3.1. Access and Sources of Weather Forecasts
Table 1 shows the distribution of respondents based on their access to forecasts on outbreak of livestock and crop pests/diseases and start of rainfall. Specifically, 62.7% and 48.4% of the farmers from East Africa respectively received information on outbreaks of pests/diseases and start of rainfall as against 29.2% and 56.4% for West Africa. These results emphasize the fact that there have been several initiatives in many African countries to transmit weather forecasts to smallholder farmers [
39].Precisely, Kenya Meteorological Department (KMD) is constitutionally charged with the responsibilities of providing daily, weekly, monthly and seasonal weather forecasts for households in urban and rural areas [
40].
However, desirability of weather forecasts to farmers is directly related to their veracity [
41]. Similarly, the lead time of weather forecasts could vary from time to time and sometimes the shorter the better [
42]. In many instances, forecasting the start of rainfall can be significantly seasonal with some intervals of months as the expected lead time [
43]. Sometimes, weather forecasts often come with associated implications for agricultural production in terms of plant and livestock pathogen development and expected incidence of some diseases.
The results further show access to weather forecasts among the farmers based on their gender. In East Africa, 23.5% and 27.6% of the respondents indicated that both sexes had access to weather forecasts on outbreak of pests/diseases and start of rainfall, respectively. However, male farmers had slightly higher access to weather forecasts than their female counterparts. In addition, in West Africa, women had very low access to weather forecasts on outbreak of pests/diseases and start of rainfall with 1.3% and 1.0%, respectively, as against 16.7% and 32.3% for men.
Inequity in access to weather forecasts across gender had been perceived as a major issue in enhancing adaptive capacity of female farmers, despite the fact that they are already deprived by inadequate access to production resources. Disparity in access to weather forecasts can be directly linked to the medium of forecast transmission. In a study by African Climate Change Resilience Alliance (ACCRA) [
44], it was found that women generally advocated for weather forecast transmission through church, local community groups and markets, while the men preferred radio stations, newspapers and local council meetings. In another study, Cherotich
et al. [
45] found that the majority of Kenyan’s vulnerable women preferred to obtain weather forecasts through radio, while those who were elderly preferred to use indigenous knowledge.
The results further reveal that the majority of the farmers with access to forecasts got them through radio transmissions. This was followed by friends. It is also important to note that involvement of extension officers in weather forecast dissemination was highest for pests/diseases in East Africa. A good number of the farmers indicated that they were monitoring the weather on their own using some form of indigenous knowledge. Radio as a foremost source of accessing weather forecasts by farm households can be strongly linked to its affordability, portability and low maintenance cost. Chavas [
46] noted that radio is among the major channels for reaching farmers with several technology-driven pieces of information. However, it was emphasized that efficiency of this medium in linking forecast information to rural farmers in sub-Saharan Africa is not well known [
47,
48].
Table 1.
Percentage distribution of households’ access to weather forecasts and their sources.
Table 1.
Percentage distribution of households’ access to weather forecasts and their sources.
| East Africa | West Africa |
---|
Pest and Disease | Start of Rainfall | Pest and Disease | Start of Rainfall |
---|
Information received | | | | |
No | 51.7 | 37.3 | 70.8 | 43.6 |
Yes | 48.4 | 62.7 | 29.2 | 56.4 |
Person that received information | | | | |
Men | 12.7 | 19.3 | 16.7 | 32.3 |
Women | 12.0 | 15.3 | 1.3 | 1.0 |
Both | 23.5 | 27.6 | 11.2 | 22.9 |
No response | 51.8 | 37.8 | 70.8 | 43.8 |
Sources of Information | | | | |
Radio | 36.2 | 53.8 | 24.2 | 51.5 |
Extension officers | 13.7 | 2.0 | 0.6 | 2.2 |
NGOs | 1.0 | 1.4 | 0.3 | 4.0 |
Friends | 14.2 | 16.0 | 11.6 | 17.2 |
Traditional forecaster and Indigenous Knowledge | 1.6 | 8.4 | 0.1 | 1.3 |
Own Observation | 6.4 | 13.2 | 10.4 | 5.6 |
Local groups | 6.0 | 3.7 | 4.0 | 1.0 |
Television | 2.3 | 2.7 | 0.9 | 2.9 |
Meteorological offices | 0.1 | 0.3 | 0.00 | 3.0 |
Newspaper | 0.9 | 1.3 | 0.7 | 1.3 |
Religious Organizations | 0.6 | 0.4 | 0.1 | 0.9 |
Cell-phone | 0.0 | 0.0 | 0.0 | 2.0 |
Others | 0.1 | 0.0 | 0.1 | 0.0 |
3.2. Weather Forecasts’ Advice and Associated Changes in Farming Systems
One of the major issues of concern in weather forecast dissemination to farmers is their inability to properly decode disseminated weather information for best farm decision making. This becomes very critical in Africa, where the majority of the farmers are illiterate or possess inadequate formal education required for proper decision making as a result of complexity in climatic scenarios [
49]. Therefore, recent dynamics in weather parameters have mesmerized the majority of African farmers and reckon ineffective their acquired indigenous adaptation knowledge. Therefore, farmers are in critical need of institutional supports from different agencies in order to cope with emerging and precarious weather conditions.
The results in
Table 2 show that weather stations and other stakeholders that are directly involved in disseminating weather information to farmers do accompany them with pieces of advice on the implications that the forecasts are having for farmers’ production decisions. The results further show that 39.3% of the farmers in East Africa indicated that the forecasts that they received on the outbreak of pests/diseases were accompanied by pieces of advice. In addition, 49.4% indicated that forecasts on start of rainfall were accompanied with advice of what the farmers could do. In West Africa, 18.2% and 41.9% of the farmers respectively indicated that weather forecasts on outbreak of pests/diseases and start of rainfalls were accompanied with advice. Due to the high illiteracy level among African farmers, the ability to properly utilize advice received on climatic forecasts would vary from one farmer to another farmer. The results, however, indicated that 33.2% and 44.2% of the farmers from East Africa were able to take some specific farm decisions based on pieces of advice received on outbreak of pests/diseases and start of rainfall, respectively. Among West African farmers, 13.3 and 34.0% of the farmers were able to use weather forecasts received on outbreak of pests/diseases and start of rainfall, respectively.
The need to supplement weather forecasts with such information is to ensure that farmers are properly guided on the implications of weather projections that they have received. The intent is also to bridge the lapses in educational attainments by farmers. However, the adoption perception model of Wossink
et al. [
50] emphasizes that ultimate use of a technology is a function of perception of its value, which is also directly linked several socio-economic characteristics such as education, farming experience, personality and their overall human values of the receivers [
51].
Table 2.
Percentage distribution of farmers that indicated inclusion of advice in forecasts and the ability to use them.
Table 2.
Percentage distribution of farmers that indicated inclusion of advice in forecasts and the ability to use them.
Region | East Africa | West Africa |
---|
Nature of Forecast | Pest and Disease | Start of Rainfall | Pest and Disease | Start of rainfall |
---|
Forecast included advice | | | | |
No | 8.4 | 12.3 | 11.0 | 14.3 |
Yes | 39.3 | 49.4 | 18.2 | 41.9 |
No response | 52.2 | 38.3 | 70.8 | 43.8 |
Able to use advice | | | | |
No | 6.3 | 5.4 | 4.6 | 7.7 |
Yes | 33.2 | 44.2 | 13.3 | 34.0 |
No response | 60.5 | 50.4 | 82.1 | 58.2 |
Information on specific farm decisions that were made by the farmers was also collected during the survey. The results in
Table 3 show that decisions on timing of farming activities were made by 24.0% and 17.6% of the farmers in East and West Africa, respectively, given that they had received weather forecasts on the start of rainfall. In addition, access to forecasts on the start of rainfall induced 20.0% and 12.7% of the farmers from East and West Africa to make land management related decisions. In East Africa, 5.2% and 5.3% of the farmers indicated that given that forecasts on the outbreak of pests/diseases were received, their farm inputs and types of livestock to be reared were altered. These results can be compared in West Africa with 4.3% and 2.4% of the farmers changing their farm inputs and livestock types respectively due to receipt of forecasts on pest/disease outbreak.
Sensitivity of farming operations to changes in weather explains to a very large extent how vulnerable the whole production systems are. The degree of understanding the nature of changes to be effected by farmers in response to access to weather forecasts underscores the degree of households’ adaptability. It should be emphasized that adaptation effectiveness depends on many factors. Among these are degree of accuracy of received forecasts, farm households’ endowment of resources such as land, other assets, family labour and receptiveness to changes. In some recent studies, Roudier
et al. [
38] found that among smallholder farmers from Senegal’s two agro-ecological zones, access to weather forecasts induced about 75.0% of the farmers to significantly change their farming practices.
Table 3.
Changes effected in farming due to forecasts and advice received by farmers.
Table 3.
Changes effected in farming due to forecasts and advice received by farmers.
Changes in Farming due to Forecast | East Africa | West Africa |
---|
Pest and Disease | Start of Rainfall | Pest and Disease | Start of Rainfall |
---|
None | 4.3 | 3.7 | 1.9 | 1.9 |
Land management | 4.4 | 20.0 | 0.9 | 12.7 |
Crop type | 4.6 | 8.7 | 0.9 | 4.6 |
Crop variety | 3.9 | 8.2 | 1.7 | 7.6 |
Use of manure/compost/mulch | 0.0 | 4.4 | 0.6 | 3.2 |
Land area | 0.0 | 0.0 | 0.3 | 0.6 |
Timing of farming activities | 3.7 | 24.0 | 1.3 | 17.6 |
Soil & water conservation | 0.3 | 4.3 | 1.0 | 2.2 |
Irrigation | 0.3 | 0.6 | 0.0 | 0.0 |
Water management | 0.0 | 1.3 | 0.0 | 0.3 |
Tree planting | 0.3 | 3.0 | 0.0 | 1.9 |
Feed management | 4.3 | 2.7 | 0.0 | 0.0 |
Change in inputs | 5.2 | 0.9 | 2.4 | 1.7 |
Field location | 0.0 | 0.0 | 0.0 | 0.0 |
Livestock type | 5.3 | 0.1 | 4.3 | 0.0 |
Livestock breed | 1.3 | 0.0 | 1.7 | 0.0 |
Field location | 0.0 | 1.0 | 0.0 | 1.0 |
Others | 13.5 | 0.4 | 1.9 | 1.4 |
3.3. Probit Regression Results of Factors Influencing Access to Weather Forecasts
Some descriptive statistics of the variables included in the Probit regressions models for East Africa, West Africa and the combined dataset are presented in
Table 4. The table also shows the diagnostic indicators for ensuring that multicollinearity does not seriously exist among the selected variables. However, testing for multicollinearity among Probit regression variables using STATA 13 software (StataCorp LP, TX, USA) is not always directly executed. The available option was to invoke the Variance Inflation Factor (vif) command after subjecting the data to linear probability modeling. The results showed that multicollinearity was not a problem based on a high level of tolerance of the variables. Specifically, VIF for the results for East Africa, West Africa and combined data are 1.17, 1.16 and 1.19, respectively.It had been indicated that multicollinearity should be worried about if VIF is considerably higher than 1.0 [
52].
However, recent econometric analyses emphasize a robustness check as a way of ensuring that parameters from regression analysis are not sensitive to removal of some important variables or reduction in the number of observations. In order to test for robustness of the results, the Probit regressions were run separately with East African data, West African dataand a combination of the two datasets. The aim was to test for sensitivity of the results to reduction in the number of observations. It also seeks to determine if there are differences in the parameters across the two African regions.
Table 5 and
Table 6 present the results of Probit regression data analyses for access to forecasts on outbreak of pests and diseases and start of rainfall respectively. Statistical significance of the Likelihood Ratio Chi Square statistics (
p < 0.05) in all of the results implies that the estimated models produced good fits for the data in all the estimated models. The results show that compared to the results across regions, estimated models for the combined data have a higher number of the parameters being statistically significant at 5% level. The implication is that the results must be interpreted with cautions. All the models are therefore considered in the final interpretation of results.
In
Table 5 and
Table 6, region-specific parameters (East Africa) which were included in the first model for the combined dataset are with positive sign and show statistical significance (
p < 0.05). These results imply that compared to those respondents from West Africa and holding other variables constant, farmers from East Africa had significantly higher access to weather forecast on start of rainfall and outbreak of pests/diseases. The issue of access to weather forecasts in African agriculture had been perceived as major keys to enhancing adaptive capacity of farmers [
53]. However, insurmountable obstacles include inadequate access as a result of limitations in transmission equipment, timeliness and usefulness of the information to the largely illiterate smallholding African farmers [
20,
53].
In
Table 5 and
Table 6, the parameters of personal land areas owned by farmers did not show statistical significance in all the models, although they consistently have negative sign (
p > 0.05).
Table 5 shows that in the East African, West African and combined dataset models for access to forecasts on incidence of pests and diseases, the parameters of personal degraded land (ha) did not show statistical significance (
p > 0.05). When the East Africa dummy was removed from the combined model, it shows statistical significance (
p < 0.05) and implies that increasing degraded land areas owned by farmers would decrease the probability of having access to forecasts on incidence of pests and diseases. In
Table 6, results for combined dataset and East Africa show that as the number of degraded land owned by farmers increased, probability of having access to forecasts on commencement of rainfall significantly decreased (
p < 0.05). Some studies have emphasized the complex interrelationships between land ownership, tenure patterns and exposure to unfavourable weather conditions [
54]. The above result emphasizes the fact that farmers with more degraded land areas would have little incentives to seek for information on weather. Similarly, land ownership patterns can influence natural resource conservation, which invariably influences some climatic parameters. In addition, the size of farm land owned and fertility status would influence farmers’ ultimate farm investment decisions and weather information seeking behaviour due to the enormity of expected losses in the event of unexpected weather [
19]. Saka
et al. [
55] and Lawal
et al. [
56] emphasized that soil fertility status would influence the pattern of crop combination although degraded plots of land are often left to fallow and regain fertility for future uses.
In
Table 5, all of the results consistently reveal that farmers with previous exposure to weather-related crises had significantly higher probability of having access to forecasts on incidence of pests and diseases (
p < 0.05).
Table 6, however, shows that, except for the parameter in West Africa’s model, farm households that previously got exposed to weather-related crises had significantly higher probability of having access to forecasts on start of rainfall. Similarly, in
Table 5, except for West African farmers, those who previously received assistance during climatic crisis had significantly higher probability (
p < 0.05) of having access to forecasts on incidence of pests and diseases. In
Table 6, except for East African farmers, those who previously received assistance during climatic crisis had significantly higher probability (
p < 0.05) of having access to forecasts on start of rainfall. Rational farmers with previous exposure to climatic shocks are expected to guard against future occurrences by seeking weather-related information from every available source.
The results in
Table 5 and
Table 6 further show that except in the results for East Africa on access to forecasts on start of rainfall, household size parameters did not show statistical significance in the estimated models. In some previous studies on factors influencing farmers’ decisions on climate change adaptation, statistically insignificant parameters had been reported. For East Africa, access to weather forecasts on start of rainfall decreases significantly as household size increases. Similar result had been reported by Oyekale [
20]. Accessing weather forecast is a form of production risk mitigation for which large households may possess high risk aversion adaptation may be risky subject to their peculiar circumstances [
57,
58,
59,
60,
61,
62].
The results presented in
Table 5 further reveal that none of the parameters of members that are 60 years old or above are statistically significant (
p > 0.05), while those for combined dataset and West Africa show statistical significance (
p < 0.05) in
Table 6. These results imply that as the number of household members that are 60 years or above increases, probabilities of accessing forecasts on the start of rainfall significantly decreases (
p < 0.05).
Table 5 shows that the probability of accessing weather forecasts on the incidence of pests and diseases increased significantly (
p < 0.05) with attainment of formal education across all the estimated models. However, in
Table 6, the parameter of education for West Africa did not show statistical significance (
p > 0.05). The observed relationship between education and access to climate forecasts is expected given the fact that education would enhance farmers’ awareness on existing media channels for sourcing weather forecast information [
20]. Education may also enhance understanding of farmers on the consequences of adverse weather conditions and impact mitigating options. However, the probability of accessing weather forecasts on incidence of pests and diseases decreased significantly (
p < 0.05) as the number of household members that are 60 years or more increased. Obayelu
et al. [
63] reported similar findings where the age of farmers reduced adoption of adaptation methods to unfavourable weather conditions in Nigeria.
The parameters of most of the livelihood sources variables in
Table 5 and
Table 6 are with a positive sign, although many did not show statistical significance (
p > 0.05) in
Table 5. Except in access to the start of the rainfall model for East Africa, parameters of farm employment income are with a positive sign and statistically significant (
p < 0.05). These results imply that farmers with production surpluses that bring some income for their households have higher probabilities of accessing forecasts on incidence of pests and diseases and the start of rainfall. This is expected since farming is among the occupations that are most vulnerable to weather vagaries [
64].
Moreover, parameters of involvement in other paid employment in
Table 5 did not show statistical significance (
p > 0.05), while only that for East African farmers did not show statistical significance in
Table 6. This implies that in the model for combined dataset and West Africa, involvement in other paid employment significantly increased the probability of having access to forecasts on start of rainfall. The parameters of access to business income also have positive sign and statistically significant (
p < 0.05), except for West Africa in
Table 5. These imply that access to business income increases probability accessing weather forecasts on incidence of pests and diseases and start of rainfall. The results go in line with the assertion of Rosenzweig and Udry [
65] that rainfall forecasts enhance labour allocation. Farmers that are engaged in other wage employments and businesses would be cautious in their labour time allocation, due to existence of serious constraints, and would ensure that consequences from weather vagaries are maximally avoided.
Parameters of access to remittances or gifts income in
Table 5 only showed statistical significance (
p < 0.05) in the model with combined dataset. In
Table 6, however, this variable shows statistical significance (
p < 0.05) in all the models, except that for East Africa. Except in the model for East Africa, access to environmental services income and project/government income significantly increased (
p < 0.05) probability of accessing forecasts on incidence of pests and diseases and start of rainfall, respectively. However, except in the model for East Africa, access to project/government incomes significantly reduced access to forecasts on incidence of pests and diseases (
p < 0.05). Similarly, access to the renting out of machineries’ income significantly increased (
p < 0.05) access to forecasts on the start of rainfall (except in East Africa).Remittances and government supports for agricultural farmers are crucial ways of adapting to weather vagaries in agricultural production [
66,
67]. It had been shown that remittances are vital options for reducing rural poverty [
68], which is the utmost development policy goal in many developing countries. The results have emphasized the need for enhanced income and farmers’ livelihood diversification in the quest towards weather vagaries’ impact mitigation. This finding is in line with that of Lyimo and Kangalawe [
69]. Farmers that obtained bank loans had significantly lower probability of having access to forecasts on pests and diseases in the combined model and the start of rainfall in the models for East and West Africa. Access to informal loans significantly increased access to forecasts on incidence of pests and diseases in East Africa but reduced it in West Africa.
Table 5 shows that perception of more erratic rainfall significantly increased the probability (
p < 0.05) of accessing forecasts on incidence of pests and diseases in the combined and West Africa models. Pests and disease causing pathogens often regain more vigour due to changes in weather parameters, thereby producing biotic and abiotic ecosystems where survival is optimized. Some pests/diseases that are associated with certain crops are often more pronounced during the rainy season or when rainfall is exceeding the usual average. Tubby and Webber [
70] noted that changes in some weather parameters influence pests and disease incidences through changes in physiology of the host plants, enhancement of pests’ and pathogens’ developmental processes, inability of the pests’ and pathogens’ predators to survive and favourability of reported changes to non-native pests and pathogens [
71].
Table 4.
Descriptive statistics of selected variables and their multicollinearity diagnostics.
Table 4.
Descriptive statistics of selected variables and their multicollinearity diagnostics.
Variables | East Africa | West Africa | All Farmers |
---|
Mean | Std. Err. | Tolerance | Mean | Std Err. | Tolerance | Mean | Std. Err. | Tolerance |
---|
Personal land (ha) | 2.83 | 0.12 | 74.75 | 5.54 | 0.37 | 88.55 | 4.18 | 0.20 | 83.44 |
Personal degraded land (ha) | 0.25 | 0.03 | 91.66 | 1.08 | 0.07 | 90.24 | 0.67 | 0.04 | 92.27 |
Exposure to weather-related crisis (previous five years) | 0.72 | 0.02 | 86.17 | 0.74 | 0.02 | 84.35 | 0.73 | 0.01 | 86.38 |
Assistance received during weather-related crisis | 0.20 | 0.02 | 94.89 | 0.17 | 0.01 | 94.56 | 0.19 | 0.01 | 96.60 |
Household size | 6.24 | 0.12 | 83.10 | 12.72 | 0.37 | 87.12 | 9.48 | 0.21 | 85.83 |
Household more 60 years | 0.38 | 0.02 | 85.62 | 0.94 | 0.04 | 88.30 | 0.66 | 0.02 | 89.39 |
Gender (male) | 0.37 | 0.02 | 89.19 | 0.42 | 0.02 | 90.63 | 0.39 | 0.01 | 84.00 |
Formal education | 0.18 | 0.01 | 85.14 | 0.17 | 0.01 | 88.05 | 0.17 | 0.01 | 90.31 |
Farm employment income | 0.34 | 0.02 | 85.78 | 0.47 | 0.02 | 95.24 | 0.41 | 0.01 | 93.85 |
Other paid employment income | 0.36 | 0.02 | 92.37 | 0.26 | 0.02 | 92.48 | 0.31 | 0.01 | 93.20 |
Business income | 0.02 | 0.01 | 87.21 | 0.03 | 0.01 | 91.88 | 0.03 | 0.00 | 94.25 |
Remittances or gifts income | 0.07 | 0.01 | 92.15 | 0.15 | 0.01 | 90.83 | 0.11 | 0.01 | 91.37 |
Environmental services income | 0.10 | 0.01 | 80.02 | 0.17 | 0.01 | 85.96 | 0.13 | 0.01 | 87.74 |
Projects/govt income | 0.17 | 0.01 | 86.19 | 0.47 | 0.02 | 86.32 | 0.32 | 0.01 | 87.67 |
Bank loan | 0.04 | 0.01 | 84.15 | 0.11 | 0.01 | 83.57 | 0.07 | 0.01 | 87.84 |
Informal loan | 0.06 | 0.01 | 77.99 | 0.04 | 0.01 | 86.18 | 0.05 | 0.01 | 84.63 |
Renting of machinery income | 0.25 | 0.02 | 82.27 | 0.44 | 0.02 | 93.57 | 0.35 | 0.01 | 82.62 |
Renting out land income | 0.74 | 0.02 | 90.68 | 0.96 | 0.01 | 75.14 | 0.85 | 0.01 | 73.43 |
More erratic rainfall | 0.90 | 0.01 | 81.10 | 0.75 | 0.02 | 68.14 | 0.83 | 0.01 | 62.69 |
Has radio | 0.54 | 0.02 | 74.22 | 0.50 | 0.02 | 82.85 | 0.52 | 0.01 | 83.66 |
Has Television | 0.18 | 0.01 | 94.86 | 0.10 | 0.01 | 88.12 | 0.14 | 0.01 | 86.13 |
Introduced new crop | 0.46 | 0.02 | 88.36 | 0.27 | 0.02 | 75.03 | 0.37 | 0.01 | 78.09 |
Testing new crops | 0.69 | 0.02 | 85.57 | 0.83 | 0.01 | 86.63 | 0.76 | 0.01 | 85.14 |
Stopped growing a crop | 0.06 | 0.01 | 86.36 | 0.06 | 0.01 | 88.36 | 0.06 | 0.01 | 89.84 |
Table 5.
Probit regression results of factors influencing access to forecasts on incidence of pests and diseases.
Table 5.
Probit regression results of factors influencing access to forecasts on incidence of pests and diseases.
Variables | Coefficient. | Z stat | Coefficient. | Z stat | Coefficient. | Z stat | Coefficient. | Z stat |
---|
Combined Data | Combined Data | East Africa | West Africa |
---|
East Africa dummy | 0.3763 | 3.84 * | - | - | - | - | - | - |
Personal land (ha) | −0.0165 | −1.87 | −0.0171 | −1.93 | −0.025 | −1.42 | −0.009 | −0.88 |
Personal degraded land (ha) | −0.0468 | −1.70 | −0.0686 | −2.52 * | −0.087 | −1.40 | −0.023 | −0.77 |
Exposure to weather-related crisis (previous five years) | 0.4701 | 5.15 * | 0.4518 | 4.98 * | 0.458 | 3.42 * | 0.544 | 3.95 * |
Assistance received during weather-related crisis | 0.2456 | 2.43 * | 0.2403 | 2.38 * | 0.557 | 3.54 * | 0.03 | 0.2 |
Household size | −0.0002 | −0.03 | −0.0054 | −0.94 | −0.025 | −1.34 | 0.003 | 0.42 |
Household more 60 years | −0.0230 | −0.48 | −0.0609 | −1.28 | −0.040 | −0.45 | 0.008 | 0.13 |
Gender (male) | 0.081 | 0.71 | 0.0076 | 0.07 | 0.084 | 0.6 | 0.161 | 0.61 |
Formal education | 0.5067 | 4.50 * | 0.6091 | 5.56 * | 0.483 | 2.16 * | 0.378 | 2.69 * |
Farm employment income | 0.2979 | 3.80 * | 0.2806 | 3.60 * | 0.352 | 2.94 * | 0.28 | 2.47 * |
Other paid employment income | 0.121 | 1.22 | 0.1312 | 1.33 | 0.181 | 1.2 | −0.126 | −0.85 |
Business income | 0.1961 | 2.43 * | 0.1917 | 2.40 * | 0.416 | 3.44 * | 0.035 | 0.3 |
Remittances or gifts income | 0.1693 | 2.00 * | 0.1985 | 2.37 * | 0.235 | 1.9 | 0.086 | 0.67 |
Environmental services income | 0.8323 | 3.30 * | 0.8239 | 3.26 * | 0.369 | 0.86 | 1.046 | 3.23 * |
Projects/govt income | −0.3601 | −2.86 * | −0.4047 | −3.26 * | 0.404 | 1.83 | −0.697 | −3.94 * |
Bank loan | −0.2795 | −2.43 * | −0.2941 | −2.57 * | −0.301 | −1.58 | −0.147 | −0.96 |
Informal loan | −0.0385 | −0.44 | −0.1085 | −1.29 | 0.493 | 3.18 * | −0.278 | −2.48 * |
Renting of machinery income | −0.2748 | −1.84 | −0.3272 | −2.22 * | 0.001 | 0 | −0.284 | −1.55 |
Renting out land income | 0.4569 | 2.64 * | 0.5036 | 2.94 * | 0.431 | 1.77 | 0.332 | 1.19 |
More erratic rainfall | 0.2174 | 2.65 * | 0.1561 | 1.95 | 0.052 | 0.38 | 0.244 | 2.17 * |
Has radio | 0.515 | 5.28 * | 0.504 | 5.16 * | 0.853 | 6.19 * | −0.082 | −0.55 |
Has Television | 0.2479 | 1.55 | 0.2779 | 1.74 | 0.306 | 1.26 | 0.203 | 0.89 |
Introduced new crop | 0.1642 | 2.06 * | 0.1748 | 2.20 * | 0.173 | 1.4 | 0.14 | 1.21 |
Testing new crops | 0.103 | 0.95 | 0.1416 | 1.32 | −0.221 | −1.48 | 0.335 | 1.88 |
Stopped growing a crop | 0.3431 | 4.24 * | 0.3969 | 5.00 * | 0.41 | 3.55 * | 0.183 | 1.46 |
Constant | −2.1988 | −1.83 * | −1.9056 | −1.29 * | −2.118 | −8.00 * | −1.528 | −4.78 * |
Log likelihood function | −769.26 | | −776.65 | | −44.448 | | −373.94 | |
LR Chi Square | 328.39 * | | 313.61 * | | 279.37 * | | 96.230 * | |
Pseudo R Square | 0.1759 | | 0.168 | | 0.289 | | 0.114 | |
Number of observations | 1398 | | 1398 | | 699 | | 699 | |
In addition,
Table 6 shows that farmers that perceived more erratic rainfalls had significantly higher probability (
p < 0.05) of having access to forecasts on the start of rainfall in all of the estimated models. This result goes in line with the expectation, given the centrality of rainfall adequacy for production activities of crop farmers. The role of accurate forecasts cannot be downplayed, although the ability of farmers to decode the provided forecasts for optimum decision making is very critical for reducing the impacts of climatic uncertainties on farm production and decision making.
Table 6.
Probit regression results of factors influencing access to forecasts on start of rainfall.
Table 6.
Probit regression results of factors influencing access to forecasts on start of rainfall.
Variables | Coefficient. | Z stat | Coefficient. | Z stat | Coefficient. | Z stat | Coefficient. | Z stat |
---|
Combined Data | Combined Data | East Africa | West Africa |
---|
East Africa | 0.3332 | 3.36 * | - | - | - | - | - | |
Personal land (ha) | 0.0105 | 1.42 | 0.0101 | 1.38 | 0.018 | 0.99 | 0.013 | 1.53 |
Personal degraded land (ha) | −0.0701 | −2.71 * | −0.0870 | −3.42 * | −0.159 | −2.43 * | −0.037 | −1.25 |
Exposure to weather-related crisis in the previous five years | 0.2296 | 2.63 * | 0.2208 | 2.54 * | 0.358 | 2.74* | 0.133 | 1.05 |
Assistance received during weather-related crisis | 0.4631 | 4.21 * | 0.4574 | 4.18 * | 0.241 | 1.46 | 0.588 | 3.74 * |
Household size | −0.0007 | −0.11 | −0.0048 | −0.85 | −0.056 | −2.98 * | 0.005 | 0.73 |
Household more 60 years | −0.1204 | −2.52 * | −0.1517 | −3.25 * | −0.078 | −0.91 | −0.157 | −2.65 * |
Gender (male) | −0.0302 | −0.26 | −0.0949 | −0.84 | −0.121 | −0.87 | 0.462 | 1.75 |
Formal education | 0.1465 | 1.41 | 0.2273 | 2.25 * | 0.413 | 2.12 * | 0.045 | 0.34 |
Farm employment income | 0.3573 | 4.45 * | 0.3442 | 4.31 * | 0.217 | 1.76 | 0.375 | 3.34 * |
Other paid employment income | 0.2527 | 2.43 * | 0.2565 | 2.47 * | 0.225 | 1.40 | 0.309 | 2.11 * |
Business income | 0.3704 | 4.55 * | 0.3624 | 4.47 * | 0.343 | 2.68 * | 0.373 | 3.33 * |
Remittances or gifts income | 0.2222 | 2.52 * | 0.2476 | 2.83 * | 0.125 | 0.98 | 0.270 | 2.08 * |
Environmental services income | 0.1053 | 0.41 | 0.1146 | 0.45 | 0.047 | 0.11 | 0.231 | 0.67 |
Projects/govt income | 0.4556 | 3.53 * | 0.4124 | 3.20 * | 0.152 | 0.67 | 0.654 | 4.04 * |
Bank loan | 0.1005 | 0.86 | 0.0904 | 0.77 | −0.39 | −2.00 * | 0.438 | 2.83 * |
Informal loan | 0.1775 | 2.03 * | 0.1111 | 1.31 | 0.293 | 1.79 | 0.112 | 1.03 |
Renting of machinery income | 0.5532 | 3.14 * | 0.5001 | 2.87 * | 0.832 | 1.91 | 0.523 | 2.60 * |
Renting out land income | 0.1346 | 0.70 | 0.1768 | 0.93 | 0.033 | 0.13 | 0.197 | 0.61 |
More erratic rainfall | 0.2631 | 3.11 * | 0.1969 | 2.41 * | 0.406 | 2.75 * | 0.234 | 2.09 * |
Has radio | 0.6232 | 6.71 * | 0.5969 | 6.45 * | 0.891 | 6.84 * | 0.288 | 1.98 * |
Has television | 0.4731 | 2.49 * | 0.5004 | 2.65 * | 0.571 | 1.97 * | 0.427 | 1.68 |
Introduced new crop | 0.1604 | 2.00 * | 0.1683 | 2.11 * | 0.139 | 1.10 | 0.138 | 1.22 |
Testing new crops | −0.0938 | −0.83 | −0.0574 | −0.52 | −0.201 | −1.34 | 0.117 | 0.64 |
Stopped growing a crop | 0.2418 | 2.88 * | 0.2868 | 3.47 * | 0.377 | 3.15 * | 0.105 | 0.83 |
Constant | −1.4419 | −8.18 * | −1.1692 | −7.52 * | −1.093 | −4.75 * | −1.625 | −5.11 * |
Log likelihood | −758.24 | | −763.92 | | −339.49 | | −392.42 | |
LR Chi Square (25) | 370.65 * | | 359.29 * | | 244.71 | | 172.820 | |
Pseudo R2 | 0.1964 | | 0.1904 | | 0.265 | | 0.181 | |
No of observations | 1398 | | 1398 | | 699 | | 699 | |
The results in
Table 5 and
Table 6 further show that access to the radio significantly increased the probabilities of receiving forecasts on the incidence of pests/diseases (except in the model for West Africa) and the start of rainfall (
p < 0.05). However, access to television also shows statistical significance (
p < 0.05) in the model of the access to forecasts on the start of rainfall in the combined and East African models. It has been noted that enhancing accessibility to climatic forecasts among farmers relies so much on access to radios due to its low cost of maintenance and wide coverage [
45,
46,
47].
Table 5 and
Table 6 also show that farmers that introduced new crops and stopped growing a crop had significantly higher probabilities of having access to forecasts on incidence of pests and diseases and the start of rainfall in the combined and East African models. These results are in line with expectations given that these underscore vulnerability to unfavourable weather situations, which may have necessitated farmers to adapt.