Is Democracy Good for the Poor?
American Journal of Political Science, 2006 · paper · Political science
Reexamines whether democracy improves poor people's material well-being, arguing earlier tests suffered from sample bias. After addressing this, the paper finds no evidence that the poor benefit from democratic government, and speculates on reasons.
Checked: PDF from fsi9-prod.s3.us-west-1.amazonaws.com, fetched 5 Oct 2026. Page numbers are those of that PDF.
18 findings
Substance Text says the first-differences model "employs 150 of the 169 states – about 89 percent coverage" and that "DEMOCRACY and DEMYRS tests include 150 states", but Table 2 reports 149 observations in columns 1, 2, 3 and 5.
- p. 24 Results, para 1
employs 150 of the 169 states – about 89 percent coverage
- p. 25 Democracy subsection, para 2
DEMOCRACY and DEMYRS tests include 150 states
- p. 43 Table 2, Observations row
149
- p. 25 Democracy section
the POLITY specification incorporates 127 states, while the DEMOCRACY and DEMYRS tests include 150 states
Judge's note: Viewed p.24, p.25 and Table 2 on p.43: the Observations row reads 149 149 149 126 149. A cross-sectional first-differences regression has one observation per state. Table 5 FD (149) and p.30 ('covers just 149 out of 169') also give 149. The paper does not explain the extra state. 89% matches 150/169, while 149/169 = 88%.
Second check: another model agreed
Report a problem with this finding
Substance Text says "the POLITY specification incorporates 127 states" (and "the same 127 states as POLITY"), but Table 2 column 4 (POLITY) reports 126 observations.
- p. 25 Democracy subsection, para 2
the POLITY specification incorporates 127 states
- p. 43 Table 2, column 4 Observations
126
Only the authors can say what is right here.
Judge's note: Viewed p.25 and Table 2 on p.43: the POLITY column shows 126 observations. One observation per state in the first-differences cross-section, and no explanation is given. This is the same off-by-one pattern as C019 but a different number. The DEMYRS run on 127 states is not tabulated, so only the author can say which count is right.
Second check: another model agreed
Report a problem with this finding
Substance "when a state's Polity score is improved by a single standard deviation (about 6.9 ...), its child mortality rate should drop by a little less than one death per thousand" cannot be reproduced from the POLITY coefficient -.013 on log child mortality (Table 2 col 4).
- p. 25 Democracy subsection, para 4
its child mortality rate should drop by a little less than one death per thousand
- p. 43 Table 2 col 4 POLITY; Table 3 col 3 DEMOCRACY
-.069*
- p. 43 Table 2, column 4, POLITY
-.013*
Only the authors can say what is right here.
Judge's note: The dependent variable is the log of child mortality (Table 2 header). 6.9 × .013 ≈ .09 log points, about a 9% fall. At the paper's own mean rates (122 per thousand in 1970, 69 in 1999, p.12), that is roughly 6–11 deaths per thousand, not under one. No conversion is stated. The 'six-month boost' remark on p.26 builds on this figure. The author should state how the magnitude was derived.
Second check: another model agreed
Report a problem with this finding
Substance "the democracy's child mortality rate should be lower by slightly less than one death in a thousand" cannot be reproduced from the fixed-effects DEMOCRACY coefficient -.069* on log child mortality (Table 3 col 3).
- p. 25 Results, Democracy, para 4
the democracy’s child mortality rate should be lower by slightly less than one death in a thousand
- p. 43 Table 3, column 3, DEMOCRACY
-.069*
Only the authors can say what is right here.
Judge's note: Table 3's dependent variable is the log of child mortality, and DEMOCRACY is -.069 (.028). A -.069 log difference is about a 6.7% gap, roughly 5–8 deaths per thousand at the paper's mean rates of 69–122. No conversion to deaths is given. The Prais-Winsten estimation does not change how a coefficient on a log outcome is read. The author should state the derivation.
Second check: another model agreed
Report a problem with this finding
Substance Table 8 marks DEMOCRACY "Yes" (significant at p=.05) for CMR UNICEF with completed data, but Table 4 col 3 (completed data, UNICEF CMR) prints DEMOCRACY "-.062 (.040)" with no star, and the text says "none of the three democracy indicators approach statistical significance (Table Four)".
- p. 45 Table 8, CMR UNICEF row, DEMOCRACY column
Yes
- p. 44 Table 4, column 3 DEMOCRACY
-.062
- p. 26 Democracy subsection
none of the three democracy indicators approach statistical significance (Table Four)
- p. 25 Democracy, para 2
Both DEMOCRACY and DEMYRS reach statistical significance at the .10 level
Only the authors can say what is right here.
Judge's note: I viewed pp.44, 45, 26 and 27. p.19 states that the full tables are run with UNICEF child mortality data, so Table 4 and the CMR UNICEF row of Table 8 appear to be the same test. No different specification for the summary tables is stated anywhere. The text's 'one out of 15' count (p.27) rests on this cell. Only the authors can say which entry is right. This cluster also mentions the Table 6 issue, which I adjudicate under C039.
Second check: another model agreed
Report a problem with this finding
Substance Table 6 marks DEMOCRACY "Yes" (p=.05) for CMR UNICEF first differences, but Table 2 col 3 prints DEMOCRACY "-.115 (.062)" with no star, and the text says "Both DEMOCRACY and DEMYRS reach statistical significance at the .10 level, and POLITY achieves significance at the .05 level".
- p. 45 Table 6, CMR UNICEF row, DEMOCRACY
Yes
- p. 43 Table 2, column 3 DEMOCRACY
-.115
- p. 25 Democracy subsection, para 2
Both DEMOCRACY and DEMYRS reach statistical significance at the .10 level, and POLITY achieves significance at the .05 level
- p. 25 Democracy, para 3
DEMYRS is close (p=.053), while POLITY
Only the authors can say what is right here.
Judge's note: I viewed pp.43, 45, 25 and 27. Table 2 is the UNICEF CMR first-difference run (p.19 says the full tables use UNICEF data). In Table 2, POLITY is starred, which matches Table 6's 'Yes'. DEMOCRACY is not starred, yet Table 6 says 'Yes'. This cell enters the '14 of 30' count. I counted 6 'Yes' in Table 6 and 8 in Table 7, which gives the paper's 14. No alternative specification is stated. The Table 7 DEMYRS issue that this cluster also raises is adjudicated under C040.
Second check: another model agreed
Report a problem with this finding
Substance Table 7 marks DEMYRS "Yes" (p=.05) for CMR UNICEF fixed effects (biased data), but Table 3 col 4 prints DEMYRS "-.005 (.003)" with no star, and the text says "DEMYRS is close (p=.053)".
- p. 45 Table 7, CMR UNICEF row, DEMYRS
Yes
- p. 25 Democracy subsection, para 3
DEMOCRACY is significant at the .05 level and DEMYRS is close (p=.053)
- p. 43 Table 3, column 4 DEMYRS
-.005
- p. 26 Democracy, para 6
none of the three democracy indicators approach statistical significance (Table Four)
Only the authors can say what is right here.
Judge's note: I viewed pp.43, 45 and 25. Table 3 is the UNICEF CMR fixed-effects biased-data run. Table 7's POLITY 'No' and DEMOCRACY 'Yes' match Table 3, but its DEMYRS 'Yes' conflicts with the stated p=.053 under the .05 threshold given on p.27. This cell feeds the '14 of 30' count. The Table 8 issue that this cluster also mentions is adjudicated under C038.
Second check: another model agreed
Report a problem with this finding
Substance "having 30 years of communist governments during this interval would produce 1.3 additional child deaths per thousand" cannot be reproduced from the FD COMYRS coefficient .009* on log child mortality (Table 5).
- p. 27 Communism subsection, para 1
having 30 years of communist governments during this interval would produce 1.3 additional child deaths per thousand
- p. 44 Table 5, FD Test COMYRS
.009*
- p. 27 Results, Communism, para 1
would produce 1.3 additional child deaths per thousand
Only the authors can say what is right here.
Judge's note: Table 5 (p.44) has the dependent variable as log child mortality, with COMYRS .009 (.004) in the FD test. COMYRS is defined as years (p.20). 30 × .009 = .27 log points, about +31%, i.e. on the order of 20+ deaths per thousand at the paper's mean rates. 1.3 equals the multiplicative factor exp(.27)≈1.31, not a death count. No rescaling is stated.
Second check: another model agreed
Report a problem with this finding
Reference or label p.7 (PDF 9) cites "[Victoria et al. 2003; Lipton and Ravallion 1995; Sen 1998]", but the reference list has only Sen 1981, 1983 and 1999.
- p. 9 Section I, para 3
[Victoria et al. 2003; Lipton and Ravallion 1995; Sen 1998]
- p. 9 Section I, para 4
Sen 1998]
- p. 37 References
Sen, Amartya. 1999. Development as Freedom.
Only the authors can say what is right here.
Judge's note: PDF p.9 shows the citation. PDF p.37 lists Sen 1981, 1983 and 1999 only, and the text search finds no 1998 Sen entry. The year may be meant as 1999, or an entry may be missing; only the authors can say.
Second check: another model agreed
Report a problem with this finding
Reference or label p.10 (PDF 12) cites "[Lake and Baum 2004; ...]", but the reference list has only Lake and Baum 2001, which the text also cites as "Lake and Baum [2001]" on PDF p.13.
- p. 12 Section II, para 4
Lake and Baum 2004
- p. 36 References
Lake, David A., and Matthew Baum. 2001.
- p. 12 Section II, para 4
[Lake and Baum 2004;
- p. 9 Section I, indicators
Sen 1998
Only the authors can say what is right here.
Judge's note: PDF p.12 shows 'Lake and Baum 2004'. The only Lake and Baum entry on PDF p.36 is 2001 (Comparative Political Studies 34(6)), and the text search shows 'Lake and Baum [2001]' on p.13. The extra items this cluster mentions (Sen 1998, Navia and Zweifel 2003, McGuire 2004, Bollen) are judged in C005, C013 and C015.
Second check: another model agreed
Report a problem with this finding
Reference or label p.17 (PDF 19) cites "[Keefer and Vlaicu 2005; McGuire 2004]", but the reference list has only McGuire 2001; 'Navia and Zweifel [2003]' (PDF 26) and 'Musgrove 1996' (PDF 31) also have no reference entries.
- p. 19 Independent variables
[Keefer and Vlaicu 2005; McGuire 2004]
- p. 19 Independent Variables, para 3
McGuire 2004]
- p. 36 References
McGuire, James W. 2001.
- p. 9 Section I
Sen 1998
Only the authors can say what is right here.
Judge's note: PDF p.19 shows 'McGuire 2004', and the only McGuire entry on PDF p.36 is 2001. The text search finds no Navia and Zweifel 2003 entry (only Zweifel and Navia 2000) and no Musgrove entry; the Musgrove citation is visible on the PDF p.31 crop. Sen 1998 is judged in C005 and Bollen in C015.
Second check: another model agreed
Report a problem with this finding
Reference or label Footnote 14 (PDF 20) cites "Bollen [1993] and Bollen and Paxton [2000]", but the reference list has no Bollen entries.
- p. 20 Footnote 14
Bollen [1993] and Bollen and Paxton [2000] show
- p. 20 Footnote 14
Bollen [1993] and Bollen and Paxton [2000]
Only the authors can say what is right here.
Judge's note: The footnote is on the PDF p.20 image. The text search for 'Bollen' finds only the footnote, and the B section of the reference list has no Bollen entry.
Second check: another model agreed
Report a problem with this finding
Reference or label Footnote 14 says "See appendix for details on the coding rules." but the document (46 PDF pages: text, references, Figures 1–5, Tables 1–10) contains no appendix.
- p. 20 Footnote 14
See appendix for details on the coding rules.
Only the authors can say what is right here.
Judge's note: Viewed p.20 footnote 14. A search of the full text finds 'appendix' only in this footnote; the document ends with the references (pp.35–38), figures (pp.39–41) and Tables 1–10 (pp.42–46). There is no appendix with coding rules. The appendix may exist in another version of the paper; only the author can supply it.
Second check: another model agreed
Report a problem with this finding
Reference or label "Navia and Zweifel [2003] suggest the gap is 4.6 infant deaths per thousand" cites a work not in the reference list, which has only "Zweifel, Thomas D., and Patricio Navia. 2000."
- p. 26 Democracy subsection
Navia and Zweifel [2003] suggest the gap is 4.6 infant deaths per thousand
- p. 26 Democracy, para 5
Navia and Zweifel [2003] suggest the gap is
- p. 38 References
Zweifel, Thomas D., and Patricio Navia. 2000.
Only the authors can say what is right here.
Judge's note: Viewed p.26 and the reference list (p.38). There is no Navia & Zweifel 2003 entry. Elsewhere the paper cites 'Zweifel and Navia [2000]'. The Pande 2003 entry is a different author. Either the reference is missing or the citation key is wrong, and only the author can say which.
Second check: another model agreed
Report a problem with this finding
Reference or label Text says Figure Four shows "mean democracy score (1970-1999) on the Y axis, and the difference between their 1999 and 1970 child mortality rates (log values) on the X axis" (and Figure Five has "democracy scores ... still on the Y axis"), but the figures plot "Number of Democratic Years, 1970-99" on the X axis and the change in log child mortality (or residuals) on the Y axis.
- p. 29 Section V, para 1
showing their mean democracy score (1970-1999) on the Y axis, and the difference between their 1999 and 1970 child mortality rates (log values) on the X axis
- p. 41 Figure 4 axes
Number of Democratic Years, 1970-99
- p. 41 Figure 4 axes
Change in log of Child Mortality, 1970-99
- p. 29 Figure Five description
democracy scores are still on the Y axis, but the X axis now plots the residuals
Judge's note: Viewed p.29 and the figures on p.41. Both figures have democratic years on the horizontal axis and the outcome on the vertical axis. The democracy measure is a count of democratic years, not a mean score. The democracies sit at the right (x=30), not 'clustered together at the top'.
Second check: another model agreed
Report a problem with this finding
Reference or label Text says for Figure Five "democracy scores are still on the Y axis, but the X axis now plots the residuals", but Figure 5 plots "Regression Residuals" on the vertical axis and "Number of Democratic Years, 1970-99" on the horizontal axis.
- p. 29 Section V, para 3
democracy scores are still on the Y axis, but the X axis now plots the residuals
- p. 41 Figure 5 axes
Regression Residuals
Judge's note: I viewed p.29 (text, para 3) and p.41 (Figure 5). The Figure 5 y-axis title is 'Regression Residuals' and its x-axis title is 'Number of Democratic Years, 1970-99', the reverse of the text. Figure 4 has the same swap (its y-axis is the change in log child mortality), but that is a separate sentence and is not part of this cluster.
Second check: another model agreed
Report a problem with this finding
Reference or label "[Musgrove 1996;" is cited in the Conclusion (p.31) but the reference list has no Musgrove entry.
- p. 31 Conclusion, para 2
[Musgrove 1996;
Judge's note: I saw the citation on the p.31 image. In the reference list, the M entries run Mellinger, Meltzer, Moon, Moore and then go straight to Pande. There is no Musgrove entry.
Second check: another model agreed
Report a problem with this finding
Wording Table 6 note is truncated and ungrammatical: "“Yes” indicates the variable statistically significant (p=.05) when the".
- p. 45 Table 6 note
“Yes” indicates the variable statistically significant (p=.05) when the
- p. 45 Table 6 note
statistically significant (p=.05) when the
Only the authors can say what is right here.
Judge's note: The p.45 page image shows the note ending at 'when the', with blank space below, so this is in the published page and is not an extraction artefact. The condition is never completed, and 'is' is missing. Only the authors can supply the intended wording.
Second check: another model agreed
Report a problem with this finding
19 minor (typos and the like; most checked by the judge only)
Reference or label In-text citation "[Heston, Summers, and Aten 2002]" (also footnote 7) does not match the reference entry "Heston, Alan, Robert Summers, and Bettina Aten. 2004."
- p. 21 Control variables
[Heston, Summers, and Aten 2002]
- p. 35 References
Heston, Alan, Robert Summers, and Bettina Aten. 2004.
Judge's note: Viewed both crops. The in-text key is 2002, but the reference list entry's year is 2004 (its body says '2002 [cited February 1 2004]'). The source is clearly the same, but the author–date key is inconsistent.
Second check: another model agreed
Report a problem with this finding
Wording p.3 (PDF 5): "In 2001, over half of the world's population – some 2.73 billion people – lived on less than $2 a day"; 2.73 billion is under half of a world population of about 6.1 billion.
- p. 5 Introduction para 2
over half of the world’s population – some 2.73 billion people – lived on less than $2 a day
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: world population in 2001 was about 6.1 billion. The paper gives no population base. On the page image, the sentence reads as quoted. 2.73/6.1 is about 0.45, so 'over half' would hold only for a base below 5.46 billion. The source figure may refer to the developing-world population. Only the authors can say which base they meant.
Second check: another model agreed
Report a problem with this finding
Wording COMYRS is defined as "the number of years of communist rule since 1900" (PDF 20), but p.25 (PDF 27) reads it as years of communist rule "between 1970 and 2000" and then lists "the total number of communist years since 1900" as an alternative measure.
- p. 20 Independent variables
the second (COMYRS) is the number of years of communist rule since 1900
- p. 27 Communism subsection
Several alternative measures of communist rule – measuring the total number of communist years since 1900
- p. 27 Communism, para 2
measuring the total number of communist years since 1900, or taking the log value of communist years
Only the authors can say what is right here.
Judge's note: Both page images were checked. In a first-differences model, differencing COMYRS could explain the 1970–2000 reading, but the text does not say so. Listing the since-1900 count as an 'alternative' to COMYRS leaves unclear which variable entered Table Five. Only the authors can clarify.
Second check: another model agreed
Report a problem with this finding
Wording Text describes Figure Four as showing "their mean democracy score (1970-1999)", while the figure's axis is "Number of Democratic Years, 1970-99" (a 0–30 count).
- p. 29 Section V, para 1
mean democracy score (1970-1999)
- p. 41 Figure 4 x-axis
Number of Democratic Years, 1970-99
Judge's note: I viewed p.29 and p.41. The axis plots a count of democratic years. For a dichotomous democracy measure, the mean times 30 equals the number of years, so the two are equivalent up to scale, but the text and the label name different quantities. The next sentence in the text uses the count (15 years). This is a minor wording inconsistency.
Second check: another model agreed
Report a problem with this finding
Typo p.10 (PDF 12): "Ghorbarah, Huth, and Russett 2004", but the reference list spells the name "Ghobarah, Hazem Adam".
- p. 12 Section II, para 4
Ghorbarah, Huth, and Russett 2004
- p. 35 References
Ghobarah, Hazem Adam
- p. 31 Conclusion
disturbingly little little correlation
- p. 12 Previous research
This arguments are consistent
Judge's note: The text spelling is on PDF p.12 and the reference spelling on PDF p.35. Other typos bundled in this cluster are also on the page images: 'disturbingly little little correlation' (PDF 31) and 'catastropic outcomes' (PDF 29). 'Iverson' is spelled the same way in the text and the reference list, so whether it is wrong rests on outside knowledge. 'This arguments', 'that'/'than' and the missing 'for' are judged in C009, C002 and C003.
Second check: none (judge only)
Report a problem with this finding
Typo p.4 (PDF 6): "Yet this example is less illuminating that it first appears"; 'that' should be 'than'.
- p. 6 Introduction, para 2
less illuminating that it first appears
- p. 6
Yet this example is less illuminating that it first appears
Judge's note: Seen on the page image of PDF p.6.
Second check: none (judge only)
Report a problem with this finding
Typo p.4 (PDF 6): "If democracy produces better outcomes those in the bottom deciles"; the word 'for' is missing before 'those'.
- p. 6 Introduction, para 3
produces better outcomes those in the bottom deciles
Judge's note: Seen on the page image of PDF p.6.
Second check: none (judge only)
Report a problem with this finding
Typo p.4 (PDF 6): "Bharitiya Janata Party"; the standard spelling is 'Bharatiya'.
- p. 6 Introduction, para 1
Bharitiya Janata Party
Judge's note: Rests on outside knowledge: the party's standard English spelling is 'Bharatiya'. The misspelling is visible on the page image of PDF p.6.
Second check: none (judge only)
Report a problem with this finding
Typo p.10 (PDF 12): "Meltzer and Richards' [1981] model", but the reference list has "Meltzer, Allan H, and Scott F Richard. 1981."
- p. 12 Section II, para 5
Meltzer and Richards’ [1981] model
- p. 12 Section II, para 5
This arguments are consistent with Meltzer and Richards’ [1981]
- p. 36 References
Meltzer, Allan H, and Scott F Richard. 1981.
- p. 12 para 3
This arguments are consistent with Meltzer
Judge's note: The text is on PDF p.12 and the reference entry on PDF p.36. The 'This arguments' part of this cluster is judged in C009.
Second check: none (judge only)
Report a problem with this finding
Typo p.10 (PDF 12): "This arguments are consistent with..."; this should read 'These arguments'.
- p. 12 Section II, para 5
This arguments are consistent
Judge's note: Seen on the page image of PDF p.12.
Second check: none (judge only)
Report a problem with this finding
Typo Footnote 9 (PDF 16): "these two variables reached statistical significant in some specifications"; 'significant' should be 'significance'.
- p. 16 Footnote 9
reached statistical significant
- p. 16 footnote 9
reached statistical significant in some specifications
Judge's note: Seen on the crop of PDF p.16.
Second check: none (judge only)
Report a problem with this finding
Typo p.16 (PDF 18): "One drawback of the fixed-effects model is that is that it requires"; the words 'is that' are duplicated.
- p. 18 Model, para 2
is that is that it requires
- p. 18 Model
drawback of the fixed-effects model is that is that it requires
- p. 18
One drawback of the fixed-effects model is that is that it requires
Judge's note: Seen on the crop of PDF p.18.
Second check: none (judge only)
Report a problem with this finding
Typo p.16 (PDF 18): "only produce data intermittenly"; this should be 'intermittently'.
- p. 18 Model
only produce data intermittenly
- p. 18
Since many countries only produce data intermittenly
Judge's note: Seen on the crop of PDF p.18.
Second check: none (judge only)
Report a problem with this finding
Typo Footnote 16: "than then first-differences model" should read "than the first-differences model".
- p. 22 Footnote 16
than then first-differences model
Judge's note: Typo visible in the footnote 16 crop on page image p.22 (printed page 20).
Second check: none (judge only)
Report a problem with this finding
Typo Text refers to "INCOME" ("both INCOME and LOGHIV are highly significant"), while the variable is defined and tabulated as LOGINCOME.
- p. 24 Results, paras 2–3
both INCOME and LOGHIV are highly significant
Judge's note: Viewed p.24: 'INCOME' appears twice. The variable is defined as LOGINCOME on p.21, and Tables 2–5 use LOGINCOME. This is an inconsistent label only.
Second check: none (judge only)
Report a problem with this finding
Typo "The greater statistical significance of POLITY may an artifact of a smaller sample" is missing "be".
- p. 25 Democracy subsection, para 2
POLITY may an artifact
Judge's note: Visible on page image p.25, line 5 of the Democracy discussion.
Second check: none (judge only)
Report a problem with this finding
Typo Misspelling "Hence to measure catastropic outcomes" (should be 'catastrophic').
- p. 29 Section V, para 4
catastropic outcomes
- p. 29 Section V, first line
Hence to measure catastropic outcomes
- p. 29 first line
to measure catastropic outcomes
Judge's note: The misspelling is visible in the first line of the p.29 page image. Elsewhere (p.27) the paper spells the word correctly as 'catastrophic'.
Second check: none (judge only)
Report a problem with this finding
Typo Duplicated word: "there is disturbingly little little correlation between public spending".
- p. 31 Conclusion, para 2
disturbingly little little correlation
- p. 31
disturbingly little little correlation between public spending
- p. 31
n fact, there is disturbingly little little correlation between
Judge's note: The p.31 page image shows 'little little'.
Second check: code check
Report a problem with this finding
Typo Reference title printed as "Counting the Worlds Poor Problems and Possible Solutions": the apostrophe in 'World's' and the colon before the subtitle are missing.
- p. 34 references
Counting the Worlds Poor Problems and Possible Solutions
Judge's note: The p.34 page image shows the title without an apostrophe or colon. The fix is evident from ordinary English punctuation.
Second check: none (judge only)