Informal Consumption & Indirect Tax Design:Consumption Diaries Evidence from 15 Countries

Pierre Bachas (World Bank Research)

With Lucie Gadenne (Warwick) & Anders Jensen (HKS)

November 12, 2018

1 / 29

Consumption Taxes in Developing Countries

I in OECD countries, large Personal Income Taxes with ⇑marginal rates do heavy lifting of equity

I Atkinson-Stiglitz (JPubE 76): consumption tax superfluous ifcan tax income non-linearly (+ separability leisure-consumption)

I In practice for most developing countries:

I Domestic Consumption Taxes most important source ofrevenue but considered regressive/neutral

I Constrained PIT

I Expenditure redistribute modestly (Commitment To Equity)

I This paper: revisits the distributional role of consumptiontaxes in LMICs

2 / 29

Consumption Taxes Represent Large Share of Total Taxes

BGD BRA

CHNEGY

ETH

FRA

DEU

IRN

ITA

JPN

KOR

MEXPAK

PHL RUS

ZAF

THA

TUR

GBR

USA

0.1

.2.3

.4.5

.6.7

.8.9

1Do

mes

tic c

onsu

mpt

ion

taxe

s (%

Tota

l Tax

)

6 7 8 9 10 11 12Log GDP per Capita

N=125 countries in 2012Source: ICTD for tax revenue, World Bank for curent USD GPD in PPP.

3 / 29

Tax Mix over GDP

02

46

810

1214

Aver

age

Tax

Reve

nue

(% G

DP)

6 7 8 9 10 11 12Log GDP per Capita

Domestic consumption taxes Personal income taxes

N=128 countries in 2012Source: ICTD for tax revenue, World Bank for curent USD GPD in PPP.

4 / 29

Consumption Taxes in Developing Countries

I in OECD countries, large Personal Income Taxes with ⇑marginal rates do heavy lifting of equity

I Atkinson-Stiglitz (JPubE 76): consumption tax superfluous ifcan tax income non-linearly (+ separability leisure-consumption)

I In practice for most developing countries:

I Domestic Consumption Taxes most important source ofrevenue but considered regressive/neutral

I Constrained PIT

I Expenditure redistribute modestly (Commitment To Equity)

I This paper: revisits the distributional role of consumptiontaxes in LMICs

5 / 29

Taking into Account Informal Consumption

In LMICs many retailers do not pay consumption taxes⇒ a share of consumption is informal:

1. Are consumption taxes de facto progressive?

2. Are differentiated commodity taxes as currently designedin LMICs (with many exemptions) useful for equity?

Standard assumption: consumption taxes get fully passed on toconsumers through higher prices

6 / 29

This Paper

Use the idea that retailer size is a key determinant of formality:

I Consumption from own production, street selling, markets &corner stores is unlikely to remit taxes

I Compared to supermarkets, chain stores & department stores

Data: Representative household expenditure surveys in 15countries which specify place of purchase for consumption(at term aim to reach ∼ 30 countries)

Empirics: Description of informal consumption along the incomedistribution (“Informality Engel Curves”) and across countries

Theory: Ramsey model of optimal commodity taxation with aninformal consumption sector ⇒ Under what conditions doesinformal consumption makes indirect taxes progressive?

7 / 29

Income & Expenditure Surveys in 15 CountriesRepresentative country surveys with open diaries of consumption &Place of purchase variable

country survey year sample size # TOR source

Brazil POF 2008 56,049 753 Stat. OfficeCameroon ECAM 2014 10,303 18 Microdatalib

Chile EPF 2017 15,237 17 Stat. OfficeColombia ENIG 2007 42,733 26 Stat. Office

Costa Rica ENIGH 2014 5,705 26 Stat. OfficeDRC E123 2005 12,098 14 Microdatalib

Ecuador ENIGHUR 2012 41,760 74 MicrodatalibMexico ENIGH 2014 19,459 18 Stat. Office

Morocco ENCDM 2001 14,243 49 MicrodatalibMozambique IOF 2009 10,659 6 Microdatalib

Papua NG HIES 2010 3,811 5 MicrodatalibPeru ENAHO 2017 43,530 40 Stat. Office

Rwanda EICV 2014 14,419 14 MicrodatalibSouth Africa IES 2011 25,325 8 Microdatalib

Tanzania HBS 2012 10,168 13 Microdatalib

8 / 29

Consumption classification idea:

“Big picture” classification:

I Not marketed: Self-production, in-kind gifts

I Marketed non “brick-and-mortar”: Street selling, markets

I Marketed small stores: Convenience stores, specialized shops

I Marketed large stores: Supermarkets, department stores

9 / 29

Number of Employees & FormalityCombine Enterprise, micro & informal surveys (Retailers only)

(a) DRC0

.1.2

.3.4

.5.6

.7.8

.91

Sh

are

of F

orm

al F

irm

s

0 1 2 3Log Employment

(b) Rwanda

0.1

.2.3

.4.5

.6.7

.8.9

1S

ha

re o

f F

orm

al F

irm

s

0 1 2 3 4 5Log Employment

10 / 29

Mexico: Store Type & Number of Employees

0 1 2 3 4 5Log of Median Number of Employees

Price club

Supermarket

Department store

Specialized store

Convenience store

Street stall

Public market

Note: Sample of stores in the merged CPI-Census data

11 / 29

Classifying Consumption: Assumptions & Limitations

Definition of Consumption for Today:

I Consumption net of housing

I Informal consumption := convenience stores & smallerI The ‘residual is large & specialized stores + unspecified

I Unspecified consumption: 21% of total consumption, of which46% utilities, telecom & gas

Limitations:

I Ignore production chains which remit part of VAT

I Can’t break down chain convenience stores from stand alone

12 / 29

Consumption classification: Mexico

Sector TOR Original TOR Recode

Formal Supermercados Large StoresTiendas departamentales Large StoresTiendas con membresıa Large StoresCompras fuera del pais Large Stores

Restaurantes RestaurantsTiendas especıficas del ramo Specialized shops

Informal Tiendas de abarrotes Corner shopPersona particular From a household

Mercado Street & MarketsTianguis o mercado sobre ruedas Street & Markets

Vendedores ambulantes Street & MarketsLoncherıas, cocinas economicas Cafeterias

Unspecified No aplica Not Applicable

13 / 29

Informality Engel Curve: Mexico

010

2030

4050

6070

8090

100

7 8 9 10 11 12 13

MX 2014

14 / 29

Informality Engel Curves Across Countries (GDP pc sorted)0

10

20

30

40

50

60

70

80

90

10

0

6 7 8 9 10 11

DRC 2005

01

02

03

04

05

06

07

08

09

01

00

0 1 2 3 4 5 6 7 8

MZ 2009

01

02

03

04

05

06

07

08

09

01

00

6 7 8 9 10 11 12 13 14

RWA 2013

01

02

03

04

05

06

07

08

09

01

00

4 5 6 7 8 9

TZ 2012

01

02

03

04

05

06

07

08

09

01

00

8 9 10 11 12 13 14

CM 20140

10

20

30

40

50

60

70

80

90

10

0

7 8 9 10 11 12 13

PNG 2010

01

02

03

04

05

06

07

08

09

01

00

12 13 14 15 16 17

MA 2001

01

02

03

04

05

06

07

08

09

01

00

8 9 10 11 12 13 14 15

SA 2011

01

02

03

04

05

06

07

08

09

01

00

10 11 12 13 14 15

ECU 2011

01

02

03

04

05

06

07

08

09

01

00

4 5 6 7 8 9 10 11

CO 2007

01

02

03

04

05

06

07

08

09

01

00

6 7 8 9 10 11 12 13 14 15 16

PE 2017

01

02

03

04

05

06

07

08

09

01

00

7 8 9 10 11 12 13

MX 2014

01

02

03

04

05

06

07

08

09

01

00

8 9 10 11 12 13 14 15 16 17

BR 2008

01

02

03

04

05

06

07

08

09

01

00

6 7 8 9 10 11 12

CR 2014

01

02

03

04

05

06

07

08

09

01

00

7 8 9 10 11 12

CL 2017

15 / 29

Average Informal Consumption on GDP per capita

BR

CL

CM

CO

CR

DRC

ECUMA

MX

MZ

PE

PNG

RWA

SA

TZ

010

2030

4050

6070

8090

100

Ave

rage

Lev

el o

f Inf

orm

al C

onsu

mpt

ion

6 7 8 9 10 11Log GDP per capita (PPP adjusted)

16 / 29

Difference Bottom to Top Decile in Informal Consumption

BR

CL

CM

CO

CR

DRC

ECU

MA

MX

MZ

PE

PNG

RWA

SATZ

05

1015

2025

3035

4045

50D

iffer

ence

bot

tom

to to

p in

com

e de

cile

6 7 8 9 10 11Log GDP per capita (PPP adjusted)

Difference bottom to top decile of Informal Consumption

I Top decile has 27% less informal consumption than bottom decile

17 / 29

Slope of IE Curves Across CountriesI For each country β: Share Informali = βln(income pp)i + ΓXi + εi

BR

CL

CM

COCR

DRC

ECU

MA

MX

MZPE

PNG

RWA

SA

TZ

02

46

810

12

Slo

pe o

f inf

orm

al c

onsu

mpt

ion

6 7 8 9 10 11

Log GDP per capita (PPP adjusted)

18 / 29

Controlling for Product CompositionHow much is driven by the type of products being consumed at 6=incomes? Important since indirect taxes often based on commodities

I Run: Share Informalip = βln(income pp)i + αp + ΓXi + εip

I Weights for product importance in household consumptionI Control at several COICOP levels, shown sequentially

19 / 29

Product Composition: No controls to COICOP1

BR

CL

CM

COCR

DRC

ECU

MA

MX

MZPE

PNG

RWA

SA

TZ

02

46

810

12

Slo

pe o

f inf

orm

al c

onsu

mpt

ion

6 7 8 9 10 11

Log GDP per capita (PPP adjusted)

Change in Slopes: COICOP0 to COICOP1

20 / 29

Product Composition: COICOP1 to COICOP3

BR

CL

CM CO

CR

DRC

ECU

MAMX

MZPE

PNG

RWA

SA

TZ

02

46

81

01

2

Slo

pe

of

info

rma

l co

nsu

mp

tio

n

6 7 8 9 10 11

Log GDP per capita (PPP adjusted)

Change in Slopes: COICOP1 to COICOP2

BR

CL

CM CO

CR

DRC

ECU

MA

MX

MZ

PNG

RWA

SA

TZ

02

46

81

01

2

Slo

pe

of

info

rma

l co

nsu

mp

tio

n

6 7 8 9 10 11

Log GDP per capita (PPP adjusted)

Change in Slopes: COICOP2 to COICOP3

21 / 29

Product Composition SummaryProduct Composition reduces slopes by 45%:

I 30% with COICOP1, 8% COICOP2 and 7% COICOP3

0.2

.4.6

.81

Rat

io o

f slo

pes

RWA MA CM PNG CO PE MZ DRC TZ ECU SA MX BR CR

22 / 29

What Explains the Informality Engel Curves Slope?

Do poor households not have access to modern retailers or do theychoose not to go?

I Supply factors/access⇒ Can use proxies from surveys to study importance ofurbanization, density, transport, revealed choice set

I Demand factors: price, quality, complementaritiesconsumption/leisure, other attributes of modern retailers?⇒ Difficult to study in most surveys

23 / 29

Morocco: Reasons for Choosing Retailer

Reason Total Formal Informal

Access 55.73% 53.65% 56.04%Price 24.10% 7.33% 26.62%

Quality 7.78% 18.35% 6.18%Retailer’s Attributes 7.78% 9.64% 7.50%

Other 4.62% 11.03% 3.65%

Access is defined as a combination of proximity and necessity. Attributesof retailer is defined as a combination of homogeneity of products,offering of credit, and quality of reception.

24 / 29

Rural (Green) vs Urban (Blue) IE Curves Slopes0

10

20

30

40

50

60

70

80

90

10

0

6 7 8 9 10 11

DRC 2005

01

02

03

04

05

06

07

08

09

01

00

0 1 2 3 4 5 6 7 8

MZ 2009

01

02

03

04

05

06

07

08

09

01

00

6 7 8 9 10 11 12 13 14

RWA 2013

01

02

03

04

05

06

07

08

09

01

00

4 5 6 7 8 9

TZ 2012

01

02

03

04

05

06

07

08

09

01

00

8 9 10 11 12 13 14

CM 20140

10

20

30

40

50

60

70

80

90

10

0

7 8 9 10 11 12 13

PNG 2010

01

02

03

04

05

06

07

08

09

01

00

12 13 14 15 16 17

MA 2001

01

02

03

04

05

06

07

08

09

01

00

8 9 10 11 12 13 14 15

SA 2011

01

02

03

04

05

06

07

08

09

01

00

10 11 12 13 14 15

ECU 2011

01

02

03

04

05

06

07

08

09

01

00

4 5 6 7 8 9 10 11

CO 2007

01

02

03

04

05

06

07

08

09

01

00

6 7 8 9 10 11 12 13 14 15 16

PE 2017

01

02

03

04

05

06

07

08

09

01

00

7 8 9 10 11 12 13

MX 2014

01

02

03

04

05

06

07

08

09

01

00

8 9 10 11 12 13 14 15 16 17

BR 2008

01

02

03

04

05

06

07

08

09

01

00

6 7 8 9 10 11 12

CR 2014

01

02

03

04

05

06

07

08

09

01

00

7 8 9 10 11 12

CL 2017

25 / 29

Set up: Ramsey optimal commodity tax model

Three goods j : {0, 1, 2}e.g. 0=street food, 1=supermarket food, 2=non-food supermarketConstraint: Tax rate τ1, levied on both goods 0 & 1, τ2 on good 2

Gov. max social welfare: W =∫i G (v(p, y i )) + µ

∑j tjqjxj

I xj =∫i x

ij total consumption of good j , j ∈ {0, 1, 2}

I Consumer prices: qj = pj(1 + τj)

Assumptions:

I Exogenous incomes & producer prices

I Homogeneous price elasticities across hhlds: εij = εj ,∀j .I Initially set cross price elasticities to 0 (Relax later)

26 / 29

Optimal Tax: No Informal Sector

No informal sector ⇒ t1 is levied on both goods 0 & 1

Taking the derivative of W wrt to t1 gives:

τ∗1 =(µ− g)−

∫i (gi − g)

x i1+x i0x1+x0

−µε=

(µ− g)−∫i (gi − g)βi

s i1+s i0s1+s0

−µε

Is i1+s i0s1+s0

hhld i’s consumption of good j relative to avg, βi = y i

y

Equity term: cov(gi − g ,s i1+s i0s1+s0

)

I Goods consumed disproportionately by poor face lower rate

With an informal sector ⇒ t1 is levied just on good 1:

τ∗∗1 =(µ− g)−

∫i (gi − g)βi

s i1s1

−µε1

27 / 29

Optimal Tax With an Informal Sector

How does the tax rate on good 1 compares with/without aninformal sector? Depends on sign of: τ∗∗1 − τ∗1

I For comparative stats assume: ε0 ≈ ε1.

τ∗∗1 − τ∗1 is of the same sign as:∫i (gi − g)(

s i0−s i1s0s1

s1+s0)

I Take good with half formal - half informal csption (s1 = s0)⇒ Proportional to Informal Engel curve’s slope within product

I e.g. 1 is food ⇒ Existence of Informal sector ⇑ optimal rate

Allow substitution informal-formal ⇒ equity efficiency tradeoff:

I τ∗1 =(µ−g)−

∫i (gi−g)βi s

i1+si0s1+s0

−µ(ε+α) vs τ∗∗1 =(µ−g)−

∫i (gi−g)βi s

i1s1

−µε1

28 / 29

Optimal Tax With an Informal Sector

How does the tax rate on good 1 compares with/without aninformal sector? Depends on sign of: τ∗∗1 − τ∗1

I For comparative stats assume: ε0 ≈ ε1.

τ∗∗1 − τ∗1 is of the same sign as:∫i (gi − g)(

s i0−s i1s0s1

s1+s0)

I Take good with half formal - half informal csption (s1 = s0)⇒ Proportional to Informal Engel curve’s slope within product

I e.g. 1 is food ⇒ Existence of Informal sector ⇑ optimal rate

Allow substitution informal-formal ⇒ equity efficiency tradeoff:

I τ∗1 =(µ−g)−

∫i (gi−g)βi s

i1+si0s1+s0

−µ(ε+α) vs τ∗∗1 =(µ−g)−

∫i (gi−g)βi s

i1s1

−µε1

28 / 29

Conclusion

Literature sees informality as a constraint on tax policy by ⇓ thebase & ⇑ the efficiency cost of taxes. Instead, we consider howan informal retail sector impacts redistribution.

I Empirically: steep Informality Engle curves in 15 countries

I Ramsey model to illustrate equity-efficiency trade-off ofcommodity taxes with an informal sector

Based on preliminary results, some early policy implications:

I Consumption taxes De facto more progressive than assumed

I Important to revisit role of commodity tax exemptions, ascurrently designed, for equity

I Other policies limiting the size of informal sector could haveredistributive impact & shift the tax burden towards the poor

29 / 29

Extra Slides

These slides show the country by country graphs

1 / 33

In Progress

Model & Calibrations:

I Calibrate commodity taxes starting from current schedule, asfunction of elasticity of substitution formal & informal

I Endogenizing supply decision for firms to formalize

I Model the demand side’s preferences

Empirics:

I Expand sample of countries

I Build from WB Enterprise surveys relation firm size - formality

I Understand what drives consumption to types of retailers

“Companion” paper: within country reform in Mexico.

I Impact on prices by store type & competition of equalizationof border municipalities’ VAT rate to rest of country.

2 / 33

Related Literature

1. Explaining macro patterns with cross-countries micro-evidence(Lagakos et al 2018, Jensen, 2018, Bick et al, 2018)⇒ New stylized facts on consumption patterns & development.

2. Impact of firm size on taxation (Gordon & Li, 2009, Kleven, Kreiner,Saez, 2015) & retailer type with hhld income (Faber & Fally, 2018)⇒ Methodology: proxying informal consumption with store type

3. Incidence of commodity taxation is neutral/regressive in OECDcountries (Warren, 2008), in LICs do consumption taxes make taxsystems regressive? (Inchauste and Lustig, 2017)⇒ We show under what condition the informal sector could makecommodity taxes progressive.

4. Large literature on optimal commodity taxation (Auerbach andHines, 2002, Cremer and Gahvari, 1993) ⇒ We adapt classic modelto provide optimal tax results relevant to LIC/MIC context.

3 / 33

Democratic Republic of Congo

010

2030

4050

6070

8090

100

6 7 8 9 10 11

DRC 2005

4 / 33

Democratic Republic of Congo: Rural vs. Urban

010

2030

4050

6070

8090

100

6 7 8 9 10 11

DRC 2005

5 / 33

Mozambique

010

2030

4050

6070

8090

100

0 1 2 3 4 5 6 7 8

MZ 2009

6 / 33

Mozambique: Rural vs. Urban

010

2030

4050

6070

8090

100

0 1 2 3 4 5 6 7 8

MZ 2009

7 / 33

Rwanda

010

2030

4050

6070

8090

100

6 7 8 9 10 11 12 13 14

RWA 2013

8 / 33

Rwanda: Rural vs. Urban

010

2030

4050

6070

8090

100

6 7 8 9 10 11 12 13 14

RWA 2013

9 / 33

Tanzania

010

2030

4050

6070

8090

100

4 5 6 7 8 9

TZ 2012

10 / 33

Tanzania: Rural vs. Urban

010

2030

4050

6070

8090

100

4 5 6 7 8 9

TZ 2012

11 / 33

Cameroon

010

2030

4050

6070

8090

100

8 9 10 11 12 13 14

CM 2014

12 / 33

Cameroon: Rural vs. Urban

010

2030

4050

6070

8090

100

8 9 10 11 12 13 14

CM 2014

13 / 33

Papua New Guinea

010

2030

4050

6070

8090

100

7 8 9 10 11 12 13

PNG 2010

14 / 33

Papua New Guinea: Rural vs. Urban

010

2030

4050

6070

8090

100

7 8 9 10 11 12 13

PNG 2010

15 / 33

Morocco

010

2030

4050

6070

8090

100

12 13 14 15 16 17

MA 2001

16 / 33

Morocco: Rural vs. Urban

010

2030

4050

6070

8090

100

12 13 14 15 16 17

MA 2001

17 / 33

South Africa

010

2030

4050

6070

8090

100

8 9 10 11 12 13 14 15

SA 2011

18 / 33

South Africa: Rural vs. Urban

010

2030

4050

6070

8090

100

8 9 10 11 12 13 14 15

SA 2011

19 / 33

Ecuador

010

2030

4050

6070

8090

100

10 11 12 13 14 15

ECU 2011

20 / 33

Ecuador: Rural vs. Urban

010

2030

4050

6070

8090

100

10 11 12 13 14 15

ECU 2011

21 / 33

Colombia

010

2030

4050

6070

8090

100

4 5 6 7 8 9 10 11

CO 2007

22 / 33

Colombia: Rural vs. Urban

010

2030

4050

6070

8090

100

4 5 6 7 8 9 10 11

CO 2007

23 / 33

Peru

010

2030

4050

6070

8090

100

6 7 8 9 10 11 12 13 14 15 16

PE 2017

24 / 33

Peru: Rural vs. Urban

010

2030

4050

6070

8090

100

6 7 8 9 10 11 12 13 14 15 16

PE 2017

25 / 33

Mexico

010

2030

4050

6070

8090

100

7 8 9 10 11 12 13

MX 2014

26 / 33

Mexico: Rural vs. Urban

010

2030

4050

6070

8090

100

7 8 9 10 11 12 13

MX 2014

27 / 33

Brazil

010

2030

4050

6070

8090

100

8 9 10 11 12 13 14 15 16 17

BR 2008

28 / 33

Brazil: Rural vs. Urban

010

2030

4050

6070

8090

100

8 9 10 11 12 13 14 15 16 17

BR 2008

29 / 33

Costa Rica

010

2030

4050

6070

8090

100

6 7 8 9 10 11 12

CR 2014

30 / 33

Costa Rica: Rural vs. Urban

010

2030

4050

6070

8090

100

6 7 8 9 10 11 12

CR 2014

31 / 33

Chile

010

2030

4050

6070

8090

100

7 8 9 10 11 12

CL 2017

32 / 33

Chile: Rural vs. Urban

010

2030

4050

6070

8090

100

7 8 9 10 11 12

CL 2017

33 / 33