You are on page 1of 20

Inflation and the Poor Author(s): William Easterly and Stanley Fischer Source: Journal of Money, Credit and

Banking, Vol. 33, No. 2, Part 1 (May, 2001), pp. 160-178 Published by: Blackwell Publishing Stable URL: Accessed: 31/07/2009 03:32
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact

Blackwell Publishing is collaborating with JSTOR to digitize, preserve and extend access to Journal of Money, Credit and Banking.


Inflationand the Poor

Using polling data for 31,869 households in thirty-eightcountries and allowing for country effects, we show that the poor are more likely than the rich to mention inflationas a top national concern. This result survives several robustnesschecks. We also find direct measures of improvementsin well-being of the poor the change in their share in national income, the percent decline in poverty, and the percent change in the real minimum wage to be negatively correlatedwith inflationin pooled cross-countrysamples. THE CLAIMthat "inflationis the cruelest tax of all" is often interpretedas meaning that inflationhurtsthe poor relatively more than the rich. It could also mean thatthe inflationtax is particularly unfairbecause, the taxing mechanism being little understood,the inflationtax can be imposed by stealth. The essential a prioriargument is thatthe rich arebetterable to protectthemselves against, or benefit from, the effects of inflationthan are the poor. In particular, the rich and more sophisticatedare likely to have better access to financialinstruments thathedge in some way againstinflation,while the (small) portfolios of the poor are likely to have a largershareof cash. The poor may also dependmore thanthe rich on state-determined income that is not fully indexed to inflation. Among the elderly poor, pensions are often not fully indexed and so inflation will directly reduce their real incomes. For the remainderof the poor, state subsidies or direct transfersmay also not be fully indexed. However,these argumentsare not decisive. Aside from the points thatthe poor are likely to hold relativelymore cash in theirportfolios, andto be less sophisticated,the relativeeffects of inflationon the rich versus the poor must be specific to the institutions and histories of each economy. Certainly,study of the long list of the potential effects of inflationon the economy outlined in Fischer and Modigliani (1978) does not lead to a clear presumptionthatit is the poor who are hurtrelativelymore by in-

This paper was prepared for the Annual World Bank Conference on Development Economics (ABCDE), April 1999. Views expressedhere are not necessarily those of the WorldBank or the International Monetary Fund. The authors are grateful for the diligent research assistance of Claire Hughes Adams and for the comments of our discussant, MartinRavallion, other ABCDE participants,and two referees.

WILLIAM EASTERLY is senior adviser at the WorldBank. STANLEY FISCHER is with the InternationalMonetaryFund. Correspondence to WEasterly@WorldBank.Org. Journal of Money,Credit,and Banking,Vol. 33, No. 2 (May 2001, Part 1) Copyright2001 by The Ohio State University


flation, especially because so many of the effects of inflationcome throughcomplicated details of the tax system, including capital taxation.The question must be an empiricalone, and the answermay well differ among economies. In this paper, we examine inflation's effects on the poor in two ways. First, we drawon the results of a global surveyof 31,869 individualsin thirty-eightcountries, which asked whether individuals think inflation is an importantnational problem. This providesan indirectway at getting at the issue of whetherinflationis more of a problemfor the poor than for the rich. Second, we assess the effects of inflationon direct measures of inequality and poverty in various cross-countryand cross-time samples. Ourevidence supportsthe views thatinflationis regardedas more of a problemby the poor thanit is by the nonpoor,and thatinflationappearsto reducethe relativeincome of the poor. It thus adds to a growing body of literaturethat on balance but not unanimously tends to supportthe view that inflationis a cruel tax. We startby reviewingthe literature,and then turnto the new evidence.

Most of the literaturedeals with the United States, using annual data on poverty rates and inflation.Powers (1995) finds that inflationworsens a consumption-based poverty measureover 1959-92, but has no significantimpact on the income-based povertyrate.Cutlerand Katz (1991), in contrast,find that an increasein inflationreduces the povertyrate over 1959-89. Blank and Blinder (1986) found that inflation increasedpovertyrates, but also slightly increasedthe income shares of the bottom two quintiles (only the second quintile was significant). On balance, Blank and Blinderarguethat "thereis little or no evidence thatinflationis the cruelesttax." Moving to other countries, Cardoso (1992) argues that the inflation tax does not affect those alreadybelow the povertyline in LatinAmericabecause of theirnegligible cash holdings. However, she finds that higher inflationis associated with lower real wages in a panel of seven LatinAmericancountries.An additionalfragmentof evidence comes from Rezende (1998, p. 568), who points out that the Gini coefficient in Brazil increasedsteadily with rising inflationin the 1980s and then declined with the successful inflation stabilizationof 1994-1996. Datt and Ravallion (1996) found in a cross-time, cross-statestudy of India that observationswith higher inflation rates also had higherpovertyrates. Romer and Romer (1998) arguethat the effects of inflationon the incomes of the poor are likely to differ between cyclical and longer-termperspectives.In the short inflation will be associated with a decline in unrun, an increase in (unanticipated) employment,that may well relatively benefit the poor. Over the longer term, however, higher inflation cannot permanentlyreduce unemployment,and the effects of inflationon the poor could then be reversed.Even in a cyclical perspective,Romer and Romer find the effects of unemployment on the income distribution to be strongerin earlier decades than in the nineties. Using an internationalpanel, they


find that lower inflation tends to increase the income of the poor over the longer term a resultthey attribute in partto the negative associationbetween inflationand economic growth.Agenor (1998) also finds poverty rates to be positively relatedto inflationin cross-countrydata. In our work using polling data,we will explore the impacton attitudesto inflation of factorsotherthanrelativeincome. The poor are less educated,andtheremay be an independenteffect of inflation'simpact on the uneducated.Ourpriorson the impact of educationon attitudesto inflationare, like those on income, ambiguous.One considerationis that humancapitalmay be a good hedge againstinflation,so those with more humancapitalfeel more protected(also stocks and bonds may be good hedges against inflation and they are also held disproportionately by the more educated). The uneducatedprobablyhave a lower weight of human capital relative to cash in their portfolios, and so dislike inflation more. But the more educated may know more aboutthe damagethat inflationcan do to the economy as a whole and so may be more likely to mentioninflationas a top concernthanthe less educated. Previous literatureusing polling data includes Fischer and Huizinga (1982), who analyzed the relative probabilitiesof mentioning inflation and unemploymentas a (or the most) serious problemfacing the nation, in the United States over the period 1939-78. They found that inflationwas consistentlymore frequentlycited as a serious problemthan unemploymentexcept duringrecessions. Apropos the question in this paper,they found a positive association between income and the probabilityof mentioning inflation as a serious problem ("inflationaversion"),although the relationship was sometimes nonmonotonic. Moreover, in regression analysis income was positively but insignificantlyrelatedto inflationaversion.Rose (1997) found no association between the standardof living and inflation aversion relative to unemploymentaversionin a sample of polling datafrom ex-Communistcountries. Fischerand Huizinga(1982) also found little relationshipbetween the level of education and inflationaversion.However,their educationvariablediscriminatedonly between high school educationand above. We will control for the national averages of inflation aversion when testing the poor's relativeinflationaversion.On the cross-sectionrelationshipbetween inflation aversion and actual inflation,Fischer (1996) found a surprisinglyweak correlation using the same survey data that we use in this paper.Likewise, Rose (1997) found little association among transitioncountries between actual inflation and inflation aversion inflationaversionrose relative to unemploymentas inflationwas falling. The Czech Republicwith its low inflationhad higherinflationaversionthanUkraine and Belarus with their quadruple-digit inflation (although causality is importantthe Czechs' inflationaversioncould be the reasonthey have low inflation).However, Fischer and Huizinga (1982) did find that the cross-time variation in the United States of mentioninginflationor unemploymentas the most seriousproblemwas associated with actualinflationand unemployment. Shiller (1996) poses a question closely related to ours, "WhyDo People Dislike Inflation?" He conducteda questionnairesurvey of 677 people in the United States, Germany,and Brazil. His answer was that people perceived inflation as reducing



their standard of living. In the U.S. sample, when asked what was their biggest concern aboutinflation,77 percentof the sample chose the response "inflationhurtsmy real buying power."Only 7 percent chose the traditionalview of economists "inflation causes a lot of inconveniences:I find it harderto comparisonshop, I feel I have to avoid holding too much cash, etc."When pressed further,the majorityin the samples in the United States, Germany,and Brazil supportedthe view that their wages would not rise as fast as the price level during the process of inflation. If Shiller's results indeed reflect most people's view of inflation,than we might expect the poor and uneducatedto dislike inflationmore because they areprobablyless protected by asset income from changes in their real wages. We will find some support for the idea that inflationreducesthe real wages of the poor in our empiricalresults.

2.A TheData Roper StarchWorldwide,a marketing,public opinion, and advertisingresearch firm,coordinatedthe surveythat we use to measureinflationconcerns.International ResearchAssociates (INRA) did the actual field work with its affiliates and partner companies. The survey was undertakenby Roper Starch during Februaryto May 1995. Table 1 lists the thirty-eightcountries nineteen industrialized,and nineteen developing and transition covered in the survey. Respondentsto the survey from all countries were classified according to their standardof living (self-assessed) and level of education. The survey question on which we focus is: Hereis a list of thingspeoplehavetoldus theyareconcerned about today. Would you readoverthe list andthentell me whichtwo or threeyou personally aremostconcerned about today. The economic concerns included in the list were "recessionand unemployment, inflationand high prices, money enough to live right and pay bills, educationalquality."There were fourteenother noneconomic concerns, and respondentscould also say "other,none of these, don't know."We define a dummy variablethat takes the value 1 if people mention"inflationandhigh prices"amongthe top two or threeconcerns (the top two or three are not rankedamong themselves), and 0 otherwise. The wording of the inflationresponse is unfortunate in that it also includes "high prices.''1 It is unclearhow the respondentwill interpret "highprices" will it be high prices comparedto the past or high prices comparedto the respondent'swage? If the latter,then the respondentmay simply be complainingaboutlow real wages. Fortunately, there is another"top concern"that directly addressesthe standardof living, which is "money enough to live right and pay bills." The correlationamong all respondentsbetween these two "topconcerns"was only .0043, with a p-value of .437. Hence, we can be moderatelyreassuredthatthe "inflationand high prices"question
1. It is not uncommonin such polls for "inflationand high prices"to be classed togetheras one issue.



Inflation & Hioh Prices



Recession/ Unemplmnt

Druo Abuse

Govn't Education ColTuption quality


RaciaV Ethnic Relations


Envi Po

Australia Austria Belgium Brazil Canada Chile China Colombia Czech Republic Denmark Hlnana France Germany Greece tiong song Hungary India
. 0 . .


Ireland Italy Japan Mexico Netherlands Norway Philippines Poland

4 7 10 7 10 5 25 9 14 3 4 4 8 9 11 19 13 12 4 6 6 17 3 2 12 11

17 16 13 17 18 10 13 15 26 12 21 12 21 13 12 17 14 24 18 10 9 13 23 19 24 22

6 10 9 14 8 17 1 14 7 7 3 19 6 14 7 2 7 9 10 12 6 7 6 2 4 6



17 8
10 11


24 2I 16 20 14

14 24

22 20 18 12 12 12 12

6 8 10 9 6 19 1 10 7 3 5 6 7 15 6 2 7 11 17 2 2 7 7 11 10 7

8 6 5 8 10 9 11 4 8 9 16 6 6 2 8 17 7 2 10 7 4 7 7 10 4 11

8 6 12 14 6 7 15 10 11 3 9 6 6 6 3 9 9 5 6 13 21 15 5 4 10 11


6 6

4 2

7 13 12

1 1 1

o o o


4 3 6

8 2 4 4

2 6

3 4
o 1 1


3 2 2

1 1

6 3 3 6

3 3
o 1

1 1


4 4
o o

7 4
o 1

TABLE 1 (Continued)
Money enouah to live on

Inflation & High Prices



Recession/ Unemplmnt

Drug Abuse

Govn't Education Corruption quality


RaciaV Ethnic Relations

Env P

llussia Singapore Spain Sweden Switzerland Taiwan Thailand Turkey Ukraine United Kingdom USA Venezuela Sample average

22 23 7 5 6 11 5 19 22 5 7 16 10

28 12 6 20 15 13 23 8 26 19 24 16 17

1 4 12 5 11 4 20 5 2 4 12 10 8

13 9 18 15 15 8 12 12 9 14 6 11 14

1 4 13 11

15 12 9 3

8 2 10 4

3 1 1

11 10 4 3 9 11 11 8

4 1 5 19 11 10 5 8

18 8 11 10 6 6 13 9

14 4 3 3 12 7 7 2

6 4
1 1 1 o

2 6 6
1 o

7 12

3 3

3 2 2 4


is really aboutinflationand not aboutreal wages. Fischerand Huizinga(1982) found no differencein poll responses in the United States to questions that mentionedjust and those thatmentioned"inflationand high prices." "inflation" The income questionon the surveyaskedthe respondentsto classify themselves in one of seven categories:"rich,very comfortable,comfortable,average,just getting by, poor, and very poor."Thus participantsare self-classifying on this question, and we should thereforeinterpretthe answers as relating to the relative income of the in his or her own country.We define dummy variablesfor each category participant that take the value 1 if the respondentsself-classify in that category and zero otherwise. Similarly the education question asked the respondentsto put themselves in higher." or less, secondary/technical, one of the following threecategories:"primary We again code three dummy variablesfor each category.We will also include country dummies in our regressions, and will later review them as indicatorsof the underlyingsensitivityto inflationin each country. Table 1 shows summarystatistics on the poll responses in each country.The percentages for each problemx are the numberof total responses that mentionedx as among the top two or three national problems, where each respondenthas two to three responses. The averageacross nations is for 10 percent of the responses to be accountfor more responses on Only crime and recession/unemployment "inflation." average. 2.B Income and EducationResults We do a probit equation, with the dependentvariableequal to one if inflation is mentioned as among the top two or three national concerns. The independentvariables are the income category dummies and the education category dummies. The resultsare shown in Table2. The category"rich"is omittedfrom the specification,so the coefficients on the income variablesmeasure the difference between the coeffiTABLE

ESTIMATED PROBIT EQUATION FOR MENTIONING INFLATIONAND HIGH PRICES AS A ToP NATIONAL CONCERN Observations: 31,869 Estimate Parameter CollStallt -1.09 Error Standard t-statistic P-value


-7.95 0.25 1.17 1.91 2.11 2.30 2.39 5.22 2.79

[.000] [.801] [.240] [.057] [.035] [.022] [.017] [.000] [.005]

Standardof living of individual("Rich" is omittedcategory) 0.14 0.03 Very Comfortable 0.13 0.15 Comfortable 0.13 0.25 Average 0.13 0.28 JUSt Getting By 0.14 0.31 Poor 0.15 0.36 Very Poor Educationalattainment("Higher education" is omittedcategory) 0.02 0.13 Primary School 0.02 0.06 Secondary School
NOTE: Countryinterceptdummies are includedbut not shown.


cient on that income category and "rich." Likewise, the category "highereducation" is omitted, so the coefficients on the educationvariablesmeasurethe differencebetween that categoryand "highereducation." Table 2 shows the results. (Individualcountryeffects are not shown at this point; they will be discussed below.) The likelihood of mentioninginflation as a top concern is decreasing in the standardof living of the respondent.The coefficient increases monotonicallyas respondentsrangefrom "verycomfortable" to "verypoor." The coefficients on "justgetting by,""poor," and "verypoor"are all statisticallysignificant,meaning that the differencebetween those categories and "rich"is statistically significant.The significanceis not overwhelminggiven the large sample, but it does pass the common statisticalthreshold.The very poor have a 10.5 percenthigher probabilityof mentioning inflation as a top concern than do the rich. The poor are thus relativelymore concernedthanthe rich aboutinflation. The patternfor the educationvariableis similar:the less educateddislike inflation more thanthe more educated.The differencebetween those who have a primaryeducationor less andthose with highereducationis highly significantstatistically, though not absolutelylarge.The coefficientimplies thatthose with only a primaryeducation have a 3.8 percenthigherprobabilityof mentioninginflationas a top concernthando those with highereducation.2 Those with a secondaryeducationare also significantly more likely to mention inflationas a top concern than those with higher education. Recalling the possibly offsettingeffects of humancapitalas a hedge againstinflation and the greaterkrlowledgeof inflation'sdamagewith highereducation,as factorsaffecting the response,our resultssuggest thatthe firsteffect dominatesthe second. 2. C RobustnessChecks Our first robustnesscheck is to split the sample between developing and developed countries.Table 3a shows that the results are still very strong in the industrial countrysample, but Table 3b shows much weakerresults in the developing country sample. The magnitudesof the coefficients are uniformly lower in the developing countrysample than in the industrialcountrysample. In the industrialcountrysample, the very poor have a 14 percenthigher probabilityof mentioninginflation as a top concern than the rich. In the developing countrysample, the very poor have a 9 percenthigherprobabilitythanthe rich. The weakness of the developing countryresults may have to do with collinearity in discriminatingamong the finely defined income categories. When we aggregate the bottom two categories as "lower class," the middle three categories as "middle class,"andthe top two categoriesas "upperclass,"we get statisticallysignificantdifferences in the developing countrysample between "lowerclass" and "upperclass," and between "middleclass" and "upperclass" (Table4). These differences are also significantin the industrialsample. The coefficient on "lowerclass" continues to be higher in the industrialsample thanin the developing countrysample.
2. The coefficient estimates are not the same as the marginalprobabilities,which vary with the righthand-sidevariables.The marginalprobabilitiesreportedhere are at the sample means.


TABLE3A (16,352 observations;countryeffects includedbut not shown) RESULTS FOR INDUSTRIAL ECONOMIES Parameter Estimate Constant -1.25 Standald Error 0.23 0.23 0.22 0.22 0.22 0.23 0.26 t-statistic -5.56 0.69 1.15 1.51 1.40 2.13 2.29 7.97 5.74 P-value [.000] [.391] [.184] [.022] [.039] [.004] [.004] [.000] [.000]

Standardof livi7zg ("Rich" is omittedcategory) 0.16 VeryComfortable 0.26 Comfortable 0.33 Average 0.31 Just GettingBy 0.49 Poor 0.59 VeryPoor

Educationallevel ("Higher education" is omittedcategozy) 0.29 0.04 PrimarySchool 0.18 0.03 SecondarySchool

TABLE3B DEVELOPING COUNTRIES (15,517 observations;countryeffects includedbut not shown) Parameter Estimate Constant -1.20 Standard Error 0.17 0.17 0.16 0.16 0.16 0.17 0.18 t-statistic -7.01 -0.23 0.61 1.28 1.65 1.50 1.44 0.01 -1.20 P-value [.000] [.822] [.539] [.202] [.099] [.134] [.150] [.993] [.230]

Standardof living ("Rich" is omittedcategory) -0.04 VeryComfortable 0.10 Comfortable 0.21 Average 0.27 Just GettingBy 0.25 Poor 0.27 VeryPoor

Educationallevel ("Higher education" is omittedcategory) 0.00 0.03 PrimarySchool -0.03 0.03 SecondarySchool

The education variables are not robust across the two samples. They are still highly significantin the industrialcountrysample,but are always insignificantin the developing countrysample. Our second robustnesscheck is to also include age and occupationalgroups.The seven age groups are 14-20 (the omitted category), 21-29, 30-39, 40-49, 50-59, 6069, and 70 and over. The occupationalcategories are student(the omitted category), professional/executive, white collar, blue collar, unemployed, homemaker, and retired.Table5 shows the results. The resultson povertyand educationarerobustto the inclusion of age groupdummies and occupationalgroupdummies.The poor and very poor are still significantly more likely than the richer to mention inflationas a top concern. Primary-educated respondentsare still more likely to cite inflation as a conand secondary-educated cern than those with higher education. All of the age groups are more likely to be concerned about inflation than teenagers.The age groupmost concernedwith inflationis thatof people in their six-



RESULTS WITH AGGREGATED INCOMECLASSES Parameter Estimate Standard Error t-statistic P-value

Industrial countries Constant


0.08 0.06 0.09

-14.65 2.52 4.07 8.36 5.99

[.000] [.012] [.000] [.000] [.000]

Income Class ("Upper Class" is omittedcategory) Middle Class 0.15 Lower Class 0.36

Educationallevel ("Higher education"is omittedcatego7ey) PrimaryEducation 0.30 0.04 SecondaryEducation 0.19 0.03

Parameter Estimate

Standard Error



Developing countries Constant


0.08 0.07 0.06

-16.28 3.69 3.91 0.83 -0.62

[.000] [.000] [.000] [.406] [.537]

Income Class ("Upper Class" is omittedcategory) Middle Class 0.26 Lower Class 0.23

Educationallevel ("Higher education" is omittedcatego7ey) PrimaryEducation 0.03 0.03 SecondaryEducation -0.02 0.03

ties, followed closely by people in their seventies and above (the differencebetween the sixties and seventies is not statisticallysignificant).This group is at the stage in the life cycle of consumingby runningdown their assets, and so may dislike the uncertaintyintroducedby inflation. The occupationalgroupmost concernedwith inflationis blue-collarworkers.This reinforcesthe finding that those who are more averse to inflation are relatively disadvantagedon several differentdimensions the poor, the uneducated,and the unskilled (blue-collar)workers. We also tried a gender dummy.Males were slightly more likely to mention inflation as a problemthan females, but the differencewas not statisticallysignificant. 2.D Other Concerns We also examined what other economic concerns the poor had, to see how their concern with inflation compares to other problems.Table 6 shows which concerns are disproportionately and significantlymore likely to be mentionedby the poor.The result on unemploymentand recession is very surprising the poor are more likely to mentionit as a problemthanthe rich, but the differenceis not significant. Otherconcernsfollow a more predictablepattern.The poor are much more likely thanthe rich to mention "moneyenough to live right"as a concern,not surprisingly. The less educatedarepredictablymuch less likely to mentionqualityof educationas a concernthanthe more educated.

TABLE5 ROBUSTNESS TO AGE AND OCCUPATION Observations: 31,443; Countrydummies are includedbut not shown Parameter Estimate Constant -1.34 Standard Error 0.14 0.14 0.13 0.13 0.13 0.14 0.15 t-statistic -9.43 0.31 1.12 1.77 1.90 2.04 2.13 3.69 3.01 4.84 5.66 4.34 4.87 6.03 4.60 1.50 1.48 2.37 1.04 1.55 1.15 P-value [.000] [.755] [.264] [.076] [.057] [.041] [.033] [.000] [.003] [.000] [.000] [.000] [.000] [.000] [.000] [.133] [.140] [.018] [.300] [.121] [.250]

Standardof living ("Rich" is omittedcatego7y) VeryComfortable 0.04 Comfortable 0.15 Average 0.23 JustGettingBy 0.25 Poor 0.28 VeryPoor 0.32

Educationalattainment("Higher education"is omittedcategory) PrimarySchool 0.10 0.03 SecondarySchool 0.06 0.02 Age Groups(14-20 age group is 07nitted categosy) People in their20s 0.17 People in their30s 0.21 People in their40s 0.17 People in their50s 0.20 People in their60s 0.28 People in their70s+ 0.26 OccupationalGroltps("Student"is omittedcategosy) Professional/Executive 0.06 White Collar 0.06 Blue Collar 0.09 Unemployed 0.05 Homemaker 0.06 Retired 0.06 0.03 0.04 0.04 0.04 0.05 0.06 0.04 0.04 0.04 0.05 0.04 0.05

TABLE6 WHAT OrHER CONCERNS Do THE POOR HAVE? t-statistic onincome oreducational level inprobit regression for mentioning concern shown: Inflation and highprices Recession and unemployment Money enough tolive right, paybills 0.05 1.43 2.88 4.89 5.27 5.81 Educational quality 0.86 0.50 -0.41 -0.47 -0.94 0.54 -17.66 -11.85 crime -0.29 -0.55 -0.35 -1.43 -2.07 -1.14 4.42 0.97

Standardof living of individual("Rich" is omittedcategory) VeryComfortable 0.25 -0.52 Comfortable 1.17 0.26 Average 1.91 1.03 Just GettingBy 2.11 1.34 Poor 2.30 1.51 VeryPoor 2.39 1.04

Educationalattainment("Higher education" is omittedcategosy) PrimarySchool 5.22 -0.43 6.01 SecondarySchool 2.79 -0.05 4.04


On a questionwhere our priorswere not so clear,we found thatthe poor were less likely than the rich to mention crime as a concern.The differencewas not very significant,however.Confusingthe pictureon crime further,the primary-educated were more likely to mention crime thanthe college-educated.3 We also tried the gender dummy in the regressions for the other economic concerns. The only significantresults were that females were more likely than males to mention "moneyenough to live right"and "educationquality"as concerns.
2.E Country EJjcects

Althoughnot directlyrelevantto our mainquestion,the patternof countryeffects is interesting[as was previously noted by Fischer (1996)]. Figure 1 graphs the countries'propensityto mentioninflationas a top nationalproblem(fromTable 1) against the actualinflationratein the decadeprecedingthe survey(1985-94). (We would get a very similarpictureusing the countrydummiesfrom the regressionin Table 2 for inflationpreferences.)Although there is a significantpositive relationshipbetween the log of averageinflation1985-94 andthe countrypropensitiesto mentioninflation as a top problem,there are some strikingoutliers.The countryin which respondents were, ceteris paribus, most concerned about inflation is not Ukraine, Russia, or Brazil in all of which the inflationrateshortlybefore the poll had been around1000 percentor higher but is China with its modest inflationrate of 12 percent.An even more strikingoutlier is Singapore,where the likelihood of mentioninginflationas a top problemis similarto those of Russia and Ukraine,even thoughinflationwas only 2 percent,which tied with Japanand the Netherlands for lowest inflationin the entire sample!Since Singapore'spopulationis largelyof Chinesedescent,we speculatethat thereis a Chinese dummyvariable.This Chinese variablemay have somethingto do with the memoryof the hyperinflation in China afterWorldWarII. (The Taiwanand Hong Kong dummiesare also at least weakly consistentwith the Chinesedummyhypothesis.)It is interestingthatanothercountrythathad a post-WWII(andpost-WWI) hyperinflation Hungary also displaysa high tendencyto mentioninflationas a top problemrelativeto a modestrecentinflation.Perhapssurprisingly, the observation for Germanyis not far out of line with the average. Outliers in the other direction are Brazil and Chile. Brazilians are a little less likely to mention inflation as a top problem than Americans, despite having had around1000 percentinflationin Brazil over the previous decade. It could be argued that since the survey was conductedin early 1995, Braziliansmay have alreadyincorporatedfavorable expectations about the success of the stabilizationplan (the Plan Real) introducedin mid-1994. Brazilians may also have been relatively well protectedfrom inflationby indexation but the election results in Brazil following the success of the Plan Real led us to expect high Brazilianinflationaversion.Chile is a similar outlier, with low concern about inflationdespite a history of high inflation this could suggest that the Chilean inflationstabilizationhad great credibility
3. Muddlingthe crime story further,there was only a weak statisticalassociationbetween the country dummiesin the crime regressionand the prevalenceof actualcrimes.



+ Ssgapore

* Llhaine


* Rllssh

Ct t:: o



S:L o

: o ._ Ct s::

lndea +

+ Menco Venenuela

* Czcch Rep

Indonesia *+

Philippines HongKong

Taiwan +

= 10.0

* Vlda *Gennany * USA Austna * * Spain *Japan * ttai?(nd VK * Sweden France + * tuStnlu * NethtrB8s dc
+ + * G+efiolombia

y o
._ ._

+ B>sl

E 5.0





, * I,

t A



_ t,


. * *



X t

FIG. 1. Probability of Mentioning Inflation as a Top National Problem and Average Previous lOYears Inflation During

AverageInOtion DuringPrevious 10 Years (Log base 10 scale)




in inflation indexation capitalmarket by 1995,andcouldalso reflectthe extensive Chile. to ancountry inflation we alsorelateactual to ourmainquestion, Moregermane asked survey. Thisquestion question askedon theRoper-Starch other poverty-related
Do you stronglyagree, mostly agree, mostly disagree,or stronglydisagreewith the following statement:"Inour society, the rich get richerand the poor get poorer."?

this whoanswer of respondents in eachcountry thepercentage InFigure 2, we graph 1985-94. Wesee a posrateof inflation agree" against theactual question "strongly Thus,notonlydo thepoor significant). (whichis highlystatistically itiveassociation wholesocietyhas a perinflation, but the complain more about withineachsociety societies.4 and poor in high-inflation gap between rich ception of a growing for theviewthatthe very strong support from the poll data provides Theevidence and that thelesseducated inflation than the rich, more dislike of poorexpress relatively 80 * Hungary 2: 70

* Gseece
+ + +

* Turicey Venezuela


4X) ,_ 60 X


Russia Uleraine

Belgium Ge tyUnitedKingdom$jlCololibia * Ftance #da * SwilteHand * Sweden * Poland * Mence


ct 50

v ,_

00 40 c) -;
ts M

+ +


Ih - rv CzechRepublic * USA * Jndia * China



20 -

NethesP^ailand * Philippines

O 3 10 _


* Indonesia





Actualinflation1985-94(log base 10 scale)

FIG. 2. Association Between Perceptionthat "RichGet Richer and Poor Get Poorer"andActual Inflation

4. This result seems to depend on the transitionand developingcountries,as can be seen from inspec-


forthe Thisprovides somesupport aremoreinflation averse thanthe moreeducated. morethanthench.Another interpretation viewthatinflation hurts thepoorrelatively the damage inflation does andto whom,the wouldbe that,whatever the factsabout thando thench.Thiswouldsuggest thatpopulist poorbelieve it to be moredamaging to appoliciesthan thoseseeking politicians arelikelyto pursue moreanti-inflationary withourex antebeliefs. pealto themiddle andupper classes,whichis notin accord on a coregroupof poorsupporters who receive Perhaps populist politicians depend thepooras a wholemaydislike inflation. benefits financed by inflation, eventhough

on the In this sectionwe turnto moredirectevidenceon the effectsof inflation wellof different measures of therelative distribution of income.Weuse a number rate,and in income,the poverty beingof the poor:the shareof the bottomquintile arecorrelated withinflation. therealminimum wage.All of thesethreeindicators
3.A Results on the Bottom Quintilein Income

average to thenextin theshare of thebottom Welookatchanges fromonedecade andSquire (1996)forthe 1970s,80s, quintile of income, usingthedataof Deininger of thebottom quintile on decadeaverage and90s.Weregress thechange in theshare growth (bothfromtheWorld Bankdatabase). CPIinflation andrealGDPpercapita ratez. [Z/(1 +X)] of thepercent inflation Weuse theinflation taxratetransformation This transformation reducesthe extentto whichextremevaluesof inflationwill in discrete time,orthe dominate theresults; it is alsothetaxrateon moneybalances overthe periodbeing annual rateof loss in the valueof moneycausedby inflation to simplyusingthe decade corlsidered. However, the resultsshownherearerobust turns outnotto be a average percent inflation, or the log changein the CPI.Growth of changes in thedistribution of income,as other statistically significant determinant Ravallion andChen1997),so in Table7 we show authors havefound(forexample, onlytheresults withtheinflation tax. in muchof the variation TheR-squared is verymodest,so we arenot explaining taxrateis However, thecoefficient on theinflation changes in bottom quintile shares.

(change in the shareof the bottom quintile of the income distribution)

Coefficient Standard Error t-Statistic Prob.

C INFLATIONTAX R-squared AdjustedR-squared

0.004195 -0.017412 0.028909 0.019918

0.001545 2.714256 0.005195 -3.351531 Mean dependentvariable S.D. dependentvariable

0.0077 0.0011 0.002084 0.013564

NOTES: Method:Least Squares Includedobservations:I 10 White Heteroskedasticity-Consistent StandardErrorsand Covariance


highly significant.We also try controllingfor growthbut it is not significantand does not change the significance of the inflationtax. A movement from zero inflationto hyperinflationwould decrease the share of the bottom quintile by 1.7 percentage points (from the coefficient on the inflation tax). This is economically significant since the sample averageshareof the poor in income is just 6.2 percent. of the inflationrate, the effect of changes in inflationis Given the transformation nonlinear:a change in the inflationrate from zero to, say, 40 percent, would reduce the share of the bottom quintile by 0.5 percent, which again is large relative to the With a positypically small share of the bottom quintile in the income distribution. tive constant,implying thatceterisparibusthe shareof the bottomincome quintilein this samplewould have increasedover time, we have the shareof the bottomquintile to an inflationrateof 31 increasingif the inflationtax is less than .24 (corresponding percent)and decreasingotherwise. There may be an argumentfor using the change in the inflation tax rate on the right-handside of this equationinstead of the level. We do not have clear priors on if some nominalincomes of the this: the level of the inflationtax is what is important inflation may effectively tax the poor are fixed. On the other hand, only "surprise" poor, so we would then want the change in inflation(as opposed to surprisesin the When we rerunthe price level, in which our original specification is appropriate). equationabove with the change in the inflationtax, it is not statisticallysignificant. Alternatively,we can run the equationin levels: the share of the bottom quintile regressed on the inflationrate (and the growthrate).The inflationtax rateis then a sigof the share of the bottom quintile;an increase of the inflation nificantdeterminant would then lower the shareof the bottom quintileby tax from zero to hyperinflation 1.7 percentagepoints. After getting this result in an earlierversion of this paper,we became awareof related resultsby Romerand Romer (1998). They show thatthe log of averageincome of the poorest fifth of the population is negatively related to log inflation across countries,and the Gini coefficient is positively relatedto log inflation. 3.B Inf ation and the PovertyRate We use data on poverty rates that span more than one point in time for forty-two developingand transitioncountriesover 1981-93, from household datacollected by Ravallion and Chen (1997). For each country, they construct a country-specific poverty line linked to mean income: it is 50 percent of the initial mean income for the household survey for that country,startingwith the initial year of the years included in the sample for that country. Ravallion and Chen present sixty-four episodes of changes in poverty rates using this country-specificpoverty line. The median length of an episode is three years. In Table 8 we regress the percentage change per year in the proportionbelow the poverty line (50 percentof initial mean income) on real GDP, per capita growth, and the inflation tax rate, over the period spannedby the change in povertyrate. The inflationtax rate has a significantpositive effect on the increase in poverty. The growth rate has a negative effect on the change in poverty, as Ravallion and


TABLE8 POVERTYCH (change in percent of households below the country-specific constructedpovertyline)

DEPENDENT VARIABLE: Variable Coefficient Standard Error t-Statistic Prob.

C GROWTH INFTAX R-squared AdjustedR-squared

7.171827 -5.328780 62.54719 0.496244 0.479727

9.541762 0.751625 1.439615 -3.701533 30.81613 2.029690 Mean dependentvariable S.D. dependentvariable

0.4552 0.0005 0.0468 35.79547 69.48502

NOTES: Method:Least Squares Includedobservations:64 White Heteroskedasticity-Consistent Standard Errorsand Covariance

Chen also found.The resulton the inflationtax rate is not robustto using the percent inflationrate or the log inflationrate (they have the same sign, and log inflation is significantat the 10 percentlevel), but the inflationtax rate does have appeal as the most appropriate functionalform.5 Once again, we are uncertainaboutwhetherthe level of the inflationtax rate or its change is more appropriate, for the same reasons mentionedbefore. In any case, the change in the inflationtax rate is insignificantin the poverty change regression, although it becomes significantof the same sign as in levels when an extreme outlier (Poland 1989-93) is omitted.6 3. C Inf ation and the Real MinimumWage The real minimumwage is not as clear an indicatorof the well-being of the poor as the two previous measures.A decrease in the real minimum wage could benefit the poor by facilitating their entry into formal sector employment, and too high a minimum wage could make the poor worse off by increasing formal sector unemployment. Nonetheless, assuming the minimum wage regulationsare observed, the real minimumwage is a welfareindicatorfor the groupof workersthatare at the bottom of the formal sector wage distribution. How might inflationaffect the real minimumwage? The governmentusually sets the nominal minimumwage. If there is downwardnominalrigidity,the government will find it easier to lower the real minimum wage during times of high inflation. There is also the arithmeticrelationshippointed out by Bacha and Lopes (1983), among others, that, given an initial real minimumwage, the averagereal minimum wage is lower the higher is inflation for a given indexation lag (for example, one month)from prices to wages. We use minimum wage data collected by Rama and Artecona (1999), using a pooled sample of annualdata for all years in which it is availablefor all countries.
5. This seems to imply that some of the extremeinflationobservationsdon't fit the regressionline very well. This conjectureis confirmed:Brazil and Peru are notable outliers to the regressionusing log inflation as the right-hand-side variable.If Brazil and Peru are omitted,then there is a significanteffect of log inflationon the change in poverty. 6. This outlier seems anomalousbecause it shows a large increase in poverty,while two other observations on Polandcovering subperiodsof this period do not show a dramaticchange in poverty.



We use the same CPI series as before.We regressthe log changein the real minimum wage on the inflationtax and on real growthper capita. High growthper capita implies rising laborproductivityand so would be expectedto translateinto higher average real wages; if the real minimumwage is sensitive to the averagereal wage, we would expect it to increase also. The results are shown in Table9. A high inflationtax rate is significantlyassociatedwith a negativepercentchange in the real wage. The real minimum wage change is positively associated with growth, as expected, with one percentagepoint more growth increasingreal minimum wages by 0.4 percent.The explanatory power of the regressionis again modest. The implied effect of inflationon the minimumreal wage is fairly strong:an increase in the inflationtax rate from zero to, say, 20 percent would reduce the real wage by eight percentage points. This-strong result depends in part on a large outlierNicaraguain 1987 when inflationwas near 1000 percentbut the nominal minimum wage only increasedby 22 percent.When this outlieris omitted,the relationshipbetween the real minimum wage change and the inflation tax is still significant, althoughthe magnitudeof the coefficient is cut in half. We also ranthe change in real minimumwage equationon the change in the inflation tax and the growth rate. The change in the inflation tax is highly significant. Thus, both the level of the inflationrate [as would be predictedby Bacha and Lopes 1983)] and its change (as would be predictedby models in which only surpriseinflation matters)are significantlyassociatedwith the real minimumwage.

This paperpresentsevidence that supportsthe view that inflationmakes the poor worse off. The primaryevidence comes from the answersto an international poll of 31,869 respondentsin 38 countries.These show thatthe disadvantaged on a number of dimensions the poor, the uneducated,the unskilled (blue-collar) worker are relatively more likely to mention inflation as a top concern than the advantagedon these dimensions. Each dimension is signiElcantwhen controlling for the others, suggesting that the differentcomponents of being disadvantagedhave independent effects on attitudesto inflation.

DEPENDENT VARIABLE: LOG PERCENT CHANGE IN REAL WAGE Variable Coefficient Standard Error t-Statistic Prob.

Inf ation TaxRate GROWTH
R-squared Adjusted R-squared

- 0.004066


0.004589 0.127369 0.122048

0.016368 0.001900 0.001547

Mean dependent S.D. dependent

2.511911 - 2.140478 2.966504

variable variable -

0.0125 0.0331 0.0032 0.000940 0.198479

NOTES: Method:Least Squares Includedobservations:331 White Heteroskedasticity-Consistent Standard Errorsand Covariance


We also examine the impact of changes in inflationon direct measuresof poverty and relatethem to inflation.We found thathigh inflationtendedto lower the shareof the bottom quintile and the real minimumwage, while tending to increase poverty. Similarresults on the directeffects of inflationon the per capitaincomes of the poor have been found recently by Romer and Romer (1998) and Agenor (1998). This paper presents evidence from surveying the poor themselves that they suffer more from inflationthanthe rich.

Paper,IMF "StabilizationPolicies, Poverty,and the Labor Market." Agenor, Pierre-Richard. andWorldBank, 1998. Bacha, Edmar,and FranciscoLopes. "Inflation,Growth,and Wage Policy: A BrazilianPerspective."Journalof DevelopmentEconomics 12 (1983). In Blank, Rebecca, and Alan Blinder. "Macroeconomics,Income Distribution,and Poverty." Fighting Poverty:WhatWorksand WhatDoesn't, edited by Sheldon Danziger and Daniel UniversityPress, 1986. Weinberg,pp. 180-208. Cambridge,Mass.: Harvard Cardoso, Eliana. "Inflationand Poverty."National Bureau of Economic Research Working PaperNo. 4006, March 1992. Cutler, David M., and Lawrence Katz. "Macroeconomic Performance and the Disadvantaged."BrookingsPapers on EconomicActivity2 (1991). Datt, Gaurav,and MartinRavallion."WhyHave Some IndianStatesDone BetterThanOthers at Reducing Rural Poverty?"World Bank Policy Research Working Paper 1594. April 1996. Deininger, Klaus, and Lyn Squire. "A New Dataset Measuring Income Inequality."World BankEconomicReview 10 (September1996), 565 -91. Fischer, Stanley. "Why are CentralBanks PursuingLong-Run Price Stability?"Federal Reserve Bank of KansasCity Symposium,AchievingPrice Stability,1996. Fischer, Stanley, and John Huizinga. "Inflation,Unemployment,and Public Opinion Polls." Journalof Money,Credit,and Banking 14 (February1982), 1-19. of the Real Effects and an Understanding Fischer,Stanley,and FrancoModigliani. "Towards (1978), 810-32. WeltwirtschaftlichesArchiv Costs of Inflation." Powers, Elizabeth T. "Inflation,Unemployment,and Poverty Revisited."Economic Review, 3 1995), 2-13. FederalReserve Bank of Cleveland(Quarter Rama, Martin,and RaquelArtecona. "A Data Base of LaborMarketIndicatorsacross Countries."WorldBank paper, 1999. andPovertyin Developing andTransition Ravallion,Martin,and ShaohuaChen. "Distribution Economies:New Data on Spells during 1981-1993." WorldBank paper, 1995. . "WhatCan New Survey Data Tell Us about Recent Changes in Distribution and Bank EconomicReview (1997), 357. Poverty?"World Rezende, Fernando."Prospectsfor Brazil's Economy."InternationalAJfairs74 (July 1998), 563-76. Policy and the Well-Being of the Poor."NaRomer,Christina,and David Romer."Monetary tional Bureauof Economic ResearchWorkingPaper6793, November 1998. UniRose, Richard."WhatIs the Demand for Price Stabilityin Post-CommunistCountries?" versity of Strathclyde,paper, 1997. NationalBureauof Economic Research Shiller,RobertJ. "WhyDo People Dislike Inflation?" WorkingPaper5539, April 1996.