Spatial Analysis of Climate Driver Impacts on Sub-Saharan African Migration Patterns in Tanzania

Abstract

Environmental problems resulting from climate change have generated negative impacts on climate-sensitive sectors of African economies. Coping with adverse situations, individuals and households adopt several strategies, including rural-urban migration. Previous literature has investigated the use of migration as a coping strategy to environmental factors. However, specific empirical assessment of links between migration and climatic factors with emphasis on spatial perspectives is not well studied. Accordingly, this study focuses on climatic driver influences on migration from statistical and spatial perspectives using logistic regression and Geographic Information Systems (GIS). We combined secondary data sets collected by the World Bank SLMS nationwide household surveys with geo-referenced sub-villages and historical gridded rainfall and temperature data. Results suggest a significant positive relationship between long-run precipitation and migration, while long-run temperature was statistically inconsequential. Results also suggest spatial patterns and climate change drivers are critical in understanding the migration determinants in Tanzania.

1. Introduction

Environmental problems resulting from climate change have continued to generate negative impacts on climate-sensitive sectors of African economies. During the twentieth century, the continent faced climate change with a rise in temperature of 0.58°C, with some areas warming even faster than others [1]. Some estimates suggest annual mean surface air temperatures in Africa had risen about 1.5 times the predicted average global increase by 2009 [2, 3]. Despite frequent intense precipitation events, African annual average precipitation has decreased, likely bringing extended droughts. In conjunction with increased prevalence of climate change driven extreme precipitation events, some regions will be increasingly susceptible to both droughts and flooding [2, 4]. This alarming pace of climate change [5] has attracted attention from national and international environmental institutions [3, 6]. Adverse effects of climate change on social and ecological systems include drought, decreased agricultural production, flooding, and hurricanes. These effects are likely to be pronounced in developing countries because their economies are overwhelmingly based on rain-fed agriculture [7, 8, 9].

To cope with adverse situations such as climate change, individuals and households adopt several strategies including urban-rural migration [10]. Several studies have recognized migration as one of the coping strategies to environmental factors [10, 11]. However, connections between migration and climate change are not well studied from a spatial perspective. Rural communities of east Africa’s Tanzania are following diverse strategies to cope with the climate change impacts, and internal migration is among the many options for regional livelihood improvement. However, complex linkages between climatic factors and migration in Tanzania have been less thoroughly investigated, especially from a spatial perspective.

Therefore, this study assesses the relationship between climate drivers and migration in Tanzania by explicitly incorporating spatial analysis. We statistically described migration pattern responses to climate drivers, and empirically assessed climate driver-migration pattern relationships using regression and geospatial techniques. Through these efforts we tested the following hypotheses: spatial patterns of temperature and precipitation will be different for migrant and non-migrant households; and temperature and precipitation influence migrant more than non-migrant households.

2. Previous studies of migration and climate change

Climate-oriented scientists agree that the likely harmful effects of climate change on rural economies have grown stronger over time. According to [12], almost 870 million globally are chronically undernourished amid climate models predicting rising temperature and declining precipitation for most of Sub-Saharan Africa [6, 12]. Many sub-Saharan economies dominantly depend on climate-sensitive agricultural production, and this region’s people are already facing starvation with more than 60% living below the poverty line [13].

According to the Intergovernmental Panel on Climate Change, about 150–200 million people will likely be displaced globally by 2050 due to anthropogenic climate changes [14]. Consensus among global scholars suggests climate change induced migration will worsen in the future [15]. Warming temperatures will have severe effects over the interior semi-arid margins of the Sahara and central southern Africa [1]. Similarly, Africa has remained hot and dry, but is warmer than it was 100 years ago [16, 17, 18].

Sub-Saharan Africa precipitation patterns are highly influenced by inter-seasonal and inter-annual climate variability including occasional El Niño-Southern Oscillation events [19, 20, 21]. Eastern equatorial Africa is heavily affected by extreme meteorology during the short October–November rainy season, while southeastern Africa experiences similar conditions during the main rainy season in November–February [22, 23, 24]. Hulme et al. [16] illustrates the nature of rainfall variability for the Sahel, East Africa, and southeastern regions, and suggests these three regions show contrasting variability. The Sahel shows large multi-decadal variability with recent drying, while east Africa demonstrates a relatively steady regime with some evidence of long-term wetting. By contrast southeast Africa tends to exhibit stability, with some noticeable inter-decadal variation [22], indicating higher latitudes are more susceptible to climate forcings.

Tanzanian climate projections show an expected mean annual temperature increase of 1.7–2.5°C by the 2060s, suggesting a stronger increase than the global average [25]. Similarly, mean annual rainfall patterns are also projected to increase across the country, but with a complex seasonal pattern highlighted by increased January and February rainfall, most dramatically in the far south. For northern Tanzania, highest rainfalls are expected for March, April, and May. During June, July, August, and September, precipitation is projected to increase in the very north of the country, while central and southern Tanzania expect declining rainfall [26].

Migration is a key rural livelihood strategy to increase household earning potential [27], reduce income risk, and shield against socioeconomic and environmental shocks [28]. Migration is a family response to income risk where migrants serve as an income insurance policy for their households of origin [28]. Households effectively diversify income sources by allocating labor to areas with a different set of risks from those faced in the source region, thereby building resilience to various livelihood shocks [29].

In their study on Ecuador, [30] show agricultural shocks are key factors in international migration. Munshi et al. [31, 32] have shown strong links between climate change, crop yields, and migration, whereas [33] discovered an opposite relationship suggesting Mexico to U.S. migration decreases as rainfall declines. Ethiopian studies show rural out-migration responds to drought sensitivity [34, 35].

There seems to be lack of consensus on the role of disasters in international migration with some arguing for a positive link [36, 37, 38, 39], while others indicate little or no or a negative relationship [40, 41]. For instance, [40] found flooding has modest to insignificant impacts on migration. On the other hand, [41] discovered people did not migrate after the occurrence of disaster in Bangladesh.

Meze-Hausken [35] showed weather anomaly impact on international migration has two channels. First, weather anomalies will lead to lower rural wages, especially if the effect of weather anomalies on agricultural production is strong. Second, lower rural wages will attract more mobile workers to move from the rural areas to cities in search of work. Consequently, more people settle in urban areas and thus increase urbanization. Therefore, weather anomalies are a key determinant of increased urbanization.

The growing literature examining the migration determinants increasingly emphasizes the role of environmental change in in-migration processes. Nawrotzkia and Maryia [42] show there are different ways to measure climatic factors. Heat waves, cold snaps, droughts, and excessive precipitation can be thresholds to construct climate measures. Schlenker and Roberts [43] investigated temperature effects on corn, soybeans, and cotton based on nationwide crop yield data and growing season climate information. Their findings indicate threshold temperatures for each crop with temperatures above the respective thresholds leading to yield reductions.

The literature therefore suggests strong links between climate change and migration. Climate change and migration relationships are relatively well researched from a non-spatial perspective. Spatially explicit empirical research in this context is in its infancy, especially for regions like Africa. Therefore, the present study seeks to address the paucity of empirical studies in the region. Another important knowledge gap is the literature’s focus on precipitation or temperature variation, while neglecting the effect they may have jointly on migrant decisions. Therefore, a clear understanding of climate change impacts on migration requires assessment of spatial links between climate change and migration. Further, impacts of temperature and precipitation variations on individual and household migration decisions need to be assessed in the same framework.

3. Study area

The Republic of Tanzania is located south of the equator at 6°00′S 35°00′E bordering the Indian Ocean and eight eastern and southern African countries. The country is bordered on the south by Mozambique, Malawi, and Zambia; on the west by Zaire, Burundi, and Rwanda; on the north by Uganda and Kenya; and on the east by the Indian Ocean (Figure 1). The country’s total area is 947,300 km², with 885,800 km² covered by land surface, and 61,500 km² of water.

Figure 1.

Study area relative location.

According to the Tanzania Bureau of Statistics [44], the country has a total population of 45 million with a population density of 51 persons/km². Tanzania has 30 geographic regions, 25 on the mainland and 5 islands (Figure 2). Administratively, the country is divided into regions, districts, wards, and villages.

Figure 2.

Tanzanian administrative regions.

Agriculture remains the economy’s largest sector. According to The World Bank’s economic outlook [45], the agricultural sector contributes almost one quarter of Tanzania’s overall gross domestic product (GDP), accounting for 85% of country’s exports, and employing about 80% of the work force. The majority of the country has a tropical climate with variations across regions. Except for the eastern coastal strip, most of the country is highland or a central plateau approximately 900–1800 m above sea level, with mountain ranges including Mount Kilimanjaro at 5895 m [26].

Tanzania’s regions exhibit topographic and seasonal variations in temperature. Accordingly, highland temperatures vary between 10 and 20°C during cold and hot seasons respectively. In the other parts of the country, temperatures rarely fall below 20°C. November through February is the hottest period, ranging between 25 and 31°C, whereas the coldest period occurs between May and August with temperature ranges of 15–20°C [13].

Seasonal Tanzanian rainfall is heavily influenced by Inter-Tropical Convergence Zone (ITCZ) migrations. The ITCZ moves southwards through Tanzania from October to December, reaching the south of the country in January and February, and returning northwards in March, April, and May. This migration brings northern and eastern Tanzania two diverse wet periods with short rains in October to December, and long rains in March to May. Conversely, the southern, western, and central parts of the country experience only one wet season that extends from October through to April or May. Seasonal rainfall varies regionally between 50 and 200 mm per month, with some regions receiving up to 300 mm per month in the wettest season [26].

4. Study data

Our study is based on high-quality household survey data called the Living Standards Measurement Survey-Integrated Survey on Agriculture (LSMS-ISA), initiated by the World Bank Development Economics Research Group, and implemented by the Tanzania Bureau of Statistics [44, 46]. National-level longitudinal data was collected between 2009 and 2013. The present study is specifically based on the 2012–2013 data. The LSMS-ISA data are composed of information on households, agriculture, and community characteristics. Each survey household is associated with a georeferenced sub-village.

Sub-villages represent enumeration areas where households were selected for the survey. However, an enumeration area is not a community from the sociological aspect; instead they are designated for information collection about the study areas where households selected for the intended study are located. A total of 26 regions and 149 sub-villages were considered in this study. Some observations were removed due to incomplete surveys and georeferencing errors, generating a final sample size of 10,461 households. Accordingly, 3968 and 6493 observations were identified for non-migrant and migrant households respectively. Non-migrants were identified in 62 enumeration areas whereas migrants were located in 75. The survey tracked all household members 15 years or older. We focused on these households and individual members aged 15–65 years. A key study variable was identification of migrant and non-migrant individuals. A migrant is an individual in a household who has left his or her initial residence and considers himself or herself to have settled in a new community. Despite the foregoing definition, the study follows the new economics of migration approach as indicated by previous work [47, 48, 49, 50], where migration is a collective action made by a household. A household member migrates not only to maximize household income for economic reasons, but also to minimize risk. Therefore, this study analysis used household-level data.

In addition to the LSMS-ISA data, this study incorporated 0.5° gridded historical climate data from the University of East Anglia’s Climate Research Unit [51] for 1983–2012. As indicated in the Results section, the 30 year (1983–2012) gridded temperature and precipitation data were downloaded and assigned into sub-villages using ArcGIS and STATA software. Temperature and precipitation data were analyzed for anomalies as well as for long-run (30 year) temperature and precipitation means.

Variable selection was based on the following studies: tenure security, land size, distance to market [52]; age, number of male and female adults, literacy, tropical livestock unit (TLU) [53]; extension of advice, soil fertility [54]; and climatic variables [11]. Additionally, [54] used social links and irrigation potential as measures of information access and land quality respectively. Similarly, we used extension advice (or access to information) and soil fertility as proxy measures of land quality.

5. Methods

This study examined the ability of 12 independent variables to explain household migration decisions. Two variables are climatic (temperature and precipitation), while four involve household demographics (household head age, household literacy, males over 15, and females over 15). Six variables define household economic characteristics: availability of extension advice, land ownership, TLU, soil quality, total land area, and distance to market.

We analyzed spatial patterns of precipitation and temperature across the regions with ArcGIS. The georeferenced datasets were converted into ArcGIS shapefiles using the Arc 1960 projection for Tanzania [55]. Households were divided into migrant and non-migrant categories. A household with at least one member who migrated during the survey year (2012) was considered a migrant household. All households in the same sub-village were assigned the same locational coordinates to merge climate data with each household. Temperature and precipitation changes were identified and first standard deviation data were incorporated into the analysis [43]. These data were mapped to show spatial migration patterns for migrant and non-migrant households across the country’s regions. Coefficients of variation (CoV), defined as σ/μ, for each sub-village’s rainfall and temperatures were displayed and analyzed using choropleth and graduated symbol maps. Average maximum temperature and rainfall represented annual extreme conditions. Long-run average rainfall and temperature represent the mean of yearly average values. Anomalies were the difference between the annual average and the 30 year annual mean. Anomalies are also shown using choropleth and graduated symbol maps. Additionally, household exposure to climatic variables was measured by the between-years rainfall CoV, for the period 1983–2012. CoV provided several advantages. First, for a given level of standard deviation, the CoV changes as the mean changes showing a lower level of variability for sub-villages with higher levels of average rainfall or temperature. Second, CoV shows dispersion of temperature and rainfall values in relative terms, allowing comparison between two sub-villages. We also computed sub-village average precipitation shortfall (i.e., the average of the annual totals’ departures from the long-run average). The same procedure was followed for study period temperature.

Migrant and non-migrant household demographic characteristics were analyzed using descriptive and inferential statistics for variables such as age, sex, household size, female head of household, literacy status of household head, male and female workforces, and households’ access to agricultural extension. Migrant and non-migrant households; distance from household farm plots to market; total land holding size; and total number of Tropical Livestock Units (TLU) were analyzed with t-statistics.

Our control variables and migration are related as follows. Household educational level is expected to increase migration because better-educated individuals are more likely to have information about migration and job availability in urban areas than are less educated individuals [56]. Total land holding is expected to decrease migration as larger land size corresponds to increased labor requirements for the household. Conversely, larger family size is expected to increase migration since households may have an excess labor supply.

The relationship between household head age and migration is expected to be mixed. If the household head is middle-aged, he may send a family member to migrate since migration may not affect family labor demand. Alternatively, an elderly household head may not be able to perform demanding tasks and may not be able to afford to expend family members on migration.

As a wealth indicator, TLU is expected to have a mixed effect on migration. Poorer households (low TLU) may lack financial resources for migration [57]. However, low income families tend to engage in short distance migration, while better income families may choose long distance migration [58]. Furthermore, gender may be important as female members of the household may not migrate long distances.

A household holding title to land is more secure than a landless household. Increased tenure security may discourage migration. Likewise, households having good soil quality could prefer working their land instead of migration. Household plot distance to market, on the other hand, may positively influence migration since a household close to market may have better labor market information and therefore opt to migrate.

Besides descriptive statistics, t-tests, and spatial analyses, we employed a logistic regression analysis to explain migration. Our analysis was framed following [59] approach, presenting migration as a determinant of a set of explanatory variables. We extended this approach to feature rainfall and temperature as key migration determinants. The logit specification assumes the household chooses migration m if the utility derived from the choice of m is greater than the decision not to migrate. As the utility from migration/non migration is unobservable, it can be expressed as a function of observables in the latent variable model, given in Eq. (1):

$${M_i}^{\ast }={\mathit{\alpha Z}}_i+{\epsilon }_i\mathit{\text{with}}{M}_i=\left\{\begin{array}{cc}1& \mathit{\text{if}}{M_i}^{\ast }\\ 0& \mathit{\text{otherwise}}\end{array}=0\right.$$E1

where $\mathit{Mi}$ is a dummy variable for the choice of migration; M$i=1$ if the household has chosen the decision to migrate, and M$i=0$ otherwise. $\alpha$ is a vector of parameters to be estimated; $\mathit{Zi}$ is a vector that represents household characteristics; and $\mathit{\epsilon i}$ is the random error term.

Based on Eq. (1), the estimated relationship between migration and temperature and precipitation is given by:

$$ln\left(m_h\right)={\alpha }_h+{\mathit{\beta x}}_h+{\mathit{\tau d}}_h+{\mathit{\rho i}}_h+{\mathit{\mu e}}_h+{\mathit{\phi j}}_h+{\epsilon }_h$$E2

where $h$ denotes household for migrant or non-migrant family. Migration is denoted by $m_h$, $x_h$ represents the temperature for each household h, while $d_h$ represents the precipitation for household h over a 30 year period. The set of socio-economic characteristics are denoted by $i_h$ while $e_h$ and $j_h$are the set of institutional and the physical capital variables respectively. The coefficients$\beta$,$\tau ,\rho$, $\mu$ and$\phi$ denote the respective vector of parameter estimates, and ${\epsilon }_h$ indicates the error term.

The model’s dependent variable is set to a 0–1 dummy variable, where 1 represents migrants and 0 represents the non-migrants household. Accordingly, the predicted values for the dependent variables will fall between 0 and 1 interval. These results will show the probability of households deciding to migrate.

As previously indicated, this study’s first hypothesis states that temperatures and precipitation show different spatial patterns between migrant and non-migrant households across sub-villages in the region. Inferential t-test statistics compared the effects of temperature and precipitation as well as the other control variables on migrant and non-migrant households. The second hypothesis states temperature and precipitation have more influence on migrant than non-migrant households with differing impacts across space. This hypothesis was tested using regression analysis. Regression coefficients quantified the (positive or negative) impact of a unit change in that variable on the propensity to migrate.

Our control variable decisions were based on [60, 61]. Since climatic variables and control variables are exogenous (unaffected by patterns of migration), a logit equation can be estimated without endogeneity concerns. Additionally, we addressed multicollinearity concerns with an independent variable correlation analysis. None of the pairs of variables were highly correlated (Table 1).

	TLU	Soil	Title	Area	Temp	Precip	Dist to market
Good soil quality	0.0165	1
Has title	0.0142	0.0132	1
Land area	−0.1986	−0.0604	−0.0449	1
Average temperature	0.0492	−0.0131	−0.0474	−0.0196	1
Average precipitation	0.0797	−0.0514	−0.0016	−0.0069	0.1582	1
Distance to market	−0.019	−0.0397	−0.0068	−0.0735	0.0271	0.1042	1

Table 1.

Key independent variable pearson correlation matrix.

6. Results and discussion

Table 2 presents demographic characteristics for non-migrant and migrant households. Average long-run temperature and precipitation for the non-migrant households were 23.59°C and 879.37 mm, respectively. The long-run temperature and precipitation for the migrant households were 23.72°C and 876.38 mm. The average age of household heads was 49 years for non-migrants and 50 for migrants, showing little between-group difference. On average there were two males and two females for non-migrant households and three each for migrant households. Seventy percent of non-migrant household heads and 74% of migrant household heads were literate. Household labor supply averaged 1.5 males and 1.6 females for migrants, and 1.2 males and 1.3 females for non-migrants. The household wealth measure (TLU) varied from 3.9 for migrant households to 2.7 for non-migrants.

Non-migrant			Migrant		Diff	p-value
Variable	Mean	s	Mean	s
Gender	1.28	0.45	1.2	0.4
Avg. Temp.	23.6	2.7	23.7	2.9	−0.12	0.03
Avg. Precip.	879.4	371.7	876.4	407.0	3.0	0.71
Age	49.4	16.9	50.1	15.2	−0.78	0.01
Literacy status	1.3	0.5	1.3	0.4	0.03	0.00
Male workers	1.3	1.0	1.5	1.1	−0.25	0.00
Female workers	1.3	0.9	1.6	1.1	−0.25	0.00
TLU	2.7	8.2	3.9	10.0	−0.27	0.00
Land area	6.7	8.5	10.4	20.0	−1.19	0.00
Extension	0.09	0.3	0.1	0.3	−3.64	0.00
Dist. to market	11.8	6.6	10.0	7.6	0.20	0.05
Good soil	0.5	0.5	0.4	0.5	0.02	0.05
Has title	0.1	0.3	0.2	0.4	−0.05	0.00

Table 2.

Summary statistics for selected variables by migration status.

Land-related analysis is also considered for both categories of households. Land ownership (tenure security) impacted a household’s migration decision. On average, 11.8% of non-migrant household and 17% of migrant households have title to land, with migrant and non-migrant households owning 3.61 and 2.33 hectares of land respectively. Household information availability was important to migration decisions as 13% of migrant households received agricultural and livestock advice, but only 9% of non-migrant households were well informed.

Bio-physical variables included household plot distance to market and soil quality. Non-migrant household plots average 11.8 km from market, while migrant household plots were only 9.9 km from market. By contrast, non-migrant households typically had better quality plots (46% good soil), compared to 44% good soil for migrant households.

Next, we show spatial patterns of climatic factors as well as distance from household plots to nearest market and land-size distributions by region. Figure 3 depicts the migration patterns by sub-village. Yellow dots depict sub-villages where non-migrants are dominant, while brown dots represent sub villages where the majority are migrants. While migrants and non-migrants are evenly distributed, there are more non-migrants in the northern part of the country as opposed to the southern. Similarly, southeastern Tanzania is non-migrant dominated compared to its northwestern section which has more migrants.

Figure 3.

Tanzanian migration patterns.

Figures 4 and 5 show average annual temperature and precipitation by sub-village for the period 1983–2012, and indicate spatial variation in highest average annual temperatures from 26.67 (southwest) to 29.52°C (northwest). Lowest temperatures ranged from 15.18 to 18.25°C as one travels northwards in Tanzania. Average annual precipitation for 1983–2012 is shown in Figure 4, which ranges between 1616.23 and 2113.44 mm, which is mainly received in parts of the southern and central sections of the country.

Figure 4.

Long-run temperature by sub-village and region.

Figure 5.

Long-run precipitation by sub-village and region.

Figures 6 and 7 illustrate long-term temperature and precipitation conditions through Tanzanian space. The coefficient of variation is defined as the long-term deviation divided by the long-term mean for temperature and precipitation taken independently (Charles et al. 2005). A negative coefficient of variation indicates an area is semi-arid or arid, while a positive coefficient of variation indicates sub-humid and humid environments. Figure 6 shows regions of low variability were observed in the central and eastern part of the regions for temperature. As shown in Figure 7 high precipitation variability areas were dominantly in the northwest, central, and southern regions.

Figure 6.

Long-run temperature coefficient of variation (CoV).

Figure 7.

Long-run precipitation coefficient of variation (CoV).

Figure 8 shows average distance from household plots to the nearest market by sub-villages. Shortest distance ranges varied from 0.00 to 5.80 km, with the greatest variability observed in the east. On the other hand, longest observed distances ranged from 81 to 190 km along the northern, central and southern regions. Figure 9 depicts household land size spatial distributions across the country. In general, plot sizes are smallest in the southeast, while plots are largest in the west and northeast.

Figure 8.

Household plot distance to nearest market.

Figure 9.

Household land size.

In general, the foregoing spatial analyses seem to show the following patterns and correlations between the migration and climatic variables. Long-run average precipitation statistical results show that the coefficient of rainfall has a significant and positive impact on migration. Spatial results indicate south and central Tanzania mainly received the highest rainfall. Migration is significantly more likely from the southern and central sections of the country where rainfall is the highest. Spatial analyses also seem to show that distance of plot to market have a significant negative impact on migration. Moreover, spatial results indicate the shortest distance recorded ranged from 0.00 to 5.80 km, mainly observed in the eastern part of the country. Therefore, migration is significantly less likely from eastern parts of the country where average distance from plots to market is the shortest. Finally, household land-size tends to be a significant migration determinant. Spatial results indicate that western and northeastern Tanzania have most of the largest farm plots. Thus, the impact of land-size on migration will be stronger in the west and northeast.

In Table 3, we show the logistic regression outputs for the determinants of migration, in which the interpretation of the results is based on the log likelihood ratio. Despite a poor fit to the data as indicated by the Hosmer-Lemeshow (HL) likelihood ratio test significance (a pseudo R² of 3.6%), all 12 original variables but two were found to be significantly affecting migration.

Variable	Coefficient
	Estimate	Std error	p-value	B(exp)
Avg. Precipitation	0.0148	0.0075	0.048	0.0034
Avg. Temperature	−0.00003	5.38x10^−5	0.632	5.99x10^−6
Household head age	0.0041	0.0014	0.002	0.0010
Literacy status	−0.1220	0.0487	0.012	−0.0284
Male workforce	0.1306	0.0218	0.000	0.0304
Female workforce	0.2025	0.0238	0.000	0.0471
Extension advice	0.1920	0.0691	0.005	0.0437
TLU	−0.0017	0.0030	0.572	−0.0004
Soil quality	−0.0945	0.0413	0.023	−0.0220
Has title	0.3511	0.0606	0.000	0.0787
Area	0.0254	0.0027	0.000	0.0059
Distance to market	−0.0110	0.0014	0.000	−0.0025
Constant	−0.4619	0.2126	0.03
Log likelihood ratio	−6693.4207		HL chi-square	499.66
Pseudo R^2	0.036		HL p-value	0.0000

Table 3.

Logistic regression results for migration determinants.

The empirical analysis indicates climate expressed by long-run precipitation had a significant positive effect on migration, where a 1 millimeter rise in precipitation led to a likelihood of 0.0034 migration increase. By contrast, long-run temperature had an insignificant impact on migration. These findings have two implications. First, households respond to rainfall instead of temperature. Second, rainfall is a potential migration driver. It should be noted that the descriptive statistics discussed above show that there is a significant difference in temperature (and not precipitation) between migrant and non-migrant households’ locations. However, the descriptive statistics suggest differences in magnitude and not climatic factors which may impact migration. This observation alone, however, because the t-test is indicative of stock differences between the two variables with respect to migration while the logistic regression shows the impact of each of variables on migration of a given variable when everything else is held constant. There are multiple possible explanations for this apparent paradox in study results including the potential influence of flooding on migration decisions which is beyond the scope of this study.

Age of household head, household head literacy status, and male and female workforce variables demonstrate significant robust impacts on migration. Access to agricultural extension services also shows significant and strong effects on migration. This analysis suggests households with older members, family with more male and female workforce, and households with access to agricultural extension services tend to migrate out of their community more so than other households. On the other hand, higher level of literacy seems to discourage migration.

The total livestock unit (TLU) variable showed insignificant impact on migration. Since livestock ownership serves here as a proxy indicator for wealth in rural areas of Tanzania, a higher TLU household is a richer family. However, larger household income seems both to encourage and discourage a given member of a family to migrate. Empirical results suggest households working on good soil quality farms and with farm plots far from market tend not to migrate. Households with good soil conditions have the opportunity to earn more revenue from farm produce than households with poor soil. Thus, the household’s tendency to improve livelihoods by sending a household member to migrate would be lower for those with better soil quality. The negative effect of distance to market indicates household access to a nearby market reduces households’ tendency to send a family member to migrate. This outcome could occur because households closer to markets do not see the need to migrate because of greater economic opportunity compared to households farther from market [60].

Study results suggest families holding title to their land are 0.0787 times more likely to migrate. Total land area is also significantly related to migration. When considering both title of land and total land area, the analyses indicate both variables show significant positive migration impacts. These outcomes suggest households with more land holdings and secured land ownership tend to migrate. Since land serves as financial capital for tenure-secured households, land ownership might be used to leverage costs associated with migration.

7. Conclusions

Our study investigated two hypotheses. First, temperature and precipitation show spatial differences in migrant and non-migrant households in Tanzania. Second, temperature and precipitation have more influence on migrant than non-migrant households with differing impacts across space. Specifically, empirical analysis shows that long-run precipitation has statistically significant positive impacts on migration and suggests increased precipitation is a key driver of migration. Conversely, long-run temperature had insignificant effects on migration. Our findings seem to suggest that the amount of rainfall is more important for Tanzanian households since their livelihood depends on it. On the other hand, some level of temperature variation may be tolerable as long as the rain is falling.

In sum, this study provides insight into spatial climate driver impacts on migration using georeferenced household survey data and gridded historical precipitation and temperature data for Tanzania. The study also contributes knowledge on spatial climate driver impacts on migration by identifying relevant determinant variables for Tanzania. This study makes an important contribution on two fronts. First, it adds to the limited empirical literature on Africa that assesses the impact of climatic factors in a detailed manner. Second, it combines econometric and spatial perspectives in the analysis to quantify key relationships while illuminating spatial pattern differences for key variables.

A shortcoming of our study is the loss of spatial resolution due to sub-village aggregation of household-level survey data. This spatial aggregation greatly limited the spatial analyses’ flexibility. Consequently, future studies with better spatial data could provide more accurate results. In addition, the statistical tests indicated overall poor model fitness which could be rectified with additional explanatory variables and well-refined data.

An additional possibility worth exploring is a gender differentiated migration response to climatic and non-climatic shocks because adult men and women have distinct roles in agricultural activities and have different levels of land tenure security [62]. Therefore, investigating gender-based differences in migration would be an interesting addition to the literature.