23WHAT FACTORS INFLUENCE HOUSEHOLDS' EXPENDITURE ON _FISH? EVIDENCE FROM URBAN MALAWI_
Fredrick Mangwaya Banda1, Abdi-Khalil Edriss2, Wilson Wesley Jere3, Horace Phiri4 *Corresponding author E-mail: [email protected]
A R T I C L E I N F O Original Article Received: 21 May 2023 Accepted: 15 June 2023 doi:10.59267/ekoPolj2302507M UDC 338.314:639.22(689.7) Keywords:
capture fishery; per capita food expenditure; per capita fish expenditure; ordinary least squares regression; Tobit regression model.
JEL: Q22
A B S T R A C T
The purpose of this study was to find the drivers of fish expenditure in urban Malawi using Blantyre as a case study. Previous demand studies on the fisheries sector in Malawi have concentrated on the determinants of consumer choices and demand for tilapia fish thereby making information regarding households' general consumption pattern of fish scanty. Results show that statistically significant negative determinants of per capita fish expenditure include the price of vegetables, the price of rice, the number of adults in the household, and the household's per capita food expenditure away from home, while, the price index of food, price of maize, the predicted value of food, and household's income level are the statistically significant positive drivers of per capita fish expenditure. General policy implications arising from this study are that policymakers need to ensure that households have higher incomes and that food prices are kept low.
Introduction
Food has always been one of the most widely used indicators of poverty such that
every single effort to investigate any country's poverty level has always included
measurements of food expenditures and food consumption. It is due to the recognition
1 Fredrick Mangwaya Banda, Ph.D. Student, Lilongwe University of Agriculture and Natural Resources, Department of Agricultural and Applied Economics, P.O. Box 219, Lilongwe, Malawi, Phone +265999721073, Email: [email protected], ORCID ID (https:// orcid.org/0000-0001-5937-1162)
2 Abdi Khalil Edriss, Professor, Lilongwe University of Agriculture and Natural Resources (LUANAR), P. O. Box 219, Lilongwe, Malawi, Phone: +265888842268, Email: abdikhalil@ yahoo.com, [email protected], ORCID ID (https://orcid.org/0000-0002-1795-1983)
3 Wilson Wesley Lazaro Jere, Associate Professor, Lilongwe University of Agriculture and Natural Resources, Department of Aquaculture and Fisheries Science, Africa Centre of Excellence in Aquaculture and Fisheries (AquaFish ACE), P.O. Box 219, Lilongwe, Malawi, Phone: +265997723715, Email: [email protected], ORCID ID (https://orcid. org/0000-0001-9353-8907)
4 Horace Phiri, Department of Agricultural and Applied Economics, Lilongwe University of Agriculture and Natural Resources, Malawi, E-mail: [email protected], ORCID ID (https://orcid.org/0000-0003-0258-3775)
of the importance of food that the United Nations lists the eradication of hunger by 2030 among its 17 Sustainable Development Goals (SDGs). Similarly, the government of Malawi regards food security as being crucial in its poverty eradication efforts. This is evidenced by the fact that, in the year 2006, it came up with the national food security policy whose overall goal is to dramatically improve the access to food by all Malawians by, among others, ensuring that Malawians have the ability to purchase enough nutritious food in a dignified manner so as to live an active and healthy life.
In Malawi, the fisheries sector is divided into three sub-sectors namely; capture fishery, aquaculture, and ornamental or aquarium trade, as explained in the subsequent paragraphs.
Further divided into artisanal and commercial production, capture fishery is practiced on Malawi's lakes and rivers such as Lake Malawi, Lake Chilwa, Lake Malombe, Lake Chiuta, and Shire River. In terms of their contributions to total catch, these water bodies contributed 93.88 percent, 1.82 percent, 2.28 percent, 0.93 percent, and 1.09per cent, respectively, to the total fish production in 2020 (GoM, 2021). Lake Malawi is, particularly, significant for fish production in that it has over 800 endemic fish species, which are of both local and international scholarly importance and also act as a source of tourism. Specifically, it is the South Eastern arm of the lake which is highly productive due to the occurrence of seasonal hydrological events that result in a plentiful supply of food for the fish.
Aquaculture, on the other hand, is mainly practiced in ponds in upland locations of the country and it is, largely, practiced by smallholder farmers. Despite there being signs of growth, Malawi's aquaculture sector is still in its nascent stage. Production rose from about 800 tons in 2005 to about 7,672 tons in 2016 and 9, 399 tons in 2020. The sector employs about 15,465 smallholder farmers, 61.51 percent of them being males and 38.49 percent females (GoM, 2021). The farmers are loosely organized in farmer clubs such that as of 2020, the total number of recorded ponds in Malawi was 10, 000 which translated to a total pond area of 251.59 hectares (GoM, 2021). Ornamental or aquarium trade concentrates on Mbuna fish which is exported live to countries such as Germany, Hong Kong, Denmark, and France (GoM, 2021).
With respect to marketing, most of the high-valued fish from both aquaculture and capture fishery are sold to customers in the urban areas such as Blantyre, Lilongwe, and Mzuzu while the low-valued fish species are sold locally around the fish ponds and the other water bodies. Traders transport these fish species using buses, pick-ups, bicycles, and motorbikes. The fish is sold either fresh or processed so as to prevent loss of quantity and quality. The main fish processing methods used include sun-drying, smoking, and salting. Fresh fish, either frozen or chilled, is particularly commonly sold in areas close to aquaculture farms and Malawi's lakes and rivers. In terms of market outlets, fish is either sold in public markets or supermarkets or retail outlets which are, mainly, owned by aquaculture companies. In the public markets, fish is usually sold on the basis of size by piece, buckets, heaps, or units while in supermarkets and retail outlets, it is sold based on weight. According to Brummett (2000), the determinants of average retail prices of fish in Malawi include the fish market factors and the fish attributes.
In terms of participation in fish marketing, fish trading is dominated by males as only a small proportion of women is engaged in fish processing and trading. Factors contributing to the lack of women's participation in fish marketing include lack of capital and the traditional division of labour. The major species of fish that are caught and sold in Malawi, on which this study focuses, include Engraulicypris sardella (usipa), Copadichromis spp (utaka), Lethrinops spp (kambuzi); Clarias gariepinus (mlamba), Rhamphochromis spp (mcheni), Barbus paludinosus (matemba) and Tilapia species of Lake Malawi tilapia (chambo), hereafter referred to as usipa, utaka, kambuzi, mlamba, mcheni, chambo and matemba, respectively.
In recognition of the significant nutritional and economic value of fish, the government of Malawi developed the National Fisheries and Aquaculture Policy (NFAP) whose main goal is to promote sustainable fisheries resource utilization and aquaculture development in order to contribute to food and nutrition security and economic growth of the country (GoM, 2016). Specifically, the NFAP aimed at increasing fish supply in the country so as to increase Malawi's per capita fish consumption from 8.12 kilogrammes in 2014 to 10 kilogrammes by 2020 (GoM, 2016) something which it has, regrettably, failed to achieve since only 9.51 kilogrammes were consumed per capita in 2020(GoM, 2021). This development turned out to be very discouraging considering the numerous interventions the government of Malawi has put in the fisheries sector.
However, information about factors that influence households' expenditure on fish is very scanty because previous demand studies on the fisheries sector in Malawi have concentrated on the determinants of consumer choices and demand for tilapia fish (Chikowi et al., 2020), or analysing factors affecting the demand for fish products, namely; smoked fish, dried fish, tinned fish, and fresh fish (Nankwenya et al., (2014). This study, therefore, bridges this knowledge gap by determining factors that influence Malawian households' expenditure on food, in general, and expenditure on fish, in particular. Furthermore, the study estimates the marginal effects of each determinant of food expenditure so as to get a clear picture of key factors that influence the likelihood of food purchases in Malawi. Research questions include: (1) what factors affect a household's expenditure on both food and fish in Malawi? (2) Are there any interdependencies between fish consumption and the consumption of other food commodities in Malawi? Providing answers to these questions will help unearth immensely useful insights for policymakers on how they can improve Malawian households' food consumption, in general, and fish consumption in particular. This, ultimately, can help improve the welfare of people in the country. Thus, this study contributes to the literature on fisheries by finding the drivers of fish consumption in Malawi, an area that has scarcely been given adequate attention by most studies in the fisheries sector in Malawi.
Materials and methods
The target population for the study was households in the city of Blantyre. The households were selected using a multistage stratified sampling procedure. Firstly, on the basis of income levels, the city was stratified into high- and low-density strata.
Secondly, using simple random sampling, clusters were selected from each stratum. Thirdly, from each cluster, households were randomly selected to form the final sample. The study targeted either household heads or other members of the household who were primarily responsible for the purchase of food in the household as respondents. To determine the sample size, the study used the formula (Zikmund et al., 2009):
Where: n is number of respondents (households),
p is proportion of the population of the households in the city that were interviewed which, following Chikowi et al. (2020), was equal to 46 percent in this study.
q = 1 - p is the estimated proportion of failures. It was equal to 54 percent in this study.
z is the statistical confidence level. This study used a 95 percent confidence level which gave a z statistic of 1.96
e = the maximum allowance for error between the true proportion and the sampling proportion. For this study, the allowance of sampling error was not greater than 4 percentage points.
Using the above information, the representative sample size for the study was calculated as follows:
However, upon factoring in the design effect, 514 households were added to the sample thereby adjusting the sample size for the study to 1,110 households. The data was collected through face-to-face interviews using semi-structured questionnaires. Specifically, the data collected included the socio-economic and demographic factors of the respondents and households' food purchase decisions. On consumption, information collected included the types and quantities of food consumed and the food expenditures in the past seven days. Consumers were also presented with Engraulicypris sardella, Copadichromis virginalis, Lethrinops spp.; Oreochromis karongae (Lake Malawi Chambo); Clarias gariepinus (Catfish); Rhamphochromis spp; Diplotaxodon; and Oreochromis shiranus (Shire Tilapia). Thereafter, quantities purchased of each fish species alongside their prices were elicited.
The study has used different variables and estimation methods on the basis of the research question being addressed. Specifically, the ordinary least squares (OLS) regression model has been used to find the drivers of households' per capita food expenditure while the determinants of households' per capita fish expenditure have been
investigated using the censored Tobit regression model. Table 1 provides a description of the variables that were used in the OLS per capita food expenditure regression model alongside their expected signs.
Table 1. A Description of the Variables that Were Used in the OLS Per Capita Food Expenditure Regression Equation
Variable Description Expected Sign
Dependent Variable
LnPCFE Natural log of per capita fish expenditure +
Independent Variables
PI Price index of fish +/-
Age Age of the household head +/-
Agesq Square of age of household head +/-
Years Years of formal schooling +/-
Primary Dummy (1-primary education, 0-otherwise) +/ -
Secondary Dummy (1-secondary education, 0-otherwise) +/ -
Hhsize Household size +/-
Children Number of children less than 7 years old in a household +/-
Adults Number of adults more than 13 years in a household +/-
Employed Number of people employed in a household +/-
Lnincome Natural log income level of a household -
Lnincomesq Natural log of the square of the income level of a household +
Female Dummy (1-if a household head is female, 0-otherwise) +/-
Unmarried Dummy (1-if a household head is married, 0-otherwise) +/-
Low Dummy (1-if a household is located in a low-density area, 0-otherwise) +/-
Source: Own compilations
As Table 1 shows that the dependent variable in the OLS per capita food regression equation was the natural log of the per capita food expenditure while the independent variables included socio-economic variables in addition to the price index of food. Table 2 presents variables that were used in the Tobit per capita fish expenditure regression model.
Table 2. A Description of the Variables that Were Used in the Tobit Per Capita Food Expenditure Regression Model.
Variable Description Expected Sign
Dependent Variable
LnFE Natural log of per capita fish expenditure +
Independent Variables
LnPO Natural lo of price index of food +/-
LnPV Natural log of price of vegetables +/-
LnPR Natural log of price of rice +/-
Variable Description Expected Sign
LnPMA Natural log of price of maize +/-
LnPCH Natural log of price of chicken +/ -
LnPE Natural log of price of eggs +/ -
FD* Predicted value of the per capita food expenditure +/-
FD*sqd Square of the predicted value of per capita food expenditure
Age Age of household head
Agesq Square of age of household head
Children Number of children less than 7 years old in a household +/-
Adults Number of adults more than 13 years in a household +/-
Employed Number of people employed in a household +/-
Lnincome Natural log income level of a household -
Lnincomesq Natural log of the square of income level of a household +
Female Dummy (1-if a household head is female, 0-otherwise) +/-
Married Dummy (1-if a household head is married, 0-otherwise) +/-
Lnpcfew Natural log of a household's per capita food expenditure away from home +/-
Source: Own compilations
Table 2 indicates that the dependent variable in Tobit per capita fish expenditure regression model was the natural log of the per capita fish expenditure. Additionally, it shows that variables such as prices of various food commodities and the household's per capita food expenditure away from home are added to the list of the independent variables used in the OLS regression model.
Results and Discussions
This section presents and discusses the results of various regression models that have been used to address the study objectives. In particular, it focuses on the results from the food expenditure regression equation and the Tobit fish expenditure regression model.
Results from Food Expenditure Regression Equation
Using variables presented in Table 1, the following OLS regression equation was estimated:
InPCFE = fi PI, Y, Z, fj) (1)
Where:
InPCFE is the natural log of per capita food expenditure; PI is the household specific food price index;
Visa vector of per capita household income which was expressed in both linear and quadratic forms;
Z is a vector of household characteristics, and Si is the error term.
Following Gujarati (2004), the study endeavoured to test and correct for the presence of the problems of non-normality, multicollinearity, and heteroskedasticity during the estimation of the OLS regression. The results of the foregoing tests are presented in the subsequent paragraphs.
Results of Normality Test
The normality assumption is one of the key assumptions underlying OLS regression model whose violation results in inflated standard errors leading to wider confidence intervals and wrong hypothesis testing. Ultimately, the violation of the normality assumption leads to biased estimates. A Kolmogorov-Smirnov test was, therefore, used to test for the presence of non-normality in the data used in the study. The Kolmogorov-Smirnov test gave a Chi-squared statistic with a p-value of 0.0000<0.01 thereby suggesting the rejection of the null hypothesis no-normality. This implies that the data was normally distributed.
Results of Multicollinearity Test
The presence of multicollinearity problem results in inefficient or inconsistent parameter estimates and inaccurate p-values thereby making it difficult to separate the independent effect of each parameter estimate on the dependent variable (Gujarati, 2004). Hence, in order to detect and measure the presence and the severity of the multicollinearity problem, the study employed the variance inflating factors (VIF) results of which are shown in Table 3.
Table 3. Results of the Multicollinearity Test
Variable VIF tolerance
Lnincomesq 271.13 0.003688
Lnincome 270.74 0.003694
Age 45.41 0.022022
Agesq 44.62 0.022413
Years 5.68 0.176206
Secondary 3.91 0.255430
Primary 3.78 0.264871
Hhsize 2.58 0.388150
Adults 2.50 0.400565
Children 1.38 0.725149
Employed 1.37 0.729368
Low 1.17 0.855924
Female 1.09 0.918474
Variable VIF tolerance
PI 1.07 0.932967
Unmarried 1.01 0.988765
Mean VIF 43.83
Source: Own compilations
As indicated by Table 3, the mean VIF value for all the explanatory variables used in the OLS regression model was 43.83 thereby suggesting that there was a multicollinearity problem in the variables used in the OLS regression model since many of the regressors had VIF values far more than 10. Therefore, to control for the multicollinearity problem, the OLS regression equation was estimated without a constant.
Results of Heteroscedasticity Test
The presence of heteroscedasticity in the data set has very damaging consequences in that it results in high standard errors, which in turn, leads to wider confidence interval problems and biased parameter estimates. The study used the Breusch-Pagan test for heteroskedasticity which yielded a chi-squared statistic value of28.23 with an associated p-value of 0.0000. This led to the rejection of the null hypothesis of constant variance thereby suggesting the presence of heteroscedasticity in the data set. To surmount the heteroscedasticity problem in the study, robust standard errors were used during the estimation of the OLS regression equation, the results of which are presented in Table 4.
Table 4. Parameter Estimates of the Food Expenditure Regression Equation
Variable coefficient robust standard error t-statistic p-value
PI 0.1587*** 0.0511 3.11 0.002
Age 0.0121 0.0130 0.93 0.352
Agesq -0.0001 0.0002 -0.74 0.463
Years 0.0061 0.0192 0.32 0.751
Primary -0.3060 0.1968 -1.55 0.121
Secondary -0.0282 0.0748 -0.38 0.706
Hhsize -0.1816*** 0.0255 -7.13 0.000
Children 0.0166 0.0529 0.31 0.754
Adults 0.0022 0.0256 0.08 0.932
Employed 0.0513 0.0338 1.52 0.129
Lnincome 1.0969*** 0.0605 18.12 0.000
Lnincomesq -0.0187*** 0.0032 -5.88 0.000
Female 0.0337 0.0308 1.09 0.275
Unmarried 0.1617** 0.0769 2.10 0.036
Low -0.0122 0.0566 -0.22 0.829
F (15, 589) = 42678.03 Prob > F = 0.0000
Source: Own compilations Note: Asterisks represent level of statistical significance: *(10% significance), **(5% significance), ***(1% significance).
As shown in Table 4, statistically significant positive determinants of the per capita food expenditure are the price index of food (p-value<0.01), the household's income level (p-value<0.01), and the household head's status of being unmarried (p-value<0.05). This suggests that an increase in the price index of food, an increase in the household's level of income, and the household head's state of being married will lead to an increase in per capita food expenditure. However, it is noteworthy that the coefficient of the squared income term is negative and statistically significant at a 1 per cent level of significance (p-value<0.01). This finding conforms to the finding by Dey et al., (2000) who found that the fish expenditure's response to changes in income is significant but non-linear. It, particularly, implies that as households' income levels progressively increase, per capita food expenditure also increases up until it reaches a maximum, beyond which, any further increase in the income levels results in a decrease in per capita food expenditure. This finding conforms to Engel's law which states that as a household's income level increases, the percentage of the income allocated for food purchases decreases.
Furthermore, Table 4 shows that household size is the statistically significant negative determinant of per capita food expenditure (p-value <0.01). This suggests that as the number of people in a household increases, there are more people who have to share in the household's food budget thereby lowering the household's per capita food expenditure. This finding is consistent with the finding by Dey et al., (2000) who found that bigger households negatively influence the per capita food expenditure in Bangladesh.
In addition to the results presented in Table 4, the study computed conditional marginal effects for each variable. By definition, conditional marginal effects are the elasticities of each variable computed at a specific value. The specific value used in this study is the mean. Table 5 presents the marginal effects of the variables.
Table 5. Marginal Effects of the Parameter Estimates of the Food Expenditure Regression
Equation
Dependent variable: Per capita food expenditure
variable coefficient standard error t-statistic p-value X
PI 0.1587*** 0.0511 3.11 0.002 -0.2066
Age 0.0121 0.0130 0.93 0.351 35.548
Agesq -0.0001 0.0002 -0.74 0.462 1328.92
Years 0.0061 0.0192 0.32 0.751 13.6722
Primary -0.3060 0.1968 -1.55 0.120 0.0497
Secondary -0.0282 0.0748 -0.38 0.706 0.3891
Hhsize -0.1816*** 0.0255 -7.13 0.000 4.3676
Dependent variable: Per capita food expenditure
variable coefficient standard error t-statistic p-value X
Children 0.0166 0.0529 0.31 0.754 1.1242
Adults 0.0022 0.0256 0.08 0.932 2.7368
Employed 0.0513 0.0338 1.52 0.129 1.3874
Lnincome 1.0969*** 0.0605 18.12 0.000 12.9607
Lnincomesq -0.0187*** 0.0032 -5.88 0.000 168.745
Female 0.0337 0.0308 1.09 0.274 0.6722
Unmarried 0.1617** 0.0769 2.10 0.035 0.0132
Low -0.0122 0.0566 -0.22 0.829 0.0912
F (15, 589) = 42678.03 Prob > F = 0.000
Source: own calculations Note: Asterisks represent the level of statistical significance: **(5% significance), ***(1% significance).
As shown in Table 5, for a unit change in the food price index, there will be a 15.87 percent increase in the per capita food expenditure. Similarly, for a unit change in the household size, there will be an 18.16 percent decrease in the per capita food expenditure. Also, Table 5, shows that for a unit change in a household's income level, there will be a 110-percentage increase in per capita food expenditure. Lastly, the change of marital status of the household head from being married to being unmarried leads to a 10-percentage increase in the per capita food expenditure.
Results from Fish Expenditure Tobit Regression Equation
In this study, the model for the determinants of per capita fish expenditure the dependent variable attains values equal to or greater than 0. In other words, the dependent variable is censored from below at 0. Consequently, applying OLS to a regression equation whose dependent variable, as in this case, is censored from below yields biased results. Hence, this study used the Tobit model to estimate the determinants of per capita fish expenditure. The Tobit model where the distribution of the dependent variable is
censored from below at point a can be expressed as follows:
Such that V; = = X$ + e, (2)
Where: Y* is the latent variable and, in this study, it represents per capita fish expenditure,
xt represents a vector of regressors as shown in Table 2, ¡3 represents a vector of slope coefficients, and
e, ™N (0, a2) is the error term which is normally distributed with mean equal to zero, and has a homoscedastic variance equal to a2.
The Tobit regression model was estimated using robust standard errors and without a constant so as to control the problems of heteroscedasticity and multicollinearity, respectively. The results of the Tobit fish expenditure regression equation are presented in Table 6.
Table 6. Parameter Estimates of the Fish Expenditure Tobit Regression Equation
Dependent variable: Per capita fish expenditure
variable coefficient robust standard error t-statistic p-value
LnPO 0.6698*** 0.0514 13.04 0.000
LnPV -0.1087* 0.6020 -1.81 0.072
LnPR -0.1779** 0.0858 -2.07 0.039
LnPMA 0.1915* 0.1118 1.71 0.083
LnPCH 0.1620 0.1134 1.45 0.148
LnPE 0.0185 0.0318 0.58 0.562
FD* 0.5545*** 0.1830 3.03 0.003
FD*sqd -0.0560*** 0.0185 -3.03 0.003
Age 0.0098 0.0243 0.40 0.688
Agesq -0.0001 0.0003 -0.40 0.687
Children 0.0330 0.0999 0.33 0.742
Adults -0.1108** 0.0477 -2.32 0.021
Employed 0.0478 0.0591 0.81 0.419
Lnincome 1.1668*** 0.1288 9.06 0.000
Lnincomesq -0.0198** 0.0083 -2.38 0.018
Lnpcfew -0.1151*** 0.0377 -3.06 0.002
Female 0.0113 0.0550 0.21 0.838
Married -0.1125 0.1240 -0.91 0.365
F (18, 387) = 8619.00 Prob > F = 0.000 Pseudo R2 = 0.7976
Source: own calculations
Note: Asterisks represent level of statistical significance: *(10% significance), **(5% significance), ***(1% significance).
As indicated in Table 6, the estimated parameter of the price index for fish is positive and statistically significant at 1 percent level of significance (p-value<0.01), thereby, suggesting that an increase in the price index of fish will result in an increase in the per capita fish expenditure. This implies that fish is a necessity so much so that even if its price rises people will still continue buying it, thereby making the total expenditure on fish increase.
This finding is consistent with the finding by Dey et al. (2000) who found that fish is a necessity in Bangladesh. Additionally, the coefficient for price of maize is positive and statistically significant (p-value <0.1) implying that an increase in the price of maize leads to an increase in per capita fish expenditure. The coefficients of the linear and squared forms of the predicted food expenditure variable exhibited statistically significant positive and negative signs, respectively. This suggests that there exists a nonlinear relationship between food expenditure and fish expenditure. Particularly, it shows that as the budget for food increases, progressively, the per capita expenditure on fish also increases up until it reaches a maximum beyond which it eventually declines. This shows the general behaviour of respect to food consumption whereby consumers have a certain threshold for food consumption such that upon reaching that threshold, consumers can never be induced to purchase more food even if their income levels have increased. This finding is consistent with the finding by Dey et al (2000) who found that there exists a non-linear relationship between food expenditure and fish expenditure in Bangladesh.
Furthermore, Table 6 shows that the coefficients of income and the squared term of income are statistically significant at a 1 per cent level of significance (p-value<0.01), even though they are positive and negative, respectively. This, again, suggests that there is a presence of a non-linear relationship between per capita food expenditure and households' income level. It, particularly, implies that as households' income levels progressively increase, per capita food expenditure also increases up until it reaches a maximum, beyond which, any further increase in the income levels results in a decrease in per capita food expenditure. This finding conforms to Engel's law which states that as a household's income level increases, the percentage of the income allocated for food purchases.
Table 6 further indicates that the price of vegetables, the number of adults in the household, and the household's per capita food expenditure away from home are statistically significant negative determinants of per capita fish expenditure. Particularly, it shows that an increase in the price of vegetables, an increase in the number of adults in the household, and an increase in the household's per capita food expenditure away from home lead to a reduction in the per capita expenditure on fish.
The finding that the number of adults in the households has a negative influence on the households' expenditure on fish is consistent with the findings by Murray et al. (2017) who found that, in British Colombia of Canada, households spend more on fish as the number of young people in the household increases. These findings are, however, inconsistent with the findings by Thong & Solgaard (2017) who found that households with more elderly people spend more on fish compared to those with a small number of elderly people.
Conclusions
This study set out to investigate the drivers of both food expenditure and fish expenditure in Malawi. It has found that the positive determinants of a household's per capita food expenditure include the price index of food, the household's income
level, and the household head's state of being unmarried while the household size is the only statistically significant negative determinant of per capita food expenditure. Again, with respect to fish expenditure, the study has found that statistically significant negative determinants of per capita fish expenditure include the price of vegetables, the price of rice, the number of adults in the household, and the household's per capita food expenditure away from home. On the other hand, the study has found that the price index of food, the price of maize, the predicted value of food, and household income level are the statistically significant positive drivers of per capita fish expenditure.
Policy implications arising from this study are that in order to increase the consumption of food in Malawi, it is imperative that policymakers should ensure that households have a small number of people with a lot of income-generating activities. Having a lot of income-generating activities will enable households to spend more on food. Additionally, with respect to fish consumption, policymakers should do the following: make sure that the prices of vegetables do not just rise anyhow, make that households do have a small number of adults, through among others, the creation of a lot of employment opportunities, and make expenditure on food away from home unattractive. One of the possible ways of making food expenditure away from home is by charging higher taxes on restaurants and other food-eating joints. Other important policy options that can help increase households' per capita expenditure on fish include making sure households have a lot of income-generating activities and maintaining the price of food low. If these policies are successfully implemented, then the households' expenditure on both food and fish will improve thereby leading to an improvement in the welfare of Malawian households.
Acknowledgements
We acknowledge Africa Centre of Excellence in Aquaculture and Fisheries (AQUAFISH ACE) for providing funds that have been used during the study's data collection stage. Without their financial assistance this study would not have been successful. However, the authors take full responsibility of all the views expressed in this study as they do not reflect the views of AQUAFISH ACE.
Conflict of interests
The authors declare no conflict of interest.
References
1. Brummett, R.E. (2000), Factors influencing fish prices in Southern Malawi. Aquaculture, 186 (3-4), 243-251.
2. Cavaliere, A., De Marchi, E., Donzelli, F., & Banterle, A. (2019), Is the Mediterranean diet for all? An analysis of socioeconomic inequalities and food consumption in Italy. British Food Journal. doi.org/10.1108/BFJ-06-2018-0373.
3. Chikowi, C.T., Ochieng, D.O. & Jumbe, C. B. (2020), Consumer choices and demand for tilapia in urban Malawi: What are the complementarities and tradeoffs? Aquaculture, 530 (20).
4. Dey, M. M. (2000), Analysis of demand for fish in Bangladesh. Aquaculture Economics and Management, 4(1/2), 63-81.
5. Dey, M. M., Garcia, T. Y., Kumar, P., Piumsombun, S., Haque, S. M., Li, L., Radam, A., Senaratne, A., Khiem, N.T. & Koeshendrajana, S. (2008), Demand for fish in Asia: A cross country analysis. The Australian Journal of Agricultural and Resource Economics, 52, 321-338.
6. Gujarati, D. (2004), Basic Econometrics. New York: McGraw-Hill Companies.
7. Lancaster, K. J. (1966), A new approach to consumer theory. Journal of Political Economy, 74(2), 132-157.
8. Malawi Government (2016), National Fisheries and Aquaculture Policy. Available at:http://www.unpei.org/malawi-government-national-fisheries-and-aquaculturepolicy-.
9. Malawi Government (2021), Annual economic report 2021, Ministry of Economic Planning, Development and Public Sector Reforms, Department of Economic Planning and Development, Lilongwe.
10. Murray, G., Wolff, K., & Patterson, M. (2017), Why eat fish? Factors influencing seafood consumer choices in British Columbia, Canada. Ocean & Coastal Management, 144,16-22. https://doi.org/10.1016lj.ocecoaman.2017.04.007.
11. Nankwenya. B., Kaunda, E. & Chimatiro, S. 2017), The demand for fish products in Malawi: An almost ideal demand system estimation. Journal of Economics and Sustainable Development, 8 (16), 63-71.
12. Thong, N. T., & Solgaard, H. S. (2017), Consumer's food motives and seafood consumption. Food Quality and Preference. 56, 181-188. doi.org/10.1016/j. foodqual.2016.10.008.
13. Phiri, L.Y., Dzanja, J., Kakota, T. & Hara, M. (2013), Value chain analysis of Lake Malawi: A case study of oreochromis spp (chambo). International Journal of Business and Social Science, 4(2).
14. Zikmund, W.G., Babin, B. J., Carr, J. & Griffin, M. (2010), Business Research Methods. 8th edition. Canada: South Western Cengage Learning.
