Impact of Climate Change Adaptation on Food Security: Evidence from Semi-Arid Lands, Kenya Impact of climate change Adaptation on food security: Evidence from Semi-Arid 1 Lands, Kenya

28 The paper investigates the impact of climate change adaptation on food security in the Semi- 29 Arid parts of Kenya. Our research used a sample of 440 households, and an endogenous 30 Switching regression is estimated to account for the heterogeneity in the decision to adapt or 31 not, and for unobservable characteristics of pastoralists. We examined how pastoralists’ 32 decision to adapt, that is to implement a set of strategies; storage/purchase of fodder, change in 33 water management, partial shift to other livelihoods, banking livestock assets and herd 34 management. The results demonstrate that climate change adaptation increases food security 35 among pastoralists significantly. Pastoralists who have not adapted are seven percentage points 36 more likely to be food secure if they had adapted to climate change while adopters are 27 37 percentage points likely to be food insecure if they had not adapted. The paper recommends 38 the strengthening of policies on adaptation to climate change in the Semi-Arid lands where 39 pastoralism is the primary means of livelihood. Although pastoralists have information on the 40 effect of climate change on their livestock, incomplete information on the mechanism of 41 adaptation remains a hurdle. Consistent climate change monitoring, timely warning systems 42 and communication of pertinent information to pastoralists is fundamental. 43

Most pastoralist systems have a tradition of communal pasture set aside as a drought 180 reserve and also crucial for pasture rehabilitation objectives. Some also provide for household 181 pasture reserves for feeding lactating and immature stock. While the household reserve system is expanding in some pastoral areas (Coppock 1994), population pressure and the 183 weakening of tribal reciprocity agreements and traditional law in many pastoral communities 184 has eliminated pasture set aside practices. However, fodder conservation does not often 185 extend beyond family initiatives and is unlikely to return to pastoralist's communal resource 186 management systems until governments improve pastoralist's land rights and strengthen 187 capacity for participatory natural resource management in pastoral areas.

188
Supplementary feeding had no place in traditional pastoralism. However, the 189 availability of industrial by-products such as oil-seed cakes and molasses has begun to 190 change this situation, and wealthier owners of more massive herds are gradually taking 191 advantage of the flexibility they offer (Blench & Marriage 1998  as milk rather than in cash from the selling livestock (Bailey et al., 1999). As such, these 205 folks will hold onto their livestock until their salvage value is higher than their income  Destocking or herd size reduction also serves to shed off weaker animals from the herd 221 resulting in keeping stronger animals to preserve capital assets to suit the household and 222 enable recovery after drought and continue with production of milk (major source of food 223 security in pastoral areas).

224
Access to water is critical for efficient tracking of feed resources. In areas with 225 permanent water supply, over-utilization and environmental damage are likely to take place.   Impact assessment using non-experimental data is very challenging due to the problem of self-245 selection and lack of counterfactual against which impact can be evaluated. In experimental 246 studies, these problems are ably addressed by assigning the treatment randomly to the target study population. However, in this current study, adaptation to climate change among the 248 studied population of pastoralists is not randomly assigned, but instead, households self-select 249 themselves into adaptation (treated) and non-adaptation (untreated) regimes. This self-selection 250 into the two treatment groups means that there could be systematic differences between the 251 treated and the untreated groups. Therefore, evaluating the impact of the treatment on food . ESR is a two-step procedure that involves first modelling the 275 household decision to adapt to climate change following the random utility formulation of the 276 non-separable household model approach. In this first step, a household is assumed to adapt to 277 climate change if its utility from adaptation (Ui1) is higher than the utility from non-adaptation 278 (Ui0), i.e. the utility derived from adoption (U*) is greater than zero: -Since this utility is unobservable, then the adoption decision can be represented as a function 281 of observable characteristics ( ) and the error term ( ) in the following latent variable model:- Where * is the unobserved binary variable indicator of climate change adaptation; is the We, therefore, applied the two-stage endogenous switching regression by 303 estimating equation (2) Where:      Descriptive statistics 416 In this study, the following climate change adaptation strategies are applied by pastoralists:

417
Purchase of fodder (Usually hay), water management, herd management and shift to other 418 livelihoods (    We also found that pastoralists who live far away from the markets probably could not adapt 477 some drought management approaches. We also found that the increase in rainfall leads to 478 climate change adaptation. We suspect that these results uncover that we do find increase in Ethiopia's pastoral areas.

483
As expected, wealthier households and those who were more endowed with livestock happen 484 to be more food secure. This is expected given that livestock production is the main livelihood 485 activity in the Semi-Arid economies. Delay in rainy season is likely to reduce food security 486 status. Household food security was also found to be enhanced by access to credit. Also, the 487 study uncovered that a high prevalence of food security among educated households. A    498 The switching regression results were used to estimate the expected conditional 499 probability of food security and to estimate the impact of climate change adaptation. The 500 results show that the probability of food security among adapters is likely to significantly drop 501 from about 82% to about 55% had they not adapted to climate change (Table 6). On the other 502 hand, the probability of being food secure among the non-adapters could significantly increase 503 from about 64% to almost 70% had they adapted to climate change. These results show that 504 climate change adaptation among the sampled pastoralists' households is crucial in ensuring 505 household food security. We also find these findings to be consistent with past studies that have  Table 6 shows the heterogeneity effect of adaptations to 508 climate changes on food security. We found that had the adapters not adapted; their food 509 security probability could have dropped to the level that it could have been significantly lower 510 than that of non-adapters in their current state of having not adapted. However, on the other 511 hand, even if the non-adapters were to adapt, their food security probability would still be 512 significantly lower than that of adapters given their current state of having adopted. These later 513 findings on heterogeneity show that some unobserved characteristics make adapters to have 514 significantly higher food security probability than their non-adapting counterparts.

515
To tease/entangle out some these differences that cause the significant food 516 security gap between adapters and non-adapters, we decompose the observed differences in 517 food security probability following Oaxaca (1973) decomposition procedure. The observed 518 food security probability (column (a) less column (b)) 0.186 can be decomposed into that 519 portion attributed to the differences in resource base and that one that is due to differences in 520 efficiency in the use of resources held between the two groups of households (adapters and 521 non-adapters). We found that if non-adapters were to keep their current resource use efficiency 522 but given the same resources like the ones held currently by adapters, their food security  To bridge this 36-percentage point gap, the efficiency in the use of resources by non-adapters 527 needs to be improved too. Therefore, to close the food security gap that exists between adapters and non-adapters, the resource base and efficiency in use of the resources by the non-adapters 529 needs to be improved. Production of hay is a suitable activity in pasture lands in large farms and ranches. Therefore, 567 the county government of Laikipia is committed to supporting and promoting hay production.

568
As expected the higher the distance from the market, the less likely to adapt to changes in 569 climate. Given that livestock production is the main economic activity in the Semi-Arid 570 economies, we found those households with more livestock to be more food secure. But one of 571 the adaptation strategies is to have optimal herd size, therefore, there is need for efforts to 572 encourage herders to reduce their herd size for more stability in the face of climate change -573 the lower the herd number, potentially the more savings and ability to get out of a drought or 574 long dry spell for example. If drought hits and a household has invested a lot into developing 575 a large herd and put little aside, this can cripple household food security and development.

576
There is also need for pastoralist to combine herd size with the keeping of livestock as a 577 business which enables them to have a plan on how to sell when animal gains the right live 578 weight and consequently reduce the herd size to the optimal level.

579
Access to credit makes a household to be more food secure. These findings are consistent with