as adaptation strategies increase significantly with household assets. This meets our expectations. However, we also find a neg-ative relationship for the use of livestock introduction. Here, households are 46% less likely, with respect to odds, to introduce livestock to cope with climate change.
In addition, we find that households that are above the poverty status are 170% more likely to use information and communication technology as a strategy, have seven times greater odds
of using moisture- conservation pits and are more than 80% more likely to use irrigation com-pared with households that are considered to be poor. The opposite is true for involvement in self-help groups. Here, households above the poverty level are more than 66% less likely to make use of this adaptation strategy to face climate change.
5.3.2.7 Farm size in hectares
The farm size of a household might be seen as a proxy for wealth, which makes it easier for a family to adapt to climate change, as adaptation processes typically are often associated with high transaction costs. We find a significantly positive relationship between farm size and en-gagement in cooperatives. Thus, increasing the farm size by 1ha increases the odds of using cooperatives by a factor of three.
5.3.2.8 Well ownership
Not owning a well indicates an inadequate supply of ground water and is considered to be positively associated with the use of adaptation strategies. Nevertheless, we do not find a sig-nificant relationship between owning a well and the use of such strategies.
5.3.2.9 Extension services
The use of extension services is hypothesized to have a positive relationship with the use of adaptation strategies (Jiri et al. 2017). We find a positive relationship between extension services and the use of irrigation and self-help groups as adaptation strategies. A household’s odds of using irrigation are almost 10.5 times higher than those for households that are not provided with extension services. In addition, the odds of using self-help groups are 6.5 times higher when the household is provided with extension services. Thus, we can conclude that extension services that provide households with assistance and information about climate change are very important for the use of adaptation strategies.
5.3.2.10 Climate-change awareness
We expect the sign between climate-change awareness and various adaptation strategies to be inconclusive. However, we find a significantly positive relationship between rainy days and Information communication technology. Furthermore, the introduction of livestock is 81% less likely, with respect to odds, if the household perceives an increase in rainy days. Although not significant, the odds are smaller 1 among crop diversification, irrigation, stone pitched con-tour bunds, moisture-conservation pits and a perceived increase in rainy days, and this is not surprising, as one could expect these strategies to be used for decreased precipitation.
We find a significant relationship between soil erosion and the use of irrigation and self-help groups. Thus, a household is 150% more likely to use irrigation and more than 200% more likely to engage in self-help groups if households are aware of soil erosion. In addition, we find that households that perceive an increase in temperature are 87% less likely to use information communication technology as a strategy, although we do not find any significant effect for the other adaptation strategies.
Being aware of water depletion has no significant effect on any of the adaptation strate-gies. However, the signs for crop diversification, irrigation, stone-pitched contour bunds and moisture conservation especially are in alignment with our expectations, as these strategies are particularly useful for coping with a water shortage.
E Appendix to Chapter 5
TABLE5.4: Regression results – Odds ratios (full table)
Information Crop Irrigation Cooperatives Livestock Self-help Stone pitched Moisture
communication diversification introduction groups contour bund conservation
technology pits
Age 0.902 0.975 0.934 0.771** 0.905 0.997 1.087 1.276
z = -0.743 z = -0.235 z= -0.692 z = -2.267 z = -0.606 z = -0.026 z = 0.390 z = 0.831
Age squared 1.001 1.001 1.001 1.002** 1.001 1.000 0.999 0.998
z = 0.702 z = 0.546 z= 0.585 z = 2.123 z = 0.459 z = -0.347 z = -0.331 z = -0.838
Education1 0.909 1.942 0.686 3.025** 0.692 1.281 1.104 0.978
z = -0.167 z = 1.355 z= -0.311 z = 2.134 z = -0.575 z = 0.410 z = 0.136 z = -0.025
Education2 3.604 4.131 1.774 9.771*** 0.212 0.399 0.00000 7.761
z = 1.147 z = 1.392 z= 0.672 z = 2.748 z = -1.026 z = -0.806 z = -0.013 z = 1.222
Gender 7.155*** 1.651 1.497 1.665 0.734 0.339* 0.220* 0.541
z = 3.438 z = 0.998 z= 0.903 z = 1.023 z = -0.451 z = -1.725 z = -1.856 z = -0.558
Household size 0.855 1.038 0.910 1.222 1.443* 1.349* 1.215 1.508
z = -0.863 z = 0.264 z= -0.742 z = 1.352 z = 1.845 z = 1.823 z = 1.006 z = 1.558
Farm Income 1.989 2.972 1.587 1.978 1.143 0.942 0.840 0.921
z = 0.665 z = 1.206 z= 0.646 z = 0.956 z = 0.154 z = -0.054 z = -0.202 z = -0.089
Household assets 1.751*** 1.848*** 1.520** 1.061 0.542** 0.797 1.928** 1.185
z = 2.777 z = 3.214 z= 2.573 z = 0.368 z = -2.380 z = -1.081 z = 2.077 z = 0.445
Livestock 2.109 1.259 1.273 3.585*** 85.446*** 2.009 0.719 0.700
z = 1.495 z = 0.610 z= 0.698 z = 3.097 z = 5.464 z = 1.558 z = -0.572 z = -0.514
Poverty status 2.709* 1.318 1.827* 0.570 0.642 0.338** 1.986 8.081**
z = 1.941 z = 0.687 z = 1.657 z = -1.316 z = -0.797 z = -2.171 z = 1.024 z = 2.529
Farm size 0.822 1.154 1.564 3.188** 2.055 0.851 0.874 0.905
z = -0.288 z = 0.237 z = 0.862 z = 2.269 z = 1.405 z = -0.347 z = -0.246 z = -0.125
Well ownership 0.178 1.340 3.360 0.223 0.594 0.951 0.424 1.142
z = -1.633 z = 0.373 z = 1.538 z = -1.151 z = -0.423 z = -0.038 z = -1.027 z = 0.152
Extension service 2.400 1.512 10.323** 1.984 0.509 6.594** 2.112 2.796
z = 0.916 z = 0.544 z = 2.134 z = 1.028 z = -0.634 z = 2.423 z = 1.083 z = 1.488
Rainy days 5.255*** 0.463 0.673 1.778 0.189** 1.872 0.342 0.221
z = 2.790 z = -1.565 z = -0.901 z = 1.209 z = -1.990 z = 1.001 z = -1.310 z = -1.490
Soil erosion 0.300 1.700 2.303* 0.933 1.570 3.172** 1.914 1.513
z = -1.643 z = 0.997 z = 1.664 z = -0.124 z = 0.559 z = 2.109 z = 0.618 z = 0.283
Temp rise 0.130** 0.827 1.459 0.684 0.763 2.284 0.912 1.163
z = -2.107 z = -0.323 z = 0.667 z = -0.559 z = -0.343 z = 0.989 z = -0.148 z = 0.221
Water depletion 1.236 1.987 2.655 0.599 0.594 1.615 3.526 7.236
z = 0.215 z = 0.704 z = 1.148 z = -0.638 z = -0.396 z = 0.355 z = 0.935 z = 1.407
Akkiyampadam 0.491 4.120** 2.731 17.481*** 0.039*** 16.144*** 0.314 0.030**
z = -0.752 z = 1.996 z = 1.515 z = 3.735 z = -2.757 z = 3.203 z = -0.981 z = -2.131
Eswaramangalam 0.031*** 8.295*** 3.918** 3.103 0.052*** 1.459 0.011*** 0.000
z = -3.610 z = 2.830 z = 2.075 z = 1.594 z = -2.703 z = 0.448 z = -2.756 z = -0.011
Note: *p<0.1; **p<0.05; ***p<0.01, significance is indicated by bold values.