In order to identify the main patterns with respect to the level of household wealth in Caiomete and Braima Sori, as well as its association with other variables, a decision was made in the context of this research project to adopt an indirect approach, whereby ownership of durable household assets was used as a proxy for household wealth. The rationale for this approach lies in the fact that direct questions on monetary wealth and income often prove more sensitive and liable to voluntary and involuntary bias on the part of the respondents than questions on the current ownership of household assets.
Additionally, the characteristics of the context under analysis (a country where only a few thousand – mostly urban – people have bank accounts and where most inhabitants of rural areas store their wealth in a limited number of physical forms) presumably increase the likelihood that a selection of durable household assets such as the one adopted here provides a valid and reliable indication of the underlying, unobserved variable ‘wealth’ (or permanent income).
The specific procedure that was used, following Filmer and Pritchett (2001), McKenzie (2005) and Johnston and Wall (2008), consisted of using principal component analysis (PCA) as the basis for the construction of an asset index, which was then computed for all 108 households in the sample and used in the subsequent analysis. The results yielded by this approach have been shown in the literature to be consistent with expenditure and income measures (cf. references indicated above).
In this particular case, the procedure consisted of the following steps:
i) preparing a subset of the survey data comprising the 108 households (units of analysis) x 11 asset variables (ten categorical, one scale variable181);
181 The ten categorical variables assumed the value “1” in the case of the households that reported owning a given asset and the value “0” for those that did not. The ten durable household assets considered were: (i) diesel generator; (ii) radio or stereo; (iii) TV; (iv) cell phone; (v) bicycle; (vi) fridge; (vii) motorcycle; (viii) car (or van or truck); (ix) gas or electric stove; and (x) zinc rooftop. The
327 ii) running PCA on this data set, subject to the specification that only one principal component (the one with the greatest eigenvalue) was to be retained;
iii) saving the 108 component scores thus computed by SPSS as the vector of values assumed by our asset index si across the 108 households.
For each household i, the value of si (which was automatically computed by SPSS) corresponds to the sum of eleven asset scores sij, each of which was computed thus:
sij = Score of household i on asset j =
= (value of asset variable j for household i – mean of asset variable j) / standard deviation of asset variable j * weight (score coefficient)
The same is to say, the asset index was constructed as a weighted sum of the standardised values assumed by the asset variables in each household, with the weights corresponding to the score coefficients obtained through principal component analysis. It has zero mean, which implies that the fact that a household exhibits an asset index score of 0 does not mean that the household in question does not own any of these items, but rather that its endowment in terms of these assets (thus weighted) is exactly average.
Using this procedure as the basis for subsequent inferences on household wealth/permanent income relies on the assumption, which has been validated in other contexts, that “household long-run wealth explains the maximum variance (and covariance) in the asset variables” (Filmer and Pritchett 2001:117) and that, as a consequence, the first principal component identified through PCA of a set of durable household asset data constitutes a valid proxy for household long-run wealth182.
eleventh variable included in the principal components analysis was the number of cows/owen owned by the household, which is a scale variable (i.e. it has a meaningful metric).
182 This assumption is especially appropriate when the asset variables have a categorical character (i.e. absence/presence, or 0/1) and range from common (almost all the households own them) to rare (few households own them) without there being substitute assets, which implies that the share of households that own each of these assets is monotonically increasing. The latter fact, along with the categorical character of the variables, causes the means and the standard deviations of the asset variables to exhibit an especially strong relationship, thus ensuring that the asset index thus computed constitutes a good indicator of long-run wealth.
328 The PCA exercise undertaken in this case yielded the SPSS output that is presented in the following pages (Tables A4.1-A4.4).
Initial Eigenvalues
Total % of Variance Cumulative %
3,779 34,357 34,357
1,661 15,096 49,453
1,064 9,674 59,127
,930 8,459 67,586
,787 7,154 74,740
,657 5,971 80,711
,549 4,991 85,702
,468 4,257 89,959
,449 4,078 94,037
,362 3,292 97,329
,294 2,671 100,000
Table A4.1: Total Variance Explained
Kaiser-Meyer-Olkin Measure of Sampling Adequacy. ,805 Bartlett's Test of
Sphericity
Approx. Chi-Square 331,462
df 55
Sig. ,000
Table A4.2: KMO and Bartlett's Test
329 Component
1
Number of cows ,739
Electricity ,480
Radio ,603
TV ,759
Cell phone ,445
Bicycle ,552
Fridge ,731
Motorcycle ,733
Car ,587
Gas or electric stove
-,260
Zinc rooftop ,276
Table A4.3: Component Matrix
Component 1
Number of cows ,196
Electricity ,127
Radio ,160
TV ,201
Cell phone ,118
Bicycle ,146
Fridge ,193
Motorcycle ,194
Car ,155
Gas or electric stove -,069
Zinc rooftop ,073
Table A4.4: Component Score Coefficient Matrix
330 The Tables presented in the previous pages provide us with a substantial amount of information. The key aspects that should be noticed are the following:
• Table A4.1 indicates that the first principal component obtained through PCA accounts for 34.4% of the variance in the data set. This implies that factors other than component 1 account for the remaining 65.6%, which might be deemed excessive. However, one should bear in mind that the implicit assumption in this exercise is not that household wealth accounts for all the variance in asset ownership, but rather that it is the single factor accounting for the largest share of the variance. Moreover, 34% may in fact be considered a relatively high figure, if we take into account that in the example provided in Filmer and Pritchett (2001) (where the use of this type of asset index is validated based on expenditure data), the first principal component only accounts for 26% of the variance.
• Table A4.2 provides the results of two standard tests of the quality of the data in the context of principal component analysis. The Kaiser-Meyer-Olkin Measure of Sampling Adequacy indicates how large the partial correlations among the variables are. It varies between 0 and 1, with a value of 0.805 (as in this case) usually being considered as indicative of “strong” common variance (>0.9 would be “very strong”), which is a condition for PCA to be appropriate. In turn, Bartlett’s Test of Sphericity is based on the Chi-Square statistic, and the resulting p-value indicates the probability that the variables are in fact uncorrelated (which, again, would render the PCA model inappropriate). In this case, the p-value is 0.000, which means that we may feel secure in the use of PCA.
• Table A4.3 shows the component loadings, i.e. the correlation coefficients between the variables and the component. When squared, these represent the extent to which the variation in a given variable is accounted for by the component:
component loadings greater than 0.7 are indicative of variables for which half or more (>0.49) of the variance is explained by the component. In this particular case, household ownership of motorcycle, TV, fridge and the number of cattle heads all have loadings greater than 0.7, which indicates that these variables are highly correlated with the underlying component 1 (“long-run wealth”). By contrast, the
331 variation in the ownership of zinc rooftops and gas/electric stoves seems to be mostly explained by factors other than “wealth”183.
• Finally, Table A4.4 contains the component score coefficients, which consist of a linear transformation of the component loadings (through division by the eigenvalue). These correspond to the weightings used to compute the household scores, in such a way as to ensure that the latter have zero mean and a standard deviation of one. They show that ownership of a fridge, TV, motorcycle or a relatively large number of cows significantly increases the (total) household score, whereas ownership of any of the other assets does not contribute as much (i.e.
those assets are weaker indicators of household wealth). In the case of gas/electric stoves, the contribution is in fact negative, reflecting the fact that households that own this item are in fact likely to be poorer in all other respects.
As a final validation exercise prior to the subsequent use of the vector of household scores as a proxy for household wealth, the following Table (A4.5) follows the procedure adopted in Filmer and Pritchett (2001), whereby the average value of each asset variable is shown for four quartile groups (ranked by asset index score).
183 This is easily explained in the case of gas/electric stoves, which are owned by as many as 54% of the households in Caiomete compared to a mere 6% in Braima Sori, despite the fact that the latter village is much wealthier. Reportedly, the Manjaco women of Caiomete – a relatively dense village – significantly appreciate the enhanced privacy afforded by the ability to cook indoors using a portable gas stove, and for that reason households are more likely to buy that item when they can afford it. In Braima Sori, by contrast, this does not seem to be considered an important benefit. Thus, as a consequence of the fact that the socially-constructed meaning and ‘utility’ of this household item differs between the two villages, most of the variation in reported ownership in the overall sample is not due to the underlying first component “long-run wealth” but rather to something else:
presumably, the difference in the social meaning of the item itself. A similar effect occurs with respect to zinc rooftops, which are also very weakly correlated with the first principal component:
the vast majority of the households in both villages report owning this item, and we find that, generally speaking, those households that do not own them are not the poorest ones (with respect to the other items), but rather households that for some other reason do not exhibit as strong a preference for zinc rooftops.
332 Means
Score
coefficient Total
1st Quartile
2nd Quartile
3rd Quartile
4th Quartile Asset index 0.0 -0,9104 -0,4627 -0,0352 1,4083
Number of cows 0,196 7.2 0.7 1.5 4 22.6
Electricity 0,127 18% 0% 4% 15% 52%
Radio 0,160 58% 4% 41% 89% 100%
TV 0,201 21% 0% 0 4% 81%
Cell phone 0,118 79% 37% 85% 96% 96%
Bycicle 0,146 52% 4% 33% 85% 85%
Fridge 0,193 9% 0% 0% 0% 37%
Motorcycle 0,194 11% 0% 0% 0% 44%
Car 0,155 7% 0% 0% 4% 26%
Zinc rooftop 0,073 93% 78% 100% 93% 100%
Gas or electric
stove -0,069 38% 56% 41% 37% 19%
Table A4.5: Mean asset variables by quartile (in the sample of 108 households, ranked according to asset index)
The Table above shows, for example, that none of the households in the 1st and 2nd quartiles own TVs, but 4% of the households in the 3rd quartile and 81% of those in the 4th quartile do. On average, households in the 1st quartile own less than one cow, but those in the 2nd, 3rd and 4th quartiles own, respectively, 1.5, 4 and 23 cows. Remarkably, the level of asset ownership increases monotonously by quartile in the case of all the asset variables except for zinc rooftop and gas/electric stove (which is accounted for by the reasons explained in footnote 183, above).
This provides additional evidence that the use of this procedure is appropriate, in the sense that the asset index thus computed constitutes a valid and consistent indicator of household wealth in the context of this research. The high level of association between the asset variables (shown by the high level of the KMO Measure in Table A1.2 and by the fact that the first principal component accounts for 34% of the variance) ensures that the PCA yields statistically significant weightings in the construction of the asset index, while the low absolute values of the score coefficients of such assets as gas/electric stove or zinc rooftop makes allowance for differences in the non-wealth social determinants of asset ownership in the two villages, without requiring us to compute two different asset indices. As a consequence of the latter, we may undertake inter-village comparisons of the level of wealth (Table A4.6), which indeed confirm that Braima Sori (with a mean asset index score of 0.87) is a much wealthier village than Caiomete (-0.43).
333 Quartile Asset Index
Total 1st
Quartile
2nd Quartile
3rd Quartile
4th Quartile Village Caiomete
(mean = -0.43)
26 23 15 8 72
Braima Sori (mean = 0.87)
1 4 12 19 36
Total (mean = 0) 27 27 27 27 108
Table A4.6: Number of households in each quartile (ranked by asset index score) in Caiomete and Braima Sori
334
Bibliography
Abranches, M. (2007) Pertenỗas Fechadas em Espaỗos Abertos: Estratộgias de (Re)Construỗóo Identitỏria de Mulheres Muỗulmanas. Colecỗóo Teses, 13, Lisboa:
Observatúrio da Imigraỗóo - Alto Comissariado para a Imigraỗóo e Minorias ẫtnicas.
Accessed online on 15 March 2011 at:
http://www.oi.acidi.gov.pt/docs/Col_Teses/13_MA.pdf
Abreu, A. (2009) As Migraỗừes Internacionais e o Desenvolvimento dos Paớses de Origam:
Impactos e Polớticas, Colecỗóo Teses, 28, Lisboa: Observatúrio da Imigraỗóo – Alto Comissariado para a Imigraỗóo e Diỏlogo Intercultural. Accessed online on 13 January 2010 at: http://www.oi.acidi.gov.pt/docs/Colec_Teses/tese_28.pdf
Abreu, A. (forthcoming) “The New Economics of Labor Migration: Beware of Neoclassicals Bearing Gifts”, Forum for Social Economics, published online first at:
https://springerlink3.metapress.com/content/q6r14x15657l1v1k/resource-
secured/?target=fulltext.pdf&sid=goyrm255jti13wv3owossubs&sh=www.springerlink.com/
Acosta, P., Fajnzylber, P. and Lopez, H. (2007) “The Impact of Remittances on Poverty and Human Capital: Evidence from Latin America”, Policy Research Working Paper Series, 4247, The World Bank.
Acosta, P., Lartey, E. and Mandelman, F. (2007b) “Remittances and the Dutch Disease”, Federal Bank of Atlanta Working Paper, 08/2007. Accessed online on 19 July 2011 at:
http://ideas.repec.org/p/fip/fedawp/2007-08.html
Agence Européenne pour le Développement et la Santé (AEDES) (2009) Rapport de Mission d’Assistance Technique: Analyse des Données de l’Enquête Agricole, Partie 2, Projet SISA – Guinée-Bissau GUB/FOOD/2005/17683, Bruxelles : AEDES. Accessed online on 9 May 2011 at: http://www.stat-guinebissau.com/publicacao/Relatorio_analis_inquerito_agricola.pdf Agnew, J. and Cox, K. (1980) “In-Migration Field Dynamics: A Conceptualization and Empirical Analysis”, Geografiska Annaler. Series B, Human Geography, 62(2):69-80.
335 Anderson, P. (1976) Considerations on Western Marxism. London: New Left Books.
Arrow, K. (1962) “The Economic Implications of Learning by Doing”, The Review of Economic Studies, 29(3):155-173.
Ball, C., Cruz-Zuniga, M., Lopez, C. and Reyes, J. (2010) “Remittances, Inflation and Exchange Rate Regimes in Small Open Economies”, University of Cincinnati Economics Working Papers, 03/2008. Accessed online on 19 July 2011 at:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1119318
Banaji, J. (1977) “Modes of Production in a Materialist Conception of History”, Capital and Class, 3:1-44.
Banaji, J. (2010) Theory as History: Essays on Modes of Production and Exploitation. Leiden:
Brill.
Banque Centrale des États de l’Afrique de l’Ouest (BCEAO) (2009) Balance des Paiements 2009 - Base de Données. Bissau: BCEAO.
Baran, P. (1957) The Political Economy of Growth, New York: Monthly Review Press.
Barry, B., Creppy, E. and Wodon, Q. (2007) Cashew Prodution, Taxation, and Poverty in Guinea-Bissau, Munich Personal RePEc Archive paper. Accessed online on 12 April 2011 at:
mpra.ub.uni-muenchen.de/11181/1/MPRA_paper_11181.pdf
Bedford, R. (1981) “Labour Circulation and the Articulation of Modes of Production in the Third World: A Preliminary Statement”, Paper Presented at the 11th New Zealand Geography Conference, mimeo.
Bender, D. and Lửwenstein, W. (2005) “Two-Gap Models: Post-Keynesian Death and Neoclassical Rebirth”, IEE Working Papers, 180, Institute of Development Research and Development Policy, Ruhr University Bochum. Accessed online on 14 July 2011 at:
http://economics.rub.de/fileadmin/IEE/pubs/2010-2020/Bender_Loewenstein_2005.pdf Berger, S. and Piore, M. (1980) Dualism and Discontinuity in Industrial Societies, Cambridge:
Cambridge University Press.
Bernstein, H. (2003) “Farewells to the Peasantry”, Transformation, 52:1-19.
336 Bernstein, H. (2010) Class Dynamics of Agrarian Change, Halifax and Winnipeg: Fernwood Publishing and Sterling, VA: Kumarian Press.
Bernstein, H. (2010b) “Can Modernity Accommodate African ‘Peasants’?”. Paper presented at the 7th Congress of African Studies, CIEA no. 7: Desarrollo Rural. Accessed online on 22 July 2011 at:
http://repositorio-
iul.iscte.pt/bitstream/10071/2337/1/CIEA7_24_BERNSTEIN_Can%20Modernity%20Accom modate%20African%20%E2%80%98Peasants%E2%80%99.pdf
Bernstein, H. and Byres, T. (2001) “From Peasant Studies to Agrarian Change”, Journal of Agrarian Change, 1(1):1-56.
Bhagwati, J. (1976) (ed.) Taxing the Brain Drain, 2 volumes, Amsterdam: North-Holland.
Bhargava, A. and Docquier, F. (2008) “HIV Pandemic, Medical Brain Drain, and Economic Development in Sub-Saharan Africa” The World Bank Economic Review, 22(2):345–366.
Bilsborrow, R., Oberai, A. and Standing, G. (1984) Migration Surveys in Low-Income Countries: Guidelines for Survey and Questionnaire Design, Beckenham: International Labour Organisation.
Bjuggren, P.-O., Dzansi, J. and Shukur, G. (2010) “Remittances and Investment”, CESIS Electronic Working Paper Series, 216. Accessed online on 25 July 2011 at:
http://scripts.abe.kth.se/cesis/documents/CESISWP216.pdf
Boyd, M. (1989) “Family and Personal Networks in International Migration: Recent Developments and New Agendas”, International Migration Review, 23(3), Special Silver Anniversary Issue: International Migration – An Assessment for the 90’s: 638-670.
Brenner, R. (1976) “Agrarian Class Structure and Economic Development in Pre-Industrial Europe”, Past & Present, 70:30-75.
Brenner, R. (1977) “The Origins of Capitalist Development: A Critique of Neo-Smithian Marxism”, New Left Review, 104:25-92.
Brenner, R. (1982) “The Agrarian Roots of European Capitalism”, Past & Present, 97:16-113.
Brito, E. (1957) “Notas sobre a Vida Religiosa dos Fulas e Mandingas”, Boletim Cultural da Guinộ Portuguesa, Vol. XII, nôº 46, p. 148-189.
337 Bryceson, D. (2002) “The Scramble in Africa: Reorienting Rural Livelihoods”, World Development, 30(5): 725-39.
Bryceson, D. (2002b) “Multiplex Livelihoods in Rural Africa: Recasting the Terms and Conditions of Gainful Employment”, Journal of Modern African Studies, 40(1): 1-28.
Bryceson, D. and Jamal, V. (1997) Farewell to Farms: De-Agrarianisation and Employment in Africa, Ashgate: Aldershot.
Byres, T. (1996) Capitalism from Above and Capitalism from Below: An Essay in Comparative Political Economy, London: Palgrave Macmillan.
Cabral, A. (1979) Unity and Struggle: Speeches and Writings, New York: Monthly Review Press.
Cabral, A. (2008) Documentário, Lisboa: Cotovia.
Callinicos, A. (2004) Making History: Agency, Structure and Change in Social Theory. Leiden:
Brill.
Cardoso, C. (1996) “Classe polớtica e transiỗóo democrỏtica na Guinộ-Bissau”. In Koudawo, F.
e Mendy, P. (coords.) Pluralismo Polớtico na Guinộ-Bissau: uma Transiỗóo em Curso (pp.
135-164). Bissau: Instituto Nacional de Estudos e Pesquisa.
Cardoso, L. (2002) “Subsídios para o Esudos dos Movimentos Migratórios na Guiné-Bissau”, Soronda – Revista de Estudos Guineenses, nova série (3):29-50.
Cardoso, L. (2003) “Sistemas de Heranỗa entre os Papộis, Manjacos e Mancanhas”, Soronda – Revista de Estudos Guineenses, nova série (6):147-178.
Carling, J. (2005) “Migrant Remittances and Development Cooperation”, PRIO Report I/2005, Oslo: International Peace Research institute. Accessed online on 20 July 2011 at:
http://www.prio.no/files/file46220_carling_2005_migrant_remittances_and_development _cooperation.pdf
Carling, J. and Akesson, L. (2004) “Mobility at the Heart of a Nation: Patterns and Meanings of Cape Verdean Migration”, International Migration Review, 47(3):123-155.
Carreira, A. and Meireles, A. (1959) “Notas sobre os Movimentos Migratúrios da Populaỗóo Natural da Guiné Portuguesa”, Boletim Cultural da Guiné Portuguesa, XIV(53):7-19.
Accessed online on 19 August 2010 at:
338 http://memoria-
africa.ua.pt/DesktopModules/MABDImg/ShowImage.aspx?q=/BCGP/BCPG-N053&p=1 Carreiro, M. (2007) “Dinõmicas Transnacionais Protagonizadas por Associaỗừes de Migrantes Guineenses em Portugal”, CIES e-Working Paper, 26, Lisboa: Lisboa: Centro de Investigaỗóo e Estudos de Sociologia – Instituto Superior das Ciờncias do Trabalho e da Empresa. Accessed online on 20 April 2011 at:
http://www.cies.iscte.pt/destaques/documents/CIES-WP26_Carreiro_.pdf
Carreiro, M. (2011) Arquitectos de um Espaỗo Transnacional Lusúfono: A Diỏspora Guineense em Portugal, Lisboa: Fundaỗao Portugal-África.
Carvalho, C. (2003) “A Revitalizaỗóo do Poder Tradicional e os Regulados Manjaco da Guinộ- Bissau”, Soronda – Revista de Estudos Guineenses, nova série(7):7-44.
Cassarino, J.-P. (2004) “Theorising Return Migration: The Conceptual Approach to Return Migrants Revisited”, International Journal on Multicultural Societies, 6(2):253-279.
Castanheira, J. (1995) Quem Matou Amílcar Cabral?, Lisboa: Relógio d’Água.
Castles, S. and Kosack, G. (1973) “The Function of Labour Immigration in Western European Capitalism”, New Left Review, 73(1):3-21.
Cesare, F. (1974) “Expectations and Reality: A Case-Study of Return Migration from the United States to Southern Italy”, International Migration Review, 8(2): 245-262.
Chabal, P. (1983) Amilcar Cabral: Revolutionary Leadership and People’s War, Cambridge:
Cambridge University Press.
Chami, R., Barajas, A., Cosimano, T., Fullenkamp, C., Gapen, M. and Montiel, P. (2003)
“Macroeconomic Consequences of Remittances”, IMF Occasional Paper, 259, International Monetary Fund. Accessed online on 25 July 2011 at:
http://www.imf.org/external/pubs/ft/op/259/op259.pdf
Chayanov, A. (1966) [1925] The Theory of the Peasant Economy, Homewood, IL:Irwin for the American Economic Association.
Chenery, H. and Bruno, M. (1962) “Development Alternatives in an Open Economy: The Case of Israel”, The Economic Journal, 72(285):79-103.