bs_bs_banner bs_bs_banner Review of Review of Income and Wealth Wealth 2015 Series 63, Number 4, December 2017 DOI: 10.1111/roiw.12212 DOI: 10.1111/roiw.12212 WHEN THE CENTRE CANNOT HOLD: PATTERNS OF POLARIZATION IN NIGERIA by F. Clementi Università degli Studi di Macerata, Macerata, Italy A. L. Dabalen and V. Molini* World Bank, Washington DC, USA F. Schettino Seconda Università degli Studi di Napoli, Napoli, Italy This paper advances the hypothesis that Nigeria is going through a process of economic polarization. The notion of polarization is concerned with the disappearance or non-consolidation of the middle class, which occurs when there is a tendency to concentrate in the tails, rather than the middle, of the income/consumption distribution. This paper uses newly available data and the relative distribution methodology (Handcock and Morris, 1998, 1999) to present new results on polarization. The findings confirm the sharp increase of polarization. Compared to 2003, the distribution of consumption has become more concentrated in upper and lower deciles in 2013, while the middle deciles have thinned. A between-group analysis shows the emergence of a macro-regional gap: while the South-South and South-West regions contribute mainly to polarization in the upper tail, households in the North East and North West zones—the conflict-stricken areas—are more likely to fall in the lower national deciles. JEL Codes: C14, D31, D63, I32, Q34 Keywords: consumption expenditure, Nigeria, polarization, poverty and inequality, relative distribution 1. Introduction Despite a stable and sustained real growth of nearly 6 percent per year for almost a decade from 2003, reduction in official poverty rate in Nigeria has not been up to general expectations. While poverty seems to have declined faster in the coastal South and around the Federal Capital, Abuja, a large belt of North-eastern states appear to have experienced a significant increase in poverty. The lack of a faster reduction in poverty despite a significant growth in GDP may be due to an increase in inequality (World Bank, 2013). The latter is, however, just one aspect of the problem. A complementary hypothesis is that Nigeria is also experiencing increasing income Note: The authors acknowledge financial support from the World Bank. We thank Federica Alfani (FAO—Food and Agriculture Organization of the United Nations) and Rose Mungai (World Bank— Poverty & Equity Global Practice) for excellent assistance with data preparation. We also thank Christoph Lakner (World Bank—Development Research Group) for comments on an earlier version of the manuscript as well as the editor and two anonymous reviewers, whose insightful suggestions helped to improve this article greatly. Of course, we are the sole responsible for all possible errors the paper may contain. *Correspondence to: Vasco Molini, Poverty & Equity Global Practice, World Bank, 1818 H Street NW, Washington DC, 20433 USA (vmolini@worldbank.org). C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth This is an open access article distributed under the terms of the Creative Commons Attribution IGO License https://creativecommons.org/ licenses/by/3.0/igo/legalcode which permits unrestricted use, distribution, and reproduction in any medium, provided that the original work is properly cited. In any reproduction of this article there should not be any suggestion that UNU or the article endorse any specific organization or products. The use of the UNU logo is not permitted. This notice should be preserved along with the article's URL. 608 1 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 polarization. Whereas inequality is the overall dispersion of the distribution, polar- ization is the combination of divergence from global and convergence on local mean incomes. When societies experience income polarization, people cluster around group means and tend to be far from the mean/median of the overall distribution. An important aspect of the income polarization analysis is that it is concerned with the disappearance or non-consolidation—in the case of Nigeria—of the middle class. This occurs when in a society there is a tendency to concentrate in the tails, rather than the middle, of the income distribution. Within each group there is increasing identification, which means income homogeneity and often declining income inequality, while between the two groups, we have “increasing alienation” instead (Duclos et al., 2004). The combined effect of the forces of alienation and identification between two significantly sized groups would tend to lead to effective opposition, a situation that might give rise to social conflicts and tensions (Esteban and Ray, 1999, 2008, 2011). This problem is made worse when the group at the top of the distribution has disproportionately more voice and power while the other group, comprising those at the bottom, are voiceless in matters that affect their welfare and the society at large. By contrast, a well-off middle class is believed to be important to every society because it is associated with many desirable and stabi- lizing outcomes such as positive contribution to economic growth, democracy, and social and political stability (Easterly, 2001; Pressman, 2007; Birdsall, 2010). Nigeria represents an interesting case for an analysis of polarization. As noted earlier, GDP and per capita income have steadily grown in the last decade. After the recent GDP re-basing, Nigeria will become the biggest African economy and yet there are clear signs of a limited consolidation of a national middle class, by which we mean a lack of stable, cohesive and secure middle. Moreover, the country is increasingly affected by sub-regional conflicts driven to a large extent by disaf- fected (alienated) groups. Studies on polarization in Nigeria are few and may have been hampered by data problems. One of the earliest studies is that of Aigbokhan (2000), who showed that polarization in Nigeria increased between 1985 and 1992, but declined in rural areas. The study also found a higher but declining degree of polarization in rural areas during the 1990s. Araar (2008) study uses the 2003/2004 round of the National Living Standard Survey (NLSS) to identify the main drivers behind polarization in the 1990s. The study concludes that the population, based on geo-ecological zones, is spatially polarized. Furthermore, employment income and non-farm business income are the two income sources that significantly contribute to total polarization. Awoyemi and Araar (2009) use the same set of data and a decomposition of the Duclos-Esteban- Ray (DER) index of polarization (Duclos et al., 2004) to conclude that, in addition to the geo-ecological zones, differences in education and occupation drive polariza- tion. Their study also shows that the ongoing polarization they find favors a prevalence of the identification component rather than alienation. The study by Awoyemi et al. (2010) also uses DER and Foster-Wolfson (2010) indices to look at the polarization dynamics over the longer time span 1996–2004, and show that polarization has declined from 0.30 to 0.25 using Foster-Wolfson index and from 0.44 to 0.38 using DER index. They also show that in the southern macro areas (South-East and South-West) indexes do not vary significantly. Ogunyemi and Oni (2011) and Ogunyemi et al. (2011) calculated the same indices on households in © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 2 609 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 rural areas only from 1980 to 2004, and find a similarly decreasing trend in polarization. More recent studies like Ogunyemi (2013), which compares the 2003/ 2004 round of the NLSS and the 2009/2010 round of the Harmonized National Living Standard Survey (HNLSS), indicate only a slight tendency towards increas- ing polarization, and mostly a general absence of movement. The present paper improves upon these studies in several aspects. First, rather than just computing and comparing polarization indices, we use a non-parametric framework (the “relative distribution” introduced by Handcock and Morris, 1998, 1999) and compare income throughout the entire income range. The relative distribution analysis requires at least two comparable survey rounds in order to investigate changes along the entire distribution. Since the lack of comparable surveys has limited the scope of previous work, the second way in which our study differs from the past studies on polarization in Nigeria is to construct two fully comparable distributions using recent innovations in the use of survey-to-survey imputation techniques, discussed in detail under data section below. Finally, the flexibility of the relative distribution tool allows an accurate analysis at macro-regional level too. Differently from previous contributions, another goal of this paper is to document not just national, but also sub-national patterns of polar- ization. Nigeria is highly heterogeneous, so that drivers of polarization can indeed differ across macro regions. It is also worth mentioning that this focus on macro- regions has the potential to inform future research on the link between polarization and regional conflicts. Our findings suggest a sharp increase in polarization between 2003 and 2013. We find that distribution of consumption is concentrated in the lower and upper deciles, while the middle deciles are shrinking. Moreover, a between-group analysis shows that while the South-South and South-West regions contribute mainly to polarization in the upper tail of the national consumption distribution, the North- East and North-West are more likely to fall in the lower deciles. The rest of the paper is organized as follows. Section 2 reviews the approaches to measuring economic polarization. Section 3 outlines the distinctive features of the relative distribution method. Section 4 presents the data and details the main findings of the study. Section 5 concludes. 2. Some Background on the Income Polarization Literature Over the last two decades, the issue of polarization has come to be assigned increasing importance in the analysis of income distribution. Notwithstanding the pains the polarization literature has suffered to distinguish itself from pure inequality measurement–see e.g. Levy and Murnane (1992); Esteban and Ray (1994); Wolfson (1994, 1997), and Foster and Wolfson (2010), it now seems to be fairly widely accepted that polarization is a distinct concept from inequality. Broadly speaking, the notion of polarization is concerned with the disappear- ance of the middle class, which occurs when there is a tendency to concentrate in the tails—rather than the middle—of the income distribution. One of the main reasons for looking at income polarization this way, which is usually referred to as “bi- polarization”, is that a well-off middle class is important to every society because it contributes significantly to economic growth, as well as to social and political stability (e.g. Easterly, 2001, and Pressman, 2007). In contrast, a society with high C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 3 610 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 degree of income polarization may give rise to social conflicts and tensions. There- fore, in order for such risks to be minimized, it is necessary to monitor the economic evolution of the society using indices that look at the dispersion of the income distribution from the middle toward either or both of the two tails. Measures of income polarization that correspond to this case have been proposed in the litera- ture by Wolfson (1994, 1997); Wang and Tsui (2000); Chakravarty and Majumder (2001); Rodríguez and Salas (2003); Chakravarty et al. (2007); Silber et al. (2007); Chakravarty (2009); Chakravarty and D’Ambrosio (2010); Foster and Wolfson (2010); Lasso De La Vega et al. (2010), and others. A more general notion of income polarization, which was originally proposed by Esteban and Ray (1994), regards the latter as “clustering” of a population around two or more poles of the distribution, irrespective of where they are located along the income scale. The notion of income polarization in a multi-group context is an attempt at capturing the degree of potential conflict inherent in a given distribution (see Esteban and Ray, 1999, 2008, 2011). The idea is to consider society as an amalgamation of groups, where the individuals in a group share similar attributes with the other members (i.e. have a mutual sense of “identifica- tion”) but in terms of the same attributes they are different from the members of the other groups (i.e. have a feeling of “alienation”). Political or social conflict is therefore more likely the more homogeneous and separate the groups are, that is when the within-group income distribution is more clustered around its local mean and the between-group income distance is longer. In addition to Esteban and Ray (1994), indices regarding the concept of income polarization as conflict among groups have been investigated, among others, by Gradín (2000); Milanovic (2000); D’Ambrosio (2001); Zhang and Kanbur (2001); Reynal-Querol (2002); Duclos et al. (2004); Lasso De La Vega and Urrutia (2006); Esteban et al. (2007); Gigliarano and Mosler (2009), and Poggi and Silber (2010). Much of the literature so far considered has analyzed summary measures of income polarization. Another strand uses kernel density estimation and mixture models in order to describe changes in polarization patterns over time, not just of personal incomes (as in Jenkins, 1995, 1996; Pittau and Zelli, 2001, 2004, 2006, and Conti et al., 2006) but also of the cross-country distribution of per capita income (see Quah, 1996a, 1996b, 1997; Bianchi, 1997; Jones, 1997; Paap and Van Dijk, 1998; Johnson, 2000; Holzmann et al., 2007; Henderson et al., 2008; Pittau et al., 2010; Anderson et al., 2012, and others). The analysis of the shape of the income distribution provides indeed a picture from which at least three important distri- butional features can be observed simultaneously (Cowell et al., 1996): income levels and changes in the location of the distribution as a whole; income inequality and changes in the spread of the distribution; clumping and polarization as well as changes in patterns of clustering at different modes. Finally, a rather recent (yet non-parametric) approach that combines the strengths of summary polarization indices with the details of distributional change offered by the kernel density estimates—the so-called “relative distribution”—has been employed by Alderson et al. (2005); Massari (2009); Massari et al. (2009a, 2009b); Alderson and Doran (2011, 2013), and Borraz et al. (2013) to assess the evolution of the middle class and the degree of household income polarization in a number of middle- and high-income countries in the world. © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 4 611 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 3. The Method Based on the Relative Distribution 3.1. The Relative Distribution: Basic Concepts In the current application, the relative distribution approach has some important advantages over the other mentioned methods of investigating income polarization. First, it readily lends itself to simple and informative graphical displays of relative data that reveal precisely where and by how much an income distribution changed over time. Second, by providing the potential for decomposition into location and shape components, it allows one to examine several hypotheses regarding the origins of distributional change—such as whether the change consists of an equal absolute subtraction or addition to all incomes that moves the overall distribution either to the left or to the right (while leaving the shape unaltered) or of shape modifications which, by definition, are independent of location shifts.1 Lastly, it allows us to quantify the degree of polarization due to changes in distributional shape only (i.e. net of location shifts), thus enabling one to isolate aspects of inter-distributional inequality that are often hidden when also changes in location are examined. Basically, the relative distribution method can be applied whenever the distri- bution of some quantity across two populations is to be compared, either cross- sectionally or over time.2 To proceed, it is necessary to single out one of the two populations, refer to it as the “comparison” population, and refer to the other as the “reference” population. More formally, let Y0 be the income variable for the reference population and Y the income variable for the comparison population. The relative distribution of Y to Y0 is defined as the distribution of the random variable: (1) R F0 Y , which is obtained from Y by transforming it by the cumulative distribution function of Y0, F0. As a random variable, R is continuous on the outcome space [0,1], and its realizations, r, are referred to as “relative data”. Intuitively, the relative data can be interpreted as the set of positions that the income observations of the comparison population would have if they were located in the income distribution of the reference population. The probability density function of R, which is called the “relative density”, can be obtained as the ratio of the density of the comparison population to the density of the reference population evaluated at the relative data r: f F01 r f yr (2) g r , 0 d r d 1, yr t 0, f0 F r 0 1 f0 yr where f(·) and f0(·) denote the density functions of Y and Y0, respectively, and yr F01 r is the quantile function of Y0. The relative density has a simple interpretation, as it describes where households at various quantiles in the 1 Of course, both the location and shape effects—named respectively as “growth” and “inequality” (or “distributional”) effect (Kakwani, 1993; Bourguignon, 2003, 2004)—may also concur together in producing the distributional change. 2 Here we limit ourselves to illustrating the basic concepts behind the use of the relative distribution method. Interested readers are referred to Handcock and Morris (1998, 1999; but see also Hao and Naiman, 2010, ch. 5) for a more detailed explication and a discussion of the relationship to alternative econometric methods for measuring distributional differences. A method very similar in spirit to the relative distribution has recently been developed by Silber et al. (2014). C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 5 612 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 comparison distribution are concentrated in terms of the quantiles of the reference distribution. As for any density function, it integrates to 1 over the unit interval, and the area under the curve between two values r1 and r2 is the proportion of the comparison population whose income values lie between the r1th and r2th quantiles of the reference population. When the relative density function shows values near to 1, it means that the two populations have a similar density at the rth quantile of the reference popula- tion, and thus R has a uniform distribution in the interval [0,1]. A relative density greater than 1 means that the comparison population has more density than the reference population at the rth quantile of the latter. Finally, a relative density function less than 1 indicates the opposite. In this way one can distinguish between growth, stability or decline at specific points of the income distribution. 3.2. The Location/Shape Decomposition of the Relative Distribution As we have said before, one of the major advantages of this method is the possibility to decompose the relative distribution into changes in location, usually associated with changes in the median (or mean) of the income distribution, and changes in shape (including differences in variance, asymmetry and/or other dis- tributional characteristics) that could be linked with several factors like, for instance, polarization. Formally, the decomposition can be written as: f yr f0 L yr f yr (3) g r u , f0 yr f0 yr f0 L yr Overall relative Density ratio for Density ratio for density the location effect the shape effect where f0L(yr) = f0(yr + ρ) is a density function adjusted by an additive shift with the same shape as the reference distribution but with the median of the comparison one.3 The value ρ is the difference between the medians of the comparison and reference distributions. If the latter two distributions have the same median, the density ratio for location differences is uniform in [0,1]. Conversely, if the two distributions have different median, the “location effect” is increasing (decreasing) in r if the comparison median is higher (lower) than the reference one. The second term, which is the “shape effect”, represents the relative density net of the location effect and is useful to isolate movements (re-distribution) occurring between the reference and comparison populations. For instance, we could observe a shape effect function with some sort of (inverse) U-shaped pattern if the comparison distribution is relatively (less) more spread around the median than the location-adjusted one. Thus, it is possible to determine whether there is polarization of the income distribution (increases in both tails), “downgrading” (increases in the lower tail), “upgrading” (increases in the upper tail) or convergence of incomes towards the median (decreases in both tails). 3 Median adjustment is preferred here to mean adjustment because of the well-known drawbacks of the mean when distributions are skewed. A multiplicative median shift can also be applied. However, the multiplicative shift has the drawback of affecting the shape of the distribution. Indeed, the equi- proportionate income changes increase the variance and the rightward shift of the distribution is accompanied by a flattening (or shrinking) of its shape (see e.g. Jenkins and Van Kerm, 2005). © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 6 613 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 3.3. Relative Polarization Indices The relative distribution approach also includes a median relative polarization index (MRP), which is based on changes in the shape of the income distribution to account for polarization. This index is normalized so that it varies between −1 and 1, with 0 representing no change in the income distribution relative to the reference year. Positive values represent more polarization—i.e. increases in the tails of the distribution—and negative values represent less polarization—i.e. convergence towards the center of the distribution. The MRP index for the comparison popu- lation can be estimated as (Morris et al., 1994, p. 217): (4) MRP 4 n ¦ ri  n i1 1 2  1, where ri is the proportion of the median-adjusted reference incomes that are less than the ith income from the comparison sample, for i = 1, . . . , n, and n is the sample size of the comparison population. The MRP index can be additively decomposed into the contributions to overall polarization made by the lower and upper halves of the median-adjusted relative distribution, enabling one to distinguish downgrading from upgrading. In terms of data, the lower relative polarization index (LRP) and the upper relative polarization index (URP) can be calculated as follows: (5) LRP n¬« i¦ 8 ªn 2 1 1 2  ri º  1, » ¼ (6) URP 8ª n « n ¬i n 2 1 ¦ ri  º 1 2 » ¼  1, with MRP = 1⁄2(LRP + URP). As the MRP, LRP and URP range from −1 to 1, and equal 0 when there is no change. 3.4. Adjustment for Covariates Similarly to what is observed for location and shape decomposition, it is possible to adjust the relative distribution for changes in the distribution of covariates measured on the households, which often vary systematically by popu- lation. The covariate adjustment technique can be used to separate the impacts of changes in population composition from changes in the covariate-response rela- tionship.4 This decomposition according to covariates draws on the definition of a counter-factual distribution for the response variable in the reference population 4 Recently, there have been several papers that have studied decomposition methods to explain changes in the unconditional distribution of an outcome variable due to either changes in the distri- bution of the covariates, or changes in the conditional distribution of the outcome given covariates, or both—see for instance the extensive survey by Fortin et al. (2011) on the wage decomposition literature. Benefits and drawbacks of some of these methods, and how they are often largely subsumed by the relative distribution framework, are reviewed in Handcock and Morris (1999, ch. 2). C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 7 614 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 that is composition-adjusted to have the same distribution of the covariates as the comparison population. Assume for simplicity that the covariate Z is categorical.5 Let ^S k `k 1 and 0 K ^S k `k 1 , where K is the number of categories of the covariate, denote the prob- K ability mass functions of Z for the reference and comparison populations, i.e. their composition according to the covariate. For conditional comparisons of the response variable Y across the two populations one can consider the density of Y0 given that Z0 = k: (7) fY0 Z0 y k , k 1, … , K , and the density of Y given that Z = k: (8) fY Z y k , k 1, … , K . These densities represent the covariate-response relationship. The marginal densities of Y0 and Y can be written, respectively, as: K K (9) f0 y ¦ k 1 Sk 0 fY Z 0 0 y k and f y ¦ k 1 S k fY Z y k . Then, the counter-factual distribution with the covariate composition of the comparison population and the covariate-response relationship of the reference population is: K (10) f0C y ¦ k 1 S k fY Z0 0 y k , and can be used to decompose the overall relative distribution into a component that represents the effect of changes in the marginal distribution of the covariate (the “composition effect”) and a component that represents the changes in the covariate-response relationship (the “residual effect”). The decomposition can be represented in the following terms: (11) g r f yr f0C y r u f y r . f0 y r f0 y r f0C y r Overall relative Density ratio for Density ratio for density the composition effect the residual effect Comparison of f(yr) to f0C(yr)—i.e. the residual effect—holds the population composition constant, and therefore isolate changes of income distribution due to the fact that returns to the selected covariate changed over time. By contrast, f0C(yr) and f0(yr) have the same covariate-response relationship, and the comparison between them—i.e. the composition effect—isolate the changes due to the different composition of the population under the assumption that the conditional distribution of income remain unchanged. 5 The extensions to continuous and multivariate covariates are considered in Handcock and Morris (1999, ch. 7). © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 8 615 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 4. Data and Results 4.1. The Nigerian Household Consumption Data We posit that a comparison of surveys separated relatively further in time is likely to capture more accurately the effect of structural changes in welfare distri- bution such as inequality, polarization or poverty. This is because in general, absent major shocks, these measures—especially polarization—tend to evolve relatively slowly. In principle, Nigeria surveys present some desirable features that are ideal for conducting long term structural changes. First, they collect consump- tion, which has proven preferable to income because it is less volatile (see e.g. Deaton and Zaidi, 2002, and Haughton and Khandker, 2009). For example, in agricultural economies income is more volatile and affected by growing and harvest seasons, so that relying on income as an indicator of welfare might under- or over-estimate living standards significantly. Second, consumption is a better measure of permanent welfare, because households can borrow, draw down savings, or get public and private transfers to smooth short-run fluctuations. Third, consumption measures what individuals have purchased, while income measures the potential claims of a person. Finally, the surveys also provide detailed information on several other modules that can be used to assess the evolution of non-income measures of well-being. The National Bureau of Statistics (NBS) has conducted two Nigeria Living Standard Surveys (henceforth NLSS) in 2003/2004 and 2009/2010 which it uses to monitor progress in poverty reduction.6 These surveys are representative at state level, use a month-long diary to collect consumption, and collect data for a year (12 month survey). But NBS also conducts other household surveys, most notably the General Household Survey (GHS) cross section and panel. The GHS panel is a randomly selected sub-sample of the GHS cross-section which was collected for the first time in 2010/2011. It consists of 5,000 households, and to date two waves have been completed: 2010/2011 (Wave 1) and 2012/2013 (Wave 2). It is representative at national, rural and urban, and zonal (geo-political) levels. In addition to the questions asked in a normal GHS cross section survey, it contains detailed data on agricultural production and other household income earning activities. Consumption data are substantially more detailed and resemble the consumption data of HNLSS but collected using a 7-day recall period. In every panel wave, households are interviewed two times: once in the “post-planting” period, ranging from August to November, and once in the “post-harvesting” period, ranging from February to April. At first glance, these diverse surveys spanning several years would seem to be the data of choice for studying polarization in Nigeria. Unfortunately, these surveys present several challenging problems for studying welfare changes. In particular, the most obvious pair of surveys to compare, the NLSS 2003/2004 and 2009/2010, could not be used because of major data quality problems (World Bank, 2013). This means that in order to enable the data comparison over a longer time span (a decade) we 6 The 2003/2004 was officially labeled Nigeria Living Standard Survey, while the 2009/2010 was dubbed Harmonized Nigeria Living Standard Survey. Because these are essentially the same types of surveys and conducted as a series with the same purpose in mind, we shall refer to them as NLSSs. C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 9 616 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 need to create two comparable data sets. To do so, we employ survey-to-survey imputation techniques derived from poverty mapping literature (Elbers et al., 2003). The surveys that fit the purpose are NLSS 2003/2004 and GHS-panels (see Appendix A for a detailed explanation of the methodology applied). 4.2. Changes in the Nigerian Household Consumption Distribution Summary measures for household total consumption expenditure per capita in 2003/2004 and 2012/2013 are presented in Appendix B (see Table B.1).7 Besides the growth of the real mean and median consumption expenditures, the most notable feature is that consumption shares of the poorest percentiles of the popu- lation decreased between approximately 1.3 and 1.6 percent a year in the period examined, in contrast to what is observed for the richest percentiles, whose shares experienced average yearly increases of around 1.7 percent. The Gini index grew at an annual average rate of 1.5 percent between 2003/ 2004 and 2012/2013, while the increment in inequality detected by the Theil index is more pronounced, with an average growth rate of 4.2 percent per annum. As for polarization, a sizeable increase is detected by both the Foster-Wolfson (2010) and Duclos-Esteban-Ray (2004) measures, which amounts to around 1.7 percent per year in the first case and almost 1.5 percent in the second. Further insight on the key changes occurring in the distribution of total per capita consumption expenditure of Nigerian households is provided by Figure 1, which shows the density overlay for the two survey waves estimated with an adaptive kernel. Two major observations are apparent from this figure: first, the whole distribution shifted rightward following the increment in the median, and second, there was also an alteration of the shape—the consumption distribution is in fact more dispersed in 2012/2013 than in 2003/2004, as it appears to be charac- terized by a smaller peak and a fatter upper tail that are quite visible in the density overlay. The declines in the mass at the lower and middle ranges of the distribution, and the concomitant spreading out of expenditures in its top half, are also noticeable from Table B.1, where the reported values of the standard deviation, skewness, and kurtosis all show a remarkable growth from one survey wave to the next. However, the graphical display above does not provide much information on the relative impact that location and shape changes had on the differences in the two distributions at every point of the expenditure scale. It also does not convey whether the upper and lower tails of the consumption distribution were growing at the same rate and for what reasons (i.e. location and/or shape driven). As already pointed out in Section 3, this is exactly what the relative distribution method is particularly good at pulling out of the data. We have chosen 2003/2004 as the reference distribution throughout the analysis. It is important to note that reversing the reference and comparison population designation will change the view provided by the relative distribution graph and the displays of the estimated effects of location and shape shifts, because these are defined in terms of the reference population scale. The relative polarization indices, however, are symmetric, meaning that they are effectively 7 In order to enhance comparability of consumption data over the years, all expenditures have been deflated across both space and time and are expressed in 2010 Naira. © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 10 617 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 Figure 1. The distribution of total household consumption expenditure in Nigeria, 2003/2004 and 2012/2013. Expenditures in the upper tiers of the densities have been truncated for better presentation of the graph, where the vertical lines denote the medians of the two survey waves invariant to whether the 2003/2004 or 2012/2013 consumption distribution is chosen as the reference: in fact, swapping the comparison and reference popu- lations yields indices of the same magnitude and opposite sign (e.g. Handcock and Morris, 1999, pp. 71–72, and Hao and Naiman, 2010, pp. 88–89). Thus, reversing the reference and comparison distributions designation will not alter our findings in a substantive way—if not for the fact that polarization would now be analyzed in the reverse direction of time. The relative density of total per capita consumption expenditure of Nigerian households between 2003/2004 and 2012/2013 is examined in Figure 2(a).8 This plot shows the fraction of households in 2012/2013 that fall into each percentile of the 2003/2004 distribution. Households in the low and middle classes moved toward high and, to a less extent, lowest deciles. Indeed, if we choose any percentile approximately between the 2nd and the 80th in the 2003/2004 distribution, the fraction of households in 2012/2013 whose consumption rank corresponds to the chosen percentile is less than the analogous fraction of households in 2003/2004. To get a more detailed picture, we decompose the relative density into loca- tion and shape effects according to Equation (3). Figure 2(b) presents the effect only due to the median shift, that is the pattern that the relative density would have 8 The relative density function has been obtained by fitting a local polynomial to the estimated relative data. C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 11 618 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 (a) Relative distribution (b) Location effect (c) Shape effect Figure 2. Changes in the Nigerian household consumption distribution between 2003/2004 and 2012/2013. The bars represent the decile breakdown of the relative distribution, showing the fraction of 2012/2013 households that fall into each 2003/2004 decile, while dotted lines indicate the 95 percent pointwise confidence limits based on the asymptotic normal approximation (Handcock and Morris, 1999, p. 144) displayed if there had been no change in distributional shape but only a location shift of the density. The effect of the median shift was quite large. This alone would have moved out of the four lowest deciles of the reference distribution a substantial fraction of GHS panel 2012/2013 households and placed them in any of the remaining deciles. Note, however, that neither tail of the observed relative distri- bution is well reproduced by the median shift. For example, the top decile of Figure 2(b) is about 1.1, below the value of 1.5 observed in the actual data, and the bottom deciles of the same figure are also substantially lower than observed. These differences are explained by the shape effect presented in Figure 2(c), which shows the relative density net of the median influence. Without the higher median, the greater dispersion of consumption expenditures in GHS panel 2012/2013 would have led to relatively more low-consuming households in 2012/2013, and this effect was mainly concentrated in the bottom decile. By contrast, at the top of the © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 12 619 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 distribution the higher spread worked in the same direction of the location shift: operating by itself, it would have increased the share of households in the top decile of the 2012/2013 consumption distribution by nearly 50 percent. In sum, once changes in real median expenditure are netted out, a U-shaped relative density is observed, indicating that income (proxied by consumption) polari- zation was hollowing out the middle of the Nigerian household consumption distribution—with a cumulative loss that more than halved the number of house- holds in deciles 2 through 8 of the 2012/2013 distribution. A link between what we have observed in the graphical analysis and the quanti- fication of the degree of polarization is captured by the relative polarization indices. These indices keep track of changes in the shape of the distribution and measure their direction and magnitude. Table 1 reports the median, lower and upper polarization indices computed from the data using Equations (4)–(6), along with their 95 percent confidence intervals and the p-values for testing the null hypothesis of no change with respect to the reference distribution. Using weighted estimates of relative polarization measures, the 95 percent confidence interval for the MRP has been calculated as: c c CI = weighted estimate ± 1.96 × SE, where SE 4 1  2 is the standard error of m n MRP based on the sample variances c1 and c2 of, respectively, ^ Qi  1 ` and m 2 i1 ^ Qi  1 2 `i 1—i.e. the absolute n deviations around the median of the location-matched quasirelative data ^Qi `i 1 (m and n denote the comparison and reference sample sizes). m,n As for the p-value, since our data sets are large survey samples for which the sample sizes tend to be large, we use the normal approximation to the exact distribution of the MRP estimate as the basis for a test for a given significance level α, that is: P(|MRP| ≤ zα/2 × SE) ≈ 1 − α, where zα/2 is the 100 × (1 − α) percentile of the standard normal distribution. Estimation of confidence intervals and p-values for the lower relative polarization (LRP) and upper relative polarization (URP) indices is similar (Handcock and Morris, 1999, ch. 10). The median index is significantly positive, implying a dispersion of the consump- tion distribution from the middle toward either or both of the two tails. The lower and upper polarization estimates indicate that both tails of the distribution are signifi- cantly positively polarized. The upper index, however, is slightly larger, indicating greater polarization in the upper tail of the distribution than in the lower tail. TABLE 1 Relative Polarization Indices Indexa Value LBb UBc p-valued MRP 0.12 0.11 0.13 0.00 LRP 0.11 0.08 0.14 0.00 URP 0.13 0.10 0.16 0.00 Notes: (a) MRP = median relative polarization index, LRP = lower relative polarization index, URP = upper relative polarization index; (b) lower bound of the 95% confidence inter- val; (c) upper bound of the 95% confidence interval; (d) refers to the null hypothesis of no change with respect to the reference distribution, i.e. that the index equals 0. C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 13 620 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 4.3. Covariate Decompositions So far we have focused on comparing the distribution of Nigerian household consumption expenditure between two points in time. However, there are often covariates measured on the households which vary over time, and the impact of these changes on the observed outcomes could be of interest to economic policy and suggest possibilities worthy of consideration by its designers. In the relative distri- bution setting, exploring the distributional impacts of changes in a covariate requires that the relative distribution is adjusted for these changes using the methods from Section 3.4. This makes it possible to separate the impacts of changes in the distribution of the covariate (the “composition effect”) from changes in the condi- tional distributions of household consumption expenditure given the covariate levels (the “residual effect”). Our Nigerian consumption microdata provide an opportunity to use this covariate adjustment technique as they contain a large set of covariates describing various socio-demographic characteristics of the respondents, household assets and characteristics of the dwelling. Here, the analysis is restricted to the following covariates: sex of household head; literacy status of household head; zone; main material used for floor; main source of drinking water; main cooking fuel; main toilet facility. This selection was inspired both from previous poverty research—which advocates the inclusion of covariates that change over time, but excluding those of them that are likely to change markedly in the face of evolving economic conditions (e.g. Stifel and Christiaensen, 2007)—and the fact that many of the covariates excluded from the analysis did not affect the statistical significance of the predicting model used to impute the 2003/2004 data. Summary statistics for the population subgroups defined by the levels of the covariates analyzed and the corresponding average percentage changes between 2003/2004 and 2012/2013 are given in Tables B.2 and B.3 of Appendix B. Both the mean and median consumption expenditures rose during the period analyzed for many population subgroups—exceptions are represented by households headed by illiterate individuals, households with inadequate housing infrastructures (such as unsafe water, low quality flooring material, no toilet facility and firewood as the main cooking device) and households living in the North East and North West zones of the country. At the same time, apart from households in the North Central region, all groups experienced increasing inequality according to both the Gini coefficient and the Theil index. Population and consumption shares changed instead more heterogeneously, following patterns of increases and decreases with different magnitudes over time. In particular, there appears to have been almost no change in the proportion of male-headed households, while female-headed households declined somewhat. By contrast, the fractions of households with a literate head and good quality housing infrastructures (such as safe water, medium-high quality flooring material and non-firewood cooking devices) grew considerably relative to their counterparts—households with no toilet facility, however, are more common in 2012/2013 than in 2003/2004. Finally, the proportions of households that consist of individuals living in the northern zones of the country increased between 2003/ 2004 and 2012/2013, whereas households in the southern regions declined slightly. The above population trends are also visible from Figure 3, which plots the relative distributions of the covariates for 2012/2013 to 2003/2004. Conceptually, © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 14 621 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 (a) Sex of household head (b) Literacy status of household head (c) Zone (d) Main material used for floor (e) Main source of drinking water (f) Main cooking fuel (g) Main toilet facility Figure 3. The relative distributions of the covariates for 2012/2013 to 2003/2004. The upper axes are labelled by the levels of categorical variables. The dotted lines are 95 percent pointwise confidence bounds C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 15 622 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 (a) Zone compositional effect (b) Zone-adjusted relative distribution Figure 4. Adjusting the relative density of Nigerian consumption expenditure for changes in the region of residence (zone) covariate distribution between 2003/2004 and 2012/2013 these relative densities are similar to the one constructed for consumption expen- diture in the previous section, though the graphs are not nearly as smooth because of natural discreteness of the covariates. By reading across the bottom axis one can see the frequencies of reference households cumulated by levels of the covariates, while reading off the y-axis for a given level of the categorical variables allows one to find the relative frequency of comparison households in each group defined by that level. The labels at the top show the categories of the covariates, and can be used for both the reference and comparison populations. However, as already mentioned earlier in this section, in order to assess the impact of changes in population characteristics on the Nigerian consumption distribution the relative density must be decomposed by the distributions of the covariates. This is shown in Figure 4 for the region of residence (zone) covariate, which presents both the covariate composition effect as well as the effect of residual changes—i.e. the expected relative density of Nigerian consumption expenditures had the covariate composition of the 2003/2004 and 2012/2013 populations been identical. The results in panel (a) are pretty close to a uniform distribution, suggesting that the observed differences in population composition according to the selected covariate had little effect on the overall changes occurred over the decade. This perception is confirmed by the adjusted distribution graphed in panel (b) which, in the absence of a major compositional effect, is not much different than the original one depicted in Figure 2(a). Results for other covariates—not shown for brevity, but available upon request—are very similar: there were slight decreases in the bottom half and tiny growth at the top of the distribution associated with some of the compositional shifts in these covariates, but the observed changes were only partly driven by modifications in the characteristics of the population. A similar conclusion can be drawn when looking at Table 2, which presents the set of relative polarization indices for each group defined by the covariates obtained by comparing their consumption distributions over time. Note that, by comparing the subgroup distributions over time, we are effectively controlling for the compo- sitional differences, even though no explicit composition effect is identified. If each © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 16 623 V TABLE 2 Relative Polarization Indices for Different Population Subgroups MRP LRP URP a b c d a b c d a b Index LB UB p-value Index LB UB p-value Index LB UBc p-valued Sex of the household head C 2015 UNU-WIDER Male 0.11 0.10 0.13 0.00 0.10 0.07 0.13 0.00 0.13 0.09 0.16 0.00 Female 0.15 0.11 0.19 0.00 0.15 0.07 0.22 0.00 0.15 0.08 0.23 0.00 Literacy status of household head Illiterate −0.01 −0.03 0.01 0.23 −0.03 −0.07 0.02 0.11 0.01 −0.03 0.05 0.33 Literate 0.10 0.08 0.12 0.00 0.08 0.04 0.13 0.00 0.12 0.08 0.16 0.00 Zone North Central 0.02 −0.01 0.06 0.10 0.03 −0.04 0.10 0.19 0.01 −0.06 0.08 0.35 North East −0.07 −0.11 −0.04 0.00 −0.07 −0.14 0.00 0.02 −0.08 −0.15 −0.01 0.02 North West −0.10 −0.14 −0.07 0.00 −0.15 −0.21 −0.09 0.00 −0.05 −0.12 0.01 0.06 South East 0.14 0.10 0.18 0.00 0.14 0.07 0.21 0.00 0.14 0.07 0.21 0.00 South South 0.21 0.17 0.25 0.00 0.23 0.15 0.31 0.00 0.19 0.11 0.26 0.00 South West 0.21 0.17 0.25 0.00 0.23 0.16 0.31 0.00 0.18 0.11 0.26 0.00 17 624 Main material used for floor Medium quality/High quality 0.08 0.06 0.09 0.00 0.05 0.02 0.09 0.00 0.10 0.06 0.13 0.00 Low quality −0.14 −0.17 −0.10 0.00 −0.26 −0.32 −0.20 0.00 −0.01 −0.10 0.07 0.36 Main source of drinking water Piped/Unprotected 0.04 0.02 0.06 0.00 −0.03 −0.07 0.01 0.10 0.11 0.07 0.16 0.00 Protected 0.16 0.14 0.19 0.00 0.21 0.17 0.25 0.00 0.12 0.07 0.16 0.00 Review of Income and Wealth, Series 63, Number 4, December 2017 Main cooking fuel Charcoal/Kerosene/Oil/ 0.21 0.18 0.24 0.00 0.28 0.22 0.34 0.00 0.15 0.09 0.21 0.00 Electricity/Gas/Other Firewood 0.03 0.01 0.04 0.00 0.01 −0.02 0.04 0.26 0.04 0.01 0.07 0.01 Main toilet facility Flush toilet/Improved pit 0.16 0.14 0.18 0.00 0.17 0.14 0.21 0.00 0.15 0.11 0.18 0.00 latrine/Uncovered pit latrine/Other No facility 0.02 −0.01 0.05 0.16 0.02 −0.04 0.08 0.30 0.01 −0.04 0.07 0.31 Notes: (a) MRP = median relative polarization index, LRP = lower relative polarization index, URP = upper relative polarization index; (b) lower bound of the 95% confidence interval; (c) upper bound of the 95% confidence interval; (d) refers to the null hypothesis of no change with respect to the reference distribution, i.e. that the index equals 0. Review of Income and Wealth 2015 © 2015 International Association for Research in Income and Wealth Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 of the group-specific polarization indices were close to 0, this would imply that after holding changes in the distributions of the covariates constant there is no residual polarization in consumption expenditures. The polarization we observe in the overall consumption distribution would then be due entirely to changing charac- teristics of the population over time. Instead, we see a different scenario. Apart from the North Central households and those with an illiterate head and no toilet facility, the estimates indicates a statistically significant increase of polarization in the subgroup distributions, except for households who reside in the North East and North West regions of the country and those with inadequate flooring in dwelling units, for whom some convergence toward the median is detected. The growth of polarization stems from a shift away from the median of both tails, and this seems to happen asymmetrically, as the LRP indices are in many cases more positive than the URPs—thus indicating more polarization in the lower than in the upper tail. Households headed by men, women or illiterates and households with good flooring material in dwellings, unsafe water and cooking with firewood, instead, are more polarized in the upper than in the lower tail of their consumption distribution—or at least they are so the same way. Overall, these patterns confirm that compositional shifts contributed little to the observed consumption polarization or, in other words, holding the changes in population characteristics constant does almost nothing to reduce overall polarization.9 The above conclusion suggests that the main drivers of polarization are to be searched elsewhere, namely in the changes occurring over the decade in the consump- tion distributions of the groups defined by the covariates. While the covariate adjust- ment technique identifies the impact of changing population characteristics on the distribution of consumption expenditures, comparing the groups defined by the covariates directly makes it possible to analyze the changes within and between these groups’ consumption distributions. As already observed, many population subgroups were both location-shifted (Tables B.2 and B.3) and more polarized (Table 2). To see what impact these location and shape shifts in the subgroups’ distributions had on their relative positions within the overall consumption distribution, we compare the changes in deciles of the between-group relative distributions for 2003/2004 and 2012/2013 to the changes that would have occurred if only the medians or shapes of the groups had changed. More specifically, for each decile we decompose the absolute change: (12) g C : R  g C0 : R0 , where g(C : R) and g(C0 : R0) denote respectively the relative density for compari- son (C) to reference (R) groups of the categorical variables in 2012/2013 and 2003/2004, into the marginal effect of the median shift from the 2003/2004 relative density: (13) g C0 L : R0 L  g C0 : R0 , and those of the shape changes in the subgroups’ consumption distributions: 9 This finding can also serve as a check of whether the observed changes in Nigerian consumption distribution are robust to sample size variations. That is, had the modifications in population charac- teristics due to artefacts of the sample size, rather than to real population trends, our results would not be affected by them. © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 18 625 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 (14) g C0 L : R  g C0 L : R0 L , (15) g C : R  g C0 L : R , where R0L and C0L denote the distributions of the reference and comparison groups adjusted to have the same median as 2012/2013 but with the same shape as 2003/2004.10 Summing up to the total difference given by Equation (12), these effects form a complete decomposition and allow us to determine what proportion of households were moved into or out of a decile of the overall distribution by changes in relative median and group-specific shape. The spatial distribution of household consumption expenditure definitely provided the most attractive results. Figure 5 presents the decomposition for each of the six Nigerian macro-regions as compared to the rest of the country. The solid bars show the total change by decile from Equation (12), and each of the lines represents one of the three components in the decomposition defined by Equations (13)–(15). We can see two ongoing distinctive patterns, both accentuating polarization. In the South South and the South West, relative to the rest of the country, residents tend to move out of the lower deciles of the distribution due to changes in relative median. More precisely, had the location effect been the only one operating, we would have seen in both cases a clear transition of Southeners from lower to upper deciles of the national distribution. However, the shape effect of both regions moved in the opposite direction, partially offsetting the positive impact of growth. Particularly in the lower deciles, the shape change is positive, indicating a clear trend of lower polarization in these areas that goes in the opposite direction vis-à-vis the national (residual) trend. This pattern is mirrored by what is going on in the upper deciles: a location effect higher than in the rest of the country (especially in the South West) and an accentuated tendency to upper polarization in both regions. For what concerns the North East and the North West, the conflict-stricken areas, had the location effect been the only operating force we would have seen a disproportionate increase of people in these regions occupying the lower national deciles compared to the rest of the country—they basically lagged behind. The increase of polarization in the rest of the country helped to offset this effect, filling the lower deciles of households from other regions too, whereas for the rest of the distribution we observe in practice a generalized decline of the relative position of these regions in the national distribution. Finally, while the North Central improves relative to the rest of the country in lower deciles, the South East comes to show a more articulated pattern of distributional change. Results for the other covariates (not shown here but available upon request) looked as expected: compared to 2003/2004, households with an illiterate head or not having good cooking material, toilet, floor and safe water were all increasingly occupying the lower deciles of the distribution, and the gap in terms of consumption with the rest was increasing. Instead, the relative fraction of households headed by females in the upper deciles of the distribution was rising during the period, whereas male-headed households were moving into the deciles below the median. In spite of 10 The decomposition follows the spirit of that presented in Bernhardt et al. (1995) and Handcock and Morris (1998, 1999), to whom we refer the reader for more details. C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 19 626 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 (a) North Central (b) North East (c) North West (d) South East (e) South South (f) South West Figure 5. Sources of distributional change in the 2012/2013 to 2003/2004 relative distribution of consumption expenditures by zone © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 20 627 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 the fact that Nigerian society is mainly patriarchal, where men have better access to productive resources than women, the poor seem more among men than women. 5. Concluding Remarks In the last two decades there have been two different narratives on Sub- Saharan Africa. The first paints a picture of an emerging continent where middle classes are expanding and prosperity is reaching large swaths of the population (African Development Bank, 2011; Fine et al., 2012). The other narrative acknowledges the relatively robust growth in the past two decades, but points to slow reduction in poverty. According to this second narrative, the lack of faster reduction in poverty may be due, in part, to increasing disparities. Nigeria, the most populous country in the African continent, has been enjoying a stable and sustained growth for over a decade since 2003. Yet despite this, the out- comes in terms of poverty reduction have not been satisfactory: while poverty seems to have declined in the coastal South and around the Federal Capital, Abuja, a large belt of North-Eastern states have experienced a clear stagnation in poverty reduction. Our conjecture is that Nigeria in the last decade has been going through signifi- cant changes in the distribution of economic resources that generated mainly, but not exclusively, an increase in polarization. In income-polarized societies people cluster around group means and tend to be far from the mean/median of the overall distri- bution, which results in the inability of the middle class to consolidate its position. This has several economic consequences for a country, but may also be the underlying cause of growing political instability seen in recent years in many middle income countries. Studies of polarization in Nigeria are surprisingly few and have tackled the topic with a narrow lense. This paper contributes to our understanding of polar- ization in Nigeria by using an approach that improves upon previous studies in several ways. First, it considers a longer time horizon—close to a decade—which allows for major changes in welfare distributions to emerge. Second, it tackles a major deficiency—lack of comparability of the available data in Nigeria—by making use of survey-to-survey estimation techniques (Elbers et al., 2003) to achieve comparability of the distributions of interest. Finally, and most impor- tantly, it employs the “relative distribution” approach (Handcock and Morris, 1998, 1999) to analyze changes in the Nigerian household consumption distribu- tion in the considered period. The novelty of this method consists in providing a non-parametric framework for taking into account all of the distributional differ- ences that could arise in the comparison of distributions over time and space. In this way, we are able to summarize multiple features of the expenditure distribu- tion that would not be detected easily from a comparison of standard measures of inequality and polarization. The analysis reveals significant changes in the consumption distribution. We find a clear rise in polarization, meaning that the distributional movements observed between 2003/2004 and 2012/2013 hollowed out the middle of the Nige- rian household consumption distribution and increased concentration of the mass toward higher and lower deciles. This pattern of distributional change, however, is not entirely homogeneous within the country, but varies from zone to zone. Through covariates analysis, C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 21 628 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 controlling for spatial characteristics of household head, we show that in the South (South-West and South-South) lower deciles tend to be emptied relative to the rest of the country, confirming the tendency for households from these zones to contribute to upper polarization. In the North-West and in the conflict-stricken North-East, we see the opposite. The overall impact was a generalized hollowing out of the center and a further accentuation of the North-South divide already characterizing the country. Understanding the political and economic consequences of these sharp distribu- tional changes is beyond the scope of this paper. However, polarization is increasingly becoming a concern in many developing countries. Recent episodes in Brazil, Egypt and Turkey suggest the existence of a link between polarization and conflict, yet so far no relevant empirical evidence has been produced to underpin the existing theoretical models (Esteban and Schneider, 2008). Nigeria is clearly an ideal candidate for such analysis, and our future research will be directed in understanding how existing con- flicts in Nigerian society can be interpreted and linked to the patterns of polarization. References African Development Bank, “Annual Report,” Statistics Department of the African Development Bank Group, 2011. Available at: http://www.afdb.org/knowledge/publications/annual-report/ annual-report-2011/. Ahmed, F. C., C. Dorji, S. Takamatsu, and N. Yoshida, “A Hybrid Survey to Improve the Reliability of Poverty Statistics in a Cost-Effective Manner,” mimeo, 2014. Aigbokhan, B. E., “Poverty, Growth and Inequality in Nigeria: A Case Study,” AERC Research Paper 102, 2000. Available at: http://www.aercafrica.org/index.php/publications/view_document/112- poverty-growth-and-inequality-in-nigeria-a-case-study. Alderson, A. S. and K. Doran, “Global Inequality, Within-Nation Inequality, and the Changing Distribution of Income in Seven Transitional and Middle-Income Societies,” in C. Suter (ed.), Inequality Beyond Globalization: Economic Changes, Social Transformations, and the Dynamics of Inequality, Transaction Publishers, New Brunswick, NJ, 183–200, 2011. ———, “How Has Income Inequality Grown? The Reshaping of the Income Distribution in LIS Countries,” in J. C. Gornick and M. Jäntti (eds), Income Inequality: Economic Disparities and the Middle Class in Affluent Countries, Stanford University Press, Stanford, CA, 51–74, 2013. Alderson, A. S., J. Beckfield, and F. Nielsen, “Exactly How Has Income Inequality Changed? Patterns of Distributional Change in Core Societies,” International Journal of Comparative Sociology, 46, 405–23, 2005. Anderson, G., O. Linton, and T. W. Leo, “A Polarization-Cohesion Perspective on Cross-Country Convergence,” Journal of Economic Growth, 17, 49–69, 2012. Araar, A., “On the Decomposition of Polarization Indices: Illustrations with Chinese and Nigerian Household Surveys,” CIRPEE Working Paper 08-06, 2008. Available at: http://www.cirpee.org/ fileadmin/documents/Cahiers_2008/CIRPEE08-06.pdf. Awoyemi, T. T. and A. Araar, “Explaining Polarization and its Dimensions in Nigeria: A DER Decomposition Approach,” paper prepared for the 14th Annual Conference on “Econometric Modelling for Africa,” Abuja, Nigeria, July 8–10, 2009. Available at: http://www .africametrics.org/documents/conference09/papers/Awoyemi_Araar.pdf. Awoyemi, T. T., I. B. Oluwatayo, and A. Oluwakemi, “Inequality, Polarization and Poverty in Nigeria,” PEP Working Paper 2010–04, 2010. Available at: http://www.pep-net.org/ communications/working-papers/. Backiny-Yetna, P., D. Steele, and I. Y. Djima, “The Impact of Household Food Consumption Data Collection Methods on Poverty and Inequality Measures in Niger,” World Bank Policy Research Working Paper 7090, 2014. Available at: http://go.worldbank.org/C6C11G2R30. Beegle, K., J. De Weerdt, J. Friedman, and J. Gibson, “Methods of Household Consumption Mea- surement through Surveys: Experimental Results from Tanzania,” World Bank Policy Research Working Paper 5501, 2010. Available at: http://go.worldbank.org/UE8VCVJXK0. Bernhardt, A., M. Morris, and M. S. Handcock, “Women’s Gains or Men’s Losses? A Closer Look at the Shrinking Gender Gap in Earnings,” American Journal of Sociology, 101, 302–28, 1995. Bianchi, M., “Testing for Convergence: Evidence from Nonparametric Multimodality Tests,” Journal of Applied Econometrics, 12, 393–409, 1997. © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 22 629 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 Birdsall, N., “The (Indispensable) Middle Class in Developing Countries; or, the Rich and the Rest, Not the Poor and the Rest,” CGD Working Paper 207, 2010. Available at: http://www.cgdev.org/ files/1423994_file_Birdsall_Indispensable_Middle_FINAL.pdf. Borraz, F., N. González, and M. Rossi, “Polarization and the Middle Class in Uruguay,” Latin American Journal of Economics, 50, 289–326, 2013. Bourguignon, F., “The Growth Elasticity of Poverty Reduction: Explaining Heterogeneity across Countries and Time Periods,” in T. S. Eicher and S. J. Turnovsky (eds), Inequality and Growth: Theory and Policy Implications, The MIT Press, Cambridge, MA, 3–26, 2003. ———, “The Poverty-Growth-Inequality Triangle,” ICRIER Working Paper 125, 2004. Available at: http://www.icrier.org/pdf/wp125.pdf. Chakravarty, S. R., Inequality, Polarization and Poverty: Advances in Distributional Analysis, Springer- Verlag Inc., New York, NY, 2009. Chakravarty, S. R. and C. D’Ambrosio, “Polarization Orderings of Income Distributions,” Review of Income and Wealth, 56, 47–64, 2010. Chakravarty, S. R. and A. Majumder, “Inequality, Polarization and Welfare: Theory and Applica- tions,” Australian Economic Papers, 40, 1–13, 2001. Chakravarty, S. R., A. Majumder, and S. Roy, “A Treatment of Absolute Indices of Polarization,” Japanese Economic Review, 58, 273–93, 2007. Conti, P. L., M. G. Pittau, and R. Zelli, “Metodi non parametrici nell’analisi della distribuzione del reddito: problemi empirici ed aspetti metodologici,” Rivista di Politica Economica, 96, 195–242, 2006. Cowell, F. A., S. P. Jenkins, and J. A. Litchfield, “The Changing Shape of the UK Income Distribution: Kernel Density Estimates,” in J. Hills (ed.), New Equalities: The Changing Distribution of Income and Wealth in the United Kingdom, Cambridge University Press, Cambridge, UK, 49–75, 1996. D’Ambrosio, C., “Household Characteristics and the Distribution of Income in Italy: An Application of Social Distance Measures,” Review of Income and Wealth, 47, 43–64, 2001. Deaton, A. and J. Drèze, “Poverty and Inequality in India: A Re-examination,” Economic and Political Weekly, 37, 3729–48, 2002. Deaton, A. and M. Grosh, “Consumption,” in M. Grosh, and P. Glewwe (eds), Designing Household Survey Questionnaires for Developing Countries: Lessons from 15 Years of the Living Standards Measurement Study, volume 1, Oxford University Press, Oxford, UK, 91–133, 2000. Deaton, A. and V. Kozel, “Data and Dogma: The Great Indian Poverty Debate,” World Bank Research Observer, 20, 177–99, 2005. Deaton, A. and S. Zaidi, “Guidelines for Constructing Consumption Aggregates for Welfare Analy- sis,” LSMS Working Paper 135, 2002. Available at: http://go.worldbank.org/JJSKLIKYW0. Doudich, M., A. Ezzrari, R. Van der Weide, and P. Verme, “Estimating Quarterly Poverty Rates Using Labor Force Surveys: A Primer,” World Bank Policy Research Working Paper 6466, 2013. Available at: http://go.worldbank.org/XH0FPIQND0. Duclos, J.-Y., J.-M. Esteban, and D. Ray, “Polarization: Concepts, Measurement, Estimation,” Econometrica, 72, 1737–72, 2004. Easterly, W., “The Middle Class Consensus and Economic Development,” Journal of Economic Growth, 6, 317–35, 2001. Elbers, C., J. O. Lanjouw, and P. Lanjouw, “Micro-Level Estimation of Poverty and Inequality,” Econometrica, 71, 355–64, 2003. Esteban, J.-M. and D. Ray, “On the Measurement of Polarization,” Econometrica, 62, 819–51, 1994. ———, “Conflict and Distribution,” Journal of Economic Theory, 87, 379–415, 1999. ———, “Polarization, Fractionalization and Conflict,” Journal of Peace Research, 45, 163–82, 2008. ———, “Linking Conflict to Inequality and Polarization,” The American Economic Review, 101, 1345–74, 2011. Esteban, J.-M. and G. Schneider, “Polarization and Conflict: Theoretical and Empirical Issues: Intro- duction,” Journal of Peace Research, 45, 131–41, 2008. Esteban, J.-M., C. Gradín, and D. Ray, “An Extension of a Measure of Polarization, with an Appli- cation to the Income Distribution of Five OECD Countries,” Journal of Economic Inequality, 5, 1–19, 2007. Ferreira, F. H. G., J. Gignoux, and M. Aran, “Measuring Inequality of Opportunity with Imperfect Data: The Case of Turkey,” Journal of Economic Inequality, 9, 651–80, 2011. Fine, D., A. van Wamelen, S. Lund, A. Cabral, M. Taoufiki, N. Dörr, A. Leke, C. Roxburgh, C. J. Schubert, and P. Cook, “Africa at Work: Job Creation and Inclusive Growth,” McKinsey Global Institute Report, 2012. Available at: http://www.mckinsey.com/insights/africa/africa_at_work. Fortin, N., T. Lemieux, and S. Firpo, “Decomposition Methods in Economics,” in O. Ashenfelter and D. Card (eds), Handbook of Labor Economics, volume 4A, chapter 1, North Holland, Amsterdam, 1–102, 2011. C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 23 630 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 Foster, J. E. and M. C. Wolfson, “Polarization and the Decline of the Middle Class: Canada and the US,” Journal of Economic Inequality, 8, 247–73, 2010. Gigliarano, C. and K. Mosler, “Constructing Indices of Multivariate Polarization,” Journal of Eco- nomic Inequality, 7, 435–60, 2009. Gradín, C., “Polarization by Sub-Populations in Spain, 1973–91,” Review of Income and Wealth, 46, 457–74, 2000. Handcock, M. S. and M. Morris, “Relative Distribution Methods,” Sociological Methodology, 28, 53–97, 1998. ———, Relative Distribution Methods in the Social Sciences, Springer-Verlag Inc., New York, NY, 1999. Hao, L. and D. Q. Naiman, Assessing Inequality, SAGE Publications Inc., Thousand Oaks, CA, 2010. Haughton, J. and S. R. Khandker, Handbook on Poverty and Inequality, The World Bank, Washington, DC, 2009. Henderson, D. J., C. Parmeter, and R. R. Russell, “Modes, Weighted Modes, and Calibrated Modes: Evidence of Clustering Using Modality Tests,” Journal of Applied Econometrics, 23, 607–38, 2008. Holzmann, H., S. Vollmer, and J. Weisbrod, “Twin Peaks or Three Components? Analyzing the World’s Cross-Country Distribution of Income”, IAI Discussion Paper 162, 2007. Available at: http://www2.vwl.wiso.uni-goettingen.de/ibero/papers/DB162.pdf. Jenkins, S. P., “Did the Middle Class Shrink During the 1980s? UK Evidence from Kernel Density Estimates,” Economic Letters, 49, 407–13, 1995. ———, “Recent Trends in the UK Income Distribution: What Happened and Why?,” Oxford Review of Economic Policy, 12, 29–46, 1996. Jenkins, S. P. and P. Van Kerm, “Accounting for Income Distribution Trends: A Density Function Decomposition Approach,” Journal of Economic Inequality, 3, 43–61, 2005. Johnson, P. A., “A Nonparametric Analysis of Income Convergence Across the US States,” Economics Letters, 69, 219–23, 2000. Jones, C. I., “On the Evolution of the World Income Distribution,” Journal of Economic Perspectives, 11, 19–36, 1997. Kakwani, N., “Poverty and Economic Growth with Application to Côte d’Ivoire,” Review of Income and Wealth, 39, 121–39, 1993. Kijima, Y. and P. Lanjouw, “Poverty in India During the 1990s—A Regional Perspective,” World Bank Policy Research Working Paper 3141, 2003. Available at: http://documents.worldbank.org/ curated/en/2003/10/2579638/poverty-india-during-1990s-regional-perspective. Lasso de la Vega, M. C. and A. M. Urrutia, “An Alternative Formulation of Esteban-Gradín-Ray Extended Measure of Polarization,” Journal of Income Distribution, 15, 42–54, 2006. Lasso de la Vega, M. C., A. M. Urrutia, and H. Díez, “Unit Consistency and Bipolarization of Income Distributions,” Review of Income and Wealth, 56, 65–83, 2010. Levy, F. and R. J. Murnane, “U.S. Earnings Levels and Earnings Inequality: A Review of Recent Trends and Proposed Explanations,” Journal of Economic Literature, 30, 1333–81, 1992. Massari, R., “Is Income Becoming More Polarized in Italy? A Closer Look with a Distributional Approach,” DSE Working Paper 1, 2009. Available at: http://phdschool-economics.dse .uniroma1.it/website/WorkingPapers/MassariWP1.pdf. Massari, R., M. G. Pittau, and R. Zelli, “A Dwindling Middle Class? Italian Evidence in the 2000s,” Journal of Economic Inequality, 7, 333–50, 2009a. ———, “Caos calmo: l’evoluzione dei redditi familiari in Italia,” in L. Cappellari, P. Naticchioni, and S. Staffolani (eds), L’Italia delle disuguaglianze, Carocci editore, Rome, 19–28, 2009b. Milanovic, B., “A New Polarization Measure and Some Applications,” mimeo, 2000. Available at: http://siteresources.worldbank.org/INTDECINEQ/Resources/polariz.pdf. Morris, M., A. D. Bernhardt, and M. S. Handcock, “Economic Inequality: New Methods for New Trends,” American Sociological Review, 59, 205–19, 1994. Ogunyemi, O. I., “Gender, Market Participation and Income Polarisation Among Farming and Non- Farming Households in Nigeria,” paper presented at the World Bank Workshop on “African Agriculture: Myths and Facts”, Washington DC, US, November 20–22, 2013. Ogunyemi, O. I. and O. A. Oni, “Income Bipolarization and Poverty: Evidence from Rural Nigeria,” Journal of American Science, 7, 722–31, 2011. Ogunyemi, O. I., O. A. Oni, T. T. Awoyemi, and S. A. Yusuf, “Income Polarization and Bipolarization Across Rural Households’ Socio-Economic Features In Nigeria,” World Rural Observations, 3, 62–72, 2011. Paap, R. and H. K. van Dijk, “Distribution and Mobility of Wealth of Nations,” European Economic Review, 42, 1269–93, 1998. Pittau, M. G. and R. Zelli, “Income Distribution in Italy: A Nonparametric Analysis,” Statistical Methods and Applications, 10, 175–89, 2001. © 2015 International Association for Research in Income and Wealth C 2015 UNU-WIDER V 24 631 Review of Income and Wealth 2015 Review of Income and Wealth, Series 63, Number 4, December 2017 ———, “Testing for Changing Shapes of Income Distribution: Italian Evidence in the 1990s from Kernel Density Estimates,” Empirical Economics, 29, 415–30, 2004. ———, “Trends in Income Distribution in Italy: A Non-Parametric and a Semi-Parametric Analysis,” Journal of Income Distribution, 15, 90–118, 2006. Pittau, M. G., R. Zelli, and P. A. Johnson, “Mixture Models, Convergence Clubs and Polarization,” Review of Income and Wealth, 56, 102–22, 2010. Poggi, A. and J. Silber, “On Polarization and Mobility: A Look at Polarization in the Wage-Career Profile in Italy,” Review of Income and Wealth, 56, 123–40, 2010. Pressman, S., “The Decline of the Middle Class: An International Perspective,” Journal of Economic Issues, 41, 181–200, 2007. Quah, D. T., “Convergence Empirics Across Economies with (Some) Capital Mobility,” Journal of Economic Growth, 1, 95–124, 1996a. ———, “Twin Peaks: Growth and Convergence in Models of Distribution Dynamics,” The Economic Journal, 106, 1045–55, 1996b. ———, “Empirics for Growth and Distribution: Stratification, Polarization, and Convergence Clubs,” Journal of Economic Growth, 2, 27–59, 1997. Reynal-Querol, M., “Ethnicity, Political Systems and Civil War,” Journal of Conflict Resolution, 46, 29–54, 2002. Rodríguez, J. G. and R. Salas, “Extended Bi-Polarization and Inequality Measures,” Research on Economic Inequality, 9, 69–83, 2003. Silber, J., J. Deutsch, and G. Yalonetzky, “On Bi-Polarization and the Middle Class in Latin America: A Look at the First Decade of the Twenty-First Century,” Review of Income and Wealth, 60, S332–52, 2014. Silber, J., M. Hanoka, and J. Deutsch, “On the Link Between the Concepts of Kurtosis and Bipolar- ization,” Economics Bulletin, 4, 1–6, 2007. Stifel, D. and L. Christiaensen, “Tracking Poverty Over Time in the Absence of Comparable Con- sumption Data,” World Bank Economic Review, 21, 317–41, 2007. Tarozzi, A., “Calculating Comparable Statistics from Incomparable Surveys, with an Application to Poverty in India”, Journal of Business & Economic Statistics, 25, 314–36, 2007. Wang, Y.-Q. and K.-Y. Tsui, “Polarization Orderings and New Classes of Polarization Indices,” Journal of Public Economic Theory, 2, 349–63, 2000. Wolfson, M. C., “When Inequalities Diverge,” The American Economic Review, 84, 353–8, 1994. Wolfson, M. C., “Divergent Inequalities: Theory and Empirical Results,” Review of Income and Wealth, 43, 401–21, 1997. World Bank, “Nigeria. Where Has All the Growth Gone? A Poverty Update,” Report 143689-NGA, 2013. Available at: http://documents.worldbank.org/curated/en/2013/08/19873872/all-growth- gone-poverty-update-nigeria-policy-note. World Bank, Nigeria Economic Report, 2014. Available at: http://documents.worldbank.org/curated/ en/2014/07/19883231/nigeria-economic-report-no-2. Zhang, X. and R. Kanbur, “What Difference Do Polarisation Measures Make? An Application to China,” Journal of Development Studies, 37, 85–98, 2001. Supporting Information Additional Supporting Information may be found in the online version of this article at the publisher’s web-site: Appendix A: Imputation Method Table A.1: Regression results of POV_MAP imputation model Table A.2: Mean and standard deviation of relative polarization indices over R = 50 simulation runs for three alternative imputation techniques Figure A.1: Post-imputation diagnostic plots Appendix B: Tables Table B.1: Summary measures of Nigerian household total consumption expenditure per capita Table B.2: Summary measures for Nigerian household consumption expenditure by population subgroups, 2003/2004 and 2012/2013 Table B.3: Summary measures for Nigerian household consumption expenditure by population subgroups, average annual compound percentage changes from 2003/2004 to 2012/2013 C 2015 UNU-WIDER V © 2015 International Association for Research in Income and Wealth 25 632