{smcl} {* *! version 1.0.0 29jun2017}{...} {findalias asfradohelp}{...} {vieweralsosee "" "--"}{...} {vieweralsosee "[R] help" "help help"}{...} {viewerjumpto "Syntax" "msdeco##syntax"}{...} {viewerjumpto "Description" "msdeco##description"}{...} {viewerjumpto "Technical details" "msdeco##techdetails"}{...} {viewerjumpto "Options" "msdeco##options"}{...} {viewerjumpto "Saved Results" "msdeco##saved"}{...} {viewerjumpto "Examples" "msdeco##examples"}{...} {viewerjumpto "Author" "msdeco##author"}{...} {viewerjumpto "References" "msdeco##references"}{...} {viewerjumpto "See also" "msdeco##seealso"}{...} {hline} help for {hi:msdeco}{right:Andrew Silva (June 2017)} {hline} {title:Title} {phang} {bf:msdeco} {hline 2} Calculates the Mookherjee & Shorrocks (1982) over-time inequality decomposition by subgroup {marker syntax}{...} {title:Syntax} {p 8 17 2} {cmd:msdeco} {varname} {ifin} {weight} {cmd:,} {cmdab:group:var(}{it:varname}{cmd:}} {cmdab:year:var(}{it:varname}{cmd:}} {cmdab:start:year(}{it:integer}{cmd:}} {cmdab:end:year(}{it:integer}{cmd:}} [{cmdab:p:reserve}] {synoptset 20 tabbed}{...} {synopthdr} {synoptline} {syntab:Main} {synopt:{opt group:var(varname)}}variable indicating group ID{p_end} {synopt:{opt year:var(varname)}}variable indicating year{p_end} {synopt:{opt start:year(#)}}starting year{p_end} {synopt:{opt end:year(#)}}ending year{p_end} {synopt:{opt p:reserve}}preserve data state before processing; greatly reduces computation time{p_end} {synoptline} {p2colreset}{...} {p 4 6 2} {cmd:fweight}s and {cmd:aweight}s are allowed; see {help weight}. {marker description}{...} {title:Description} {pstd} {cmd:msdeco} calculates the Mookherjee & Shorrocks over-time decomposition for the specified {it:varname} with respect to groups in {cmdab:group:var}, from {cmdab:star:tyear} to {cmdab:end:year}. The decomposition is based on the {it:GE(0)} (Generalized Entropy of class 0) inequality measure, also known as the mean logarithmic deviation. As such, values of the primary input variable are restricted to positive numbers. {cmd:msdeco} depends upon {cmd:ineqdeco}, an excellent user-written routine by Jenkins (2015), which computes a panel of inequality indices and associated statistics that provide the building blocks of the Mookherjee & Shorrocks decomposition. {marker techdetails}{...} {title:Technical details} {p 4 4 2} The Mookherjee & Shorrocks (1982) decomposition is an approximate decomposition of the GE(0) (mean logarithmic deviation) measure of inequality. In a weighted sample with sample weights w_i, income y_i, mean income m, and f_i = w_i / SUM (w_i), GE(0) (I_0 in the author's notation) is defined as: {p 4 4 2}GE(0) = I_0 = SUM f_i log(m / y_i). {p 4 4 2} See Jenkins (1999) for more coverage of the GE indices and their computation in discrete data. {p 4 4 2} {cmd:msdeco} reports the cecomposition results in four components, labeled A, B, C, and D, which correspond to equations (14a), (14b), (14c), and (14d) in the original paper, respectively. This is a slight approximation of the exact decomposition, but offers a more practical interpretation that the exact decomposition. Using the author's original notation, the components are: {p 8 8 2} A: SUM { m(v_k) D(I_k) } {break}(Effect due to changes in within subgroup inequality) {p 8 8 2} B: SUM { m(I_k) D(v_k) } {break}(Effect due to changes in population shares of within component) {p 8 8 2} C: SUM { [ m(lambda_k) - m(log lambda_k) ] D(v_k) } {break}(Effect due to changes in population shares of between component) {p 8 8 2} D: SUM { [m(theta_k) - m(v_k)] D(log mu_k) } {break}(Effect due to relative changes in subgroup means) {p 4 4 2} where SUM is the summation operator, summing over index k, m() is the mean-over-time operator and D() is the "delta" difference-over-time operator. v_k = n_k/n, or the proportion of the population in group k; I_k is the GE(0) of group k; lambda_k = mu_k/mu, the relative mean income of group k to the mean of the entire population; and theta_k = v_k * u_k, the income share of group k. {p 4 4 2} {cmd:msdeco} also computes the exact decomposition, provided in equation (12) of the original paper. This form unfortunately does not provide the same neatly-identified conceptual components, yet is easily presented in four distinct terms, written as: {p 8 8 2} A: SUM { m(v_k) D(I_k) } (same as approximate decomposition above) {p 8 8 2} B: SUM { m(I_k) D(v_k) } (same as approximate decomposition above) {p 8 8 2} C: -SUM { m(log lambda_k) D(v_k) } {p 8 8 2} D: -SUM {m(v_k) D(log lambda_k) } {p 4 4 2} Note the leading "-" sign on the C and D terms, implying that if the reported C or D term is negative, the summation term (with out the "-" sign) is positive. {marker options}{...} {title:Options} {dlgtab:Main} {phang} {opt group:var(varname)} specifies the variable indicating the group ID (mandatory). Group IDs values must be non-negative integers, and must have the same levels in both {cmd:startyear} and {cmd:endyear} among {it:positive} values of the primary variable to be decomposed. {phang} {opt year:var(varname)} specifies the variable indicating the year (mandatory). The year variable should be of type numeric. {phang} {opt start:year(#)} specifies the starting year for the decompostion (mandatory), the value of which will be looked up in variable specified by {cmdab:year:var()}. {phang} {opt end:year(#)} specifies the ending year for the decompostion (mandatory), the value of which will be looked up in variable specified by {cmdab:year:var()}. {phang} {opt p:reserve} uses the {cmd:preserve} and {cmd:restore} system commands to restore the original data state after the command has completed. This allows the routine to avoid passing {cmd:if} and {cmd:in} qualifiers to subroutines, and instead drops unused observations, which greatly reduces computation time. If the {cmdab:p:reserve} option is used, {cmd:msdeco} may not be used within an existing preserve-restore block, as Stata does not permit nesting of preserved data states. {marker saved}{...} {title:Saved results} Scalars r(A) effect of changes in within subgroup inequality r(B) effect of changes in population shares of within component r(C) effect of changes in population shares of between component r(D) effect of relative changes in subgroup means r(I0_dif_approx) sum of A, B, C, and D components (from approximate decomposition) r(Aexact) effect of changes in within subgroup inequality; same as r(A) r(Bexact) effect of changes in population shares of within component; same as r(B) r(Cexact) -SUM [ m(log lambda_k) D(v_k) ] (from exact decomposition) r(Dexact) -SUM [m(v_k) D(log lambda_k) ] (from exact decomposition) r(I0_dif_exact_sum) sum of Aexact, Bexact, Cexact, and Dexact (from exact decomposition) r(I0_dif_exact) difference between GE(0) at endyear and GE(0) at startyear r(I0_t1) GE(0) in startyear r(I0_t2) GE(0) in endyear r(N_t1) observations in startyear r(N_t2) observations in endyear {marker examples}{...} {title:Examples} {phang}{cmd:. msdeco marketIncome, groupvar(sector) yearvar(year) startyear(1975) endyear(2015)}{p_end} {phang}{cmd:. msdeco marketIncome [w=hwtsupp], groupvar(sector) yearvar(year) startyear(1975) endyear(2015) preserve}{p_end} {marker author}{...} {title:Author} {pstd} Andrew Silva {break} The Graduate Institute of International and Development Studies {marker references}{...} {title:References} {pstd} Jenkins, Stephen P. 1999. "Analysis of income distributions." {it:Stata Technical Bulletin} 8.48. {pstd} Jenkins, Stephen P. 2015. "INEQDECO: Stata module to calculate inequality indices with decomposition by subgroup." Statistical Software Components. {pstd} Mookherjee, Dilip. and Shorrocks, Anthony. 1982. "A decomposition analysis of the trend in UK inequality." {it:Economic Journal} 92: 886{c -}992. {marker seealso}{...} {title:See also} {pstd} {help ineqdeco} if installed.