 Well, good afternoon everyone. My name is Ronald Schill. I'm from the University of Cape Town and today I have the honor of presenting the paper Assessing the impact of social grants and inequality a South African case study and I'm presenting this on behalf of my fellow court authors Mary Lee Brunt and David Lamb. The purpose of this paper is to investigate the expansion of the social grant system in South Africa post the fall of apartheid and the effect that's had on inequality and to use this case study as an example as an opportunity to compare income decomposition techniques across a variety of techniques in order to gain insight into what these techniques can tell us and the advantages and disadvantages of the various techniques So in order to cover that I will be following the roughly the next the following outline First we'll cover some introductory comments just to provide some context and background to the paper Next we'll briefly discuss the data employed before continuing on to the analysis in order The analysis is broken down into three sections first. We apply a static decomposition technique followed by dynamic income decomposition technique and then finally dynamic income decomposition technique Using simulations before concluding this allows us to not only assess the impact of social grants on inequality in South Africa But also allows us to put each decomposition technique against one another or each other Thus we can see what the effect social grants has had on inequality By viewing the issue through different lenses and in doing so learning something about the lenses themselves Since the transition to democracy in 1994 the South African government has dramatically expanded the social system the system of social grants in South Africa building on The existing ratio lead by a social grant system Developed by the apartheid government the new government instituted a variety of new grants as well as increase the targeting of various grants What this figures effectively shows is that in 1993 in 1993? The percentage of income from the government was not that dramatic across all the deciles. However, fast forward to 1993 2008 and we can see that the expansion of the social grant system Both in terms of the volume in terms of number of grants issued and number of grants instituted has dramatically increased over Across all day cells and particularly for the poorer day cells where the majority of their income is now Attributed to transfers from government. This shows the impact that the policy has had on poor income poor individuals especially The rollout or the expansion the system has come through two avenues firstly the institution of the child support grant which as we can see over the period over the period of 1999 to 2010 Has dramatically increased in expenditure and the number of Grants claimed the child support grant is a grant targeted specifically at low-income mothers who have children in addition the States all day pension or the pension which is targeted at individuals who are tired of Over the age of 60 has increased to include individuals of all racial groups So to assess the impact of these grants on inequality, we just briefly plot the distributions to give us a taste of what the situation is like So in 1993 We can see that the income total household income per capita before Taking into account grants lies likely to the left if we exclude the grants in total This implies that for the individuals at this end of the distribution the grant is the grants for the social transit system has greatly aided them However in 2008 we can do the exact same analysis and you can see that there's a bigger shift when we include grants in comparing total household income So this gives the idea that as previously showed that grants have mainly been helping the individuals at the lower end and the expansion in particular has made great strides in doing so what is interesting now is to relate this back to inequality We are curious to know what the impact of inequality of these grants are on inequality or income inequality in South Africa And so just looking at the distribution So I would expect that seeing that the distribution which includes income income from grants is narrower You would expect that the grants has had some Equalizing effect in terms of income inequality as a whole so in order to sufficiently assess the impact of these grants on In the quality we turn to we require we have data which we require two specific areas of Firstly we need data of two time periods one before the expansion of the social grant system And secondly one post the expansion and what we require of these two peer These data sources is that the master quite detailed information on the exact source of income So that within a period we can decompose total household income into various buckets or various categories And this is how these buckets have been changing over time And so as a result we turn to the project for statistics on stems of living conducted 1993 as our pre-democracy Snapshot and then using the 2008 15 years later national income dynamics study to give us our picture of South Africa with which would include the expansion of the social grant system The table one provides some basic descriptive statistics on the total household income per capita and shows that pretty much the Genie coefficient has increased slightly but not dramatically and that despite the expansion of the social grant system We do not see a decrease in inequality on an aggregate level But we've done now is to in order to assess the impact of social grants on inequality We've broken down total household income into four buckets or four main categories to which to assess Firstly we look at house of labor income per capita Secondly into the first social grant the state's old age pension as I mentioned was targeted all the individuals Secondly Sorry our third bucket is then government chances or other government chances which will include this massive expansion in child support grant and Finally any other income which individuals may have So it's a couple of things just to point out From the descriptive statistics, which gives us some insight into the situation is a large share that labor income Attributes to the total share of household income roughly across the period Or with in within both periods labor income makes up the vast majority of income of a household and that's roughly at 60% however Within labor income is quite large inequality in both periods The expansion of the social grant system can be easily seen here when we look at the change of average From three to 28 in other government transits which captures this effect of the massive expansion in the social in the child support grant of the period But we can also see from these two is that the two government trances Has negative correlations with total income which hints at the means tested nature of the grants in South Africa So in order to assess what these grants have on inequality We apply different lenses we look at it through different lenses in South Africa and the first Decomposition technique that we use is the lemon chisaki approach And what this approach does is it breaks down that in the Genie coefficient in particular into three main drivers per component or per income category that we've used So the first the first component over here is the correlation that the genie coefficient has between income between the income component itself and Total income Secondly, it looks at the genie coefficient for each income source and finally it looks at the share of That particular source in total income in terms of total income So collectively what it's trying to capture is what is inequality within a particular household income component What is the inequality within that component and how does that relate to income overall? This is a static approach implying that it's only done within a particular period And so using the stock extension of the method We can take the partial derivative approach to see what would happen if we change Always we increase the particular income component slightly within the period and that is given by this formula over here So here we can see our results given the static income position Approach given here, and I'd just like to draw your attention to some Some findings in particular Firstly, you can see that labor income or the income obtained from labor market activities has greatly Resulted in increasing inequality over time in the 1993 period The two government transfers Firstly the old age pension as well as the other government transfers Which captures the child support grant has resulted to a decrease in inequality within the period But because this is a static Decomposition approach we have to do it for both 1993 and 2008 Separately and contrast the results across the two data sets So in 2008 we can see that once again labor has had quite a large Disequalising effect and in comparing this to the 93 we can see that it's this equalizing effect has somewhat increased Interestingly enough the old age pension in 2008 has a this equalizing effect This is sort of counter-intuitive you would expect that a grant could have an equalizing effect But this is not the results that obtained from the static Decomposition in 2008 period however other government transfers keeps Remains somewhat equalizing although the curve the number is quite dramatically small however, the main disadvantage that the static approach has is it's one dimensionality as it provides a Short of the drivers of income inequality in a particular period and the story is quite limited and it says in how these things change Over the course of the period in question And so we move on from a static decomposition technique to a dynamic decomposition technique Which explicitly takes into account the change in the in the genie coefficient So we can assess how changes in each income decomposition technique has affected the change in the genie overall So what the WAN approach does is Is that it calculates first a concentration index of a particular Component as well as the share of the component of total income and using these two divides them into three separate categories The first category which is outlined over here Is labeled the structural effect which mainly captures the effect of changes in shares of income on the change in genie coefficient The second is how the change within a component in equality Affects the change in genie and that is kept named the real in Inequality effect and then finally over here is how changes in both effects Affects the change in genie and it's appropriately labeled the interaction effect Applying the WAN approach that the Dynamic decomposition approach to the solid data we obtained the following results. I Draw your attention to the final column here the full contribution Effective you can see that labor income has had quite a positive which represents a disequalizing effect on income inequality over the period The two government transfers both indicate negative effects which shows that they are quite equalizing in fact They've reduced income inequality over the course of the period and that other income has also related has resulted in disequalizing effect So in essence what we capture is is that the changes in the In the genie coefficient over the period has been joined mainly by Labor income and that the two social grants or the two buckets of government income that we've created has Rulter has resulted in decrease in income inequality However, there's a disadvantage of the WAN approach and is that that it doesn't isolate the exact impact of a change in income inequality In ideal situation, we would like to compare the extension of the social grant system over the period Over the period between 1993 and 2008 to a counterfactual situation in essence We would like to look at how the 1993 social grant system would operate in the 2008 world So a novel approach by Barros tries to create this counterfactual which we can use as a comparison By basing it on a series of simulation. So similar to the previous approaches. We can divide per capita household Income into two broad categories into various categories. So just for illustrative purposes over here I've just broken it down into government income and non-government income while the analysis breaks down into our original four components then assuming leading FB The cumulative distribution function of income which is then dependent of course on these two components We can calculate any welfare indicator which is based on some function of the distribution cumulative distribution as a whole So in essence what the approach does is that it plays around with these components in calculating the various effects And so to give you like an illustrative example Well, let me check you through the technique Firstly the indicator is calculated as it is for the 2008 period Then subsequently The indicator is recalculated, but one particular component is changed. So in essence The 1993 component for non-government income is substituted into the income into the cumulative distribution function of the 2008 income and then the inequality figure is recalculated The difference between the 2008 which contains only 2008 components and then 2008 which includes the one component Which is 1993 then gives us the impact of these two Of the change of that particular component across the period Let me talk you through this in a bit more of a practical example Firstly we hustles are ranked according to the It can be ranked according to either the value in the total income or by their values within a particular component and then We calculate the Gini coefficient for the 2008 period Then we swap, we aim to recalculate the 2008 Gini coefficient But by swapping one component within its 1993 value and the way this is achieved is by doing it within quantiles So if a household is ranked in 2008 in a particular quantile We replace that value with its 1993 counterpart and then recalculate the curve The Gini coefficient and the difference within it gives us the impact of that the change in a component So in essence we capture what would have happened if that 1993 So as component was operating in the 2008 world This technique is quite useful because it tries to aim at that capturing the counterfactual Which we are looking for but it also gives us allows us to make two additions to the such situation Firstly we can take into we can explicitly take into account A household composition changes So in particular how we can break down this equation over here into a more detailed equation Which allows us to take into our household decomposition such as the number of individuals or number with in our household the number of adults with inner households and the number of employed individuals in a household So you can see how changes within a household itself or how households change has had an effect on income inequality Secondly because this technique is highly based on re-ranking of other ranking of the distribution and substituting across the time period We can rank individuals and do the substitution based not only on total income But also an income component itself And so you can see how changes in the distribution of a particular income component has had an impact on inequality and this particularly useful with regards to social grants which Basically relates to how targeting or the change in targeting within some particular social grants has had an impact on inequality And so just applying the simple approach over here without taking into account the Household decompositions or the re-ranking's we get the following results for our four categories We can see that labor that changes in labor of the period has contributed to an increase in the genie coefficient similar to our previous approaches Secondly the states the change in old age pensions grant from the government has resulted in a slight increase in Inequality over the period However, transits other Transits from the government which mainly captures the child support grant has dramatically resulted in a decrease in inequality over the period Over here we to do the more nuanced approach we take to account changes in household composition as well as try and aim It's affecting how changes in targeting in a particular grant system or the distribution changes Which we change through the re-ranking is obtained So just first focusing on household composition We can see that as the share of adults has changed of the period It has resulted in a slight increase in inequality or income inequality within the how across the period However, the share of individuals the share of the proportion of adults that are employed has resulted in a decrease in inequality total labor income Has similarly using this more advanced technique which now takes to account Hustle composition we see that labor income actually results in a small change in income a small decrease in inequality across the period The old age pension once again as our previous results I suggested has somewhat Resulted in a disequalizing effect across the period while other government transfers our second grant Has left to a decrease or an equalizing effect across the period So finally just to conclude and wrap up what these three techniques have shown us not only about South Africa But also about themselves we can see that each of these activities often takes a Different approach in trying to answer the exact same question either through explicitly accounting for the dynamic nation across the period Or by explicitly accounting for a counterfactual situation which allows it to control on isolate their faiths The dynamic decompositions would seem to have to offer more in terms of looking at the impact of social grants and inequality especially decomposition techniques that consider the impact that changes in household composition has on Changes in real income value across the period Still we're not getting much traction as we'd like to see given that those densities initially showed a decrease in the distribution That we had first. Thank you