This section presents and discusses empirical results of the study. Table 2 presents the summary of the descriptive statistics for the variables.
Table 2
Variable
|
N
|
Mean
|
Maximum
|
Minimum
|
Standard Deviation
|
AGTRADE
|
120
|
3.50
|
9.33
|
0.54
|
2.20
|
TEMP
|
120
|
21.24
|
24.81
|
12.26
|
2.99
|
PREC
|
120
|
930.29
|
3061.66
|
231.80
|
463.88
|
AGDP
|
120
|
2.78
|
53.00
|
-21.18
|
8.90
|
FI
|
120
|
15.02
|
601.02
|
-3.34
|
55.60
|
FPI
|
120
|
102.43
|
141.74
|
33.60
|
17.11
|
PG
|
120
|
2.04
|
3.76
|
0.00
|
1.06
|
AGTRADE- agriculture trade balance; TEMP- temperature; PREC- precipitation; AGDP- agriculture GDP; FI- food inflation; FPI- food production index; PG- population growth; N- number of observations
Source: Author's constructions
From Table 2, AGTRADE had a mean of 3.50 showing that for the period 2012 to 2021, agriculture trade balance averaged about 3.5% of GDP. More so, for TEMP, the mean of 21.24 show that on average mean annual temperatures were around 21.24 Degrees Celsius. The mean of 930.29 for average annual precipitation also imply that mean annual precipitation in Southern Africa was approximately 930.29 millimetres. The mean of 2.78 for AGDP show that on average, the value of agricultural GDP has been 2.78% of national GDP in Southern Africa. Additionally, the mean of 15.02 for food inflation (FI) indicate food inflation in the Southern African region averaged 15.02% though it peaked to 601.02% during the same period. Lastly, the mean of 2.04 for PG indicate the on average, the population in Southern African countries grew by approximately 2.04% on an annual basis. Table 3 shows the results of the unit root tests.
Table 3
Variable | Level | First difference |
LLC | ADF | PP | LLC | ADF | PP |
lnAGTRADEit | -1.17 | 25.25 | 77.82*** | -5.91*** | 48.97** | 115.55*** |
lnTEMPit | 0.78 | 9.11 | 8.80 | -14.02*** | 130.74*** | 148.71*** |
lnPRECit | 0.60 | 10.28 | 7.62 | -12.69*** | 111.13*** | 179.62*** |
AGDPit | -7.16*** | 63.97*** | 109.41*** | | | |
FIit | -3.05** | 29.13 | 51.99 | -4.43**** | 38.98* | 126.43*** |
lnFPIit | 4.46 | 12.44 | 9.16 | -7.13*** | 79.47*** | 201.90*** |
lnPGit | -2.72** | 40.53* | 79.91*** | | | |
* P < 0.05; ** P < 0.01; *** P < 0.001; LLC- Levin, Lin and Chu; ADF- Augmented Dickey Fuller; PP- Phillips-Perron; AGTRADE- agriculture trade balance; TEMP- temperature; PREC- precipitation; AGDP- agriculture GDP; FI- food inflation; FPI- food production index; PG- population growth |
Source: Author's constructions |
The results presented in Table 3 show that two variables (AGDP and PG) were found to contain no unit root tests at level whilst the remaining five variables (AGTRADE, TEMP, PREC and FPI) were found to unit root tests at level. However, the series for the five variables became stationary after first differencing. In this regard, robust and unbiased estimates were obtained. Table 4 presents the collinearity matrix.
Table 4
| AGDPit | AGDPit | lnAGTRADEit | lnPRECit | lnTEMPit | lnPGit |
AGDPit | 1.00 | | | | | |
AGDPit | 0.16 | 1.00 | | | | |
lnAGTRADEit | 0.00 | -0.03 | 1.00 | | | |
lnPRECit | 0.05 | 0.06 | 0.09 | 1.00 | | |
lnTEMPit | -0.00 | 0.04 | -0.30 | 0.23 | 1.00 | |
lnPGit | 0.05 | 0.07 | -0.16 | -0.38 | 0.02 | 1.00 |
AGTRADE- agriculture trade balance; TEMP- temperature; PREC- precipitation; AGDP- agriculture GDP; FI- food inflation; PG- annual population growth
Source: Author's constructions
From Table 4, the correlations for paired independent variables were significantly less than 0.8 implying that there were no serious problems of collinearity among the independent variables. More so, baseline panel GMM regression estimations were done. From these, the Hausman test was undertaken and the results are presented in Table 5.
Table 5
Cross-section random | Chi-Square Statistic | Chi-Square Degrees of Freedom | Prob. | Decision at 5% level |
Model (1) | 0.001 | 6 | 1.00 | RE GMM is most appropriate |
Model (2) | 337.32 | 6 | 0.00*** | FE GMM is most appropriate |
* P < 0.05; ** P < 0.01; *** P < 0.001; RE- Random effects; GMM- Generalised Method of Moments; FE- Fixed effects
Source: Author's constructions
From Table 5, for model (1), a Hausman statistic of 0.01 was estimated with a p-value of 1.00 (P > 0.05) indicating that the RE GMM model was the most appropriate. On the other hand, for model (2), a Hausman statistic of 337.32 with a p-value of 0.000 (P < 0.001) implying that FE GMM model was the most appropriate. Besides, the JB test was undertaken to test for normality of residuals and the results are presented in Table 6.
Table 6
Jarque-Bera normality test
Model | Jarque-Bera statistic | Kurtosis | Skewness | Prob. |
Model (1) | 1.30 | 3.56 | 0.04 | 0.52 |
Model (2) | 0.89 | 2.93 | -0.23 | 0.64 |
Source: Author's constructions |
The results shown in Table 6 show that for model (1) (RE model) and model (2) (FE model), the Jarque-Bera statistics were 1.30 (P = 0.52 > 0.05) and 0.89 (P = 0.64 > 0.05). These show that the residuals followed a normal distribution. More so, the Kurtosis and Skewness statistics are close to the values three and zero respectively showing normal distribution (Hsiao, 2022). The results of the robust panel GMM regressions are presented in Table 7. For each of the two models, three models (pooled ordinal least square, RE GMM and FE GMM models) were estimated. However, basing on the results of the Hausman tests, results for the RE model and FE model were interpreted for models (1) and (2) respectively. As indicated in Table 7, model (1) and model (2) had coefficients of determination (R-squared) of 0.86 and 0.52 respectively. These results show goodness of fit. More so, models (1) and (2) estimated DW statistics of 2.31 and 1.66 respectively (Table 7). These DW statistics fell in the range 1.5 to 2.5 showing that the models did not suffer from autocorrelation (Hsiao, 2022).
Table 7
Results of robust panel GMM regression model) (Models 1–2)
Variable | Model (1) Dependent variable: lnAGTRADEit | Model (2) Dependent variable: lnFPIit |
Pooled OLS | RE model | FE model | Pooled OLS | RE model | FE model |
C | 1.31** (0.478) | 1.31*** (0.105) | -4.00 (2.642) | 3.46*** (0.480) | 3.51*** (0.038) | 2.76** (0.885) |
AGTRADE | 0.87*** (0.038) | 0.87*** (0.008) | 0.230* (0.120) | 0.001 (0.012) | -0.01*** (0.001) | -0.13*** (0.024) |
FPI | - | - | - | 0.16* (0.064) | 0.16*** (0.004) | -0.12** (0.04) |
TEMP | -0.25* (0.018) | -0.25** (0.008) | 1.54 (1.113) | 0.064 (0.067) | 0.04*** (0.006) | 0.09 (0.471) |
PREC | -0.06 (0.06) | -0.06*** (0.013) | 0.01 (0.281) | 0.04** (0.010) | 0.04*** (0.01) | 0.17* (0.087) |
AGDP | 0.01* (0.002) | 0.01*** (0.001) | 0.004 (0.003) | - | - | - |
FI | -0.001 (0.002) | -0.001*** (0.003) | -0.003 (0.003) | 0.001 (0.001) | -0.022*** (0.003) | -0.001 (0.001) |
PG | -0.04 (0.025) | -0.04*** (0.005) | -0.03 (0.021) | 0.04** (0.010) | 0.03*** (0.001) | 0.02** (0.001) |
Observations | 96 | 96 | 96 | 96 | 96 | 96 |
R-squared | 0.86 | 0.86 | 0.87 | 0.32 | 0.28 | 0.52 |
DW statistic | 2.31 | 2.31 | 1.78 | 1.24 | 1.27 | 1.66 |
* P < 0.05; ** P < 0.01; *** P < 0.001; robust standard errors in parentheses; AGTRADE- agriculture trade balance; TEMP- temperature; PREC- precipitation; AGDP- agriculture GDP; FI- food inflation; FPI- food production index; PG- population growth; DW-Durbin Watson; R-squared- coefficient of determination; RE- Random effects; GMM- Generalised Method of Moments; FE- Fixed effects; OLS- ordinary least square; C- Constant |
Source: Author's constructions |
As shown in Table 7, the RE GMM model (1) estimated the impacts of climate change on agricultural trade whilst the FE GMM model (2) estimated the impacts of climate change on food security in the Southern African region. For the RE model (1), the one-period lagged variables for the climate change variables (TEMP and PREC) were found to have statistically significant negative coefficients. For TEMP, the coefficient of -0.25 (P = < 0.01) show that a percent increase in mean temperatures in one year can result in about a 0.25% decline in agricultural trade flows in the succeeding season. This is because, increase in temperatures may significantly reduce agricultural productivity leading to low output supplied to the market. On the other hand, for PREC, the coefficient of -0.06 (P < 0.01) show that changes in precipitation negatively impacts agricultural trade. Precisely, a percentage change in the amount of precipitation can cause agricultural trade to fall be approximately 0.06%. These results show that climate change has significant impacts on agricultural trade in the Southern African region. Basing on these results, the hypothesis that climate change has significant negative impacts on agricultural trade could not be rejected. In other words, the results have proven that climate change is threat to agricultural trade flows in the region such that region may continue to be a net-food importer. The results corroborate those of Tekce and Deniz (2016) who also found negative impacts of climate change on agricultural trade.
Besides the negative climate change effects, other factors negatively impacting agricultural trade in Southern Africa have been found to include food prices proxied by food inflation (FI) (β= -0.001; P < 0.001) and PG (β= -0.04; P < 0.001). These have further implications in food security. Similar findings were also obtained by Adeste et al. (2022) and Affoh et al. (2022).
More so, from the FE GMM model (2), temperature (TEMP) was found to have insignificant effects on food security in the Southern African region (α = 0.09; P > 0.05). This could be a pointer of adoption of adaptation strategies to temperature changes in the Southern African region. On the other side, changes in precipitation (PREC) have significant positive effects on food security (α = 0.17; P < 0.05). These results show that a percentage increase in precipitation may cause an increase in agricultural trade flows by approximately 0.17%. The results show inconclusive effects of climate change on food security leading to the partial acceptance of the research hypothesis that climate change negatively impacts food security. The findings contradict findings of Mharous (2019) and Brenton et al. (2022) who found significant negative impacts of climate change on food security.