C1 - [Schneider & Medina 2018] - Shadow Economies Around the World: What Did We Learn Over the Last 20 Years?

 

ANALYSIS OF ARTICLE

Didactic article describing different methodologies to estimate the size of the Shadow Economy, focusing on MIMIC and some alternatives over it.

It provides estimates for 158 countries from 1991 to 2015, using different methods.

Article also discuss possible research areas for those methods.

 

 Section 1 (Introduction) –.

The paper uses the shadow economy definition of legal economic and productive activities that would contribute to national GDP.

Goals:

1)      Evaluate and discuss latest estimation methods: CDA, MIMIC, PMM

2)      Present shadow economy estimates for 158 countries from 1991 to 2015

3)      Compare results of different estimation methods.

 

 

Section 2 (Theoretical Considerations)

Shadow Economy SE depends on:

a)      p = Probability of detection

a.      determined by enforcement actions A and facilitating activities F

b)      f = Potential fines

c)      B = Opportunity costs of remaining formal

a.      Determined by taxation T and higher labor costs W

The Article presents literature exploring the main causes of informality (shadow economy).

Section 3 (Estimation Methods and MIMIC Estimation Results):

Provides an overview of estimation methods.

Main Direct approaches (Micro):

1)      Micro Approach – measurement by the System of National Accounts Statistics SNA (Discrepancy Method) – described in detail in the OECD 2010 Handbook for Measuring the Non-Observed Economy NOE.

2)      Micro Approach – representative surveys – leverage on targeted surveys to get micro knowledge about the size of the shadow economy and shadow labor markets

3)      Micro Approach – survey with company managers -

Main Indirect approaches (indicators, Macro):

1)      Macro Approach – discrepancy between national expenditure and income stats – use the gap between income and expenditure to size the shadow economy, considering SE can hide their incomes for tax purposes, but not their expenditure

2)      Macro Approach – discrepancy between official and actual labor force – fluctuation on the official labor force vis-à-vis total labor force estimates can indicate SE

3)      Macro Approach – electricity consumption as the single best indicator of overall (official and unofficial) economic activity

4)      Macro Approach – transaction using Fischer’s equation. Considering Prices * Transactions constant (official GDP + SE), Money * Velocity would also be constant. Stock of Money and GDP are known, and Velocity can be estimated, resulting in an estimation of SE.

5)      Macro Approach – currency demand approach (CDA). Increased demand for currency considering informal transactions would be done in cash.

6)      Macro Approach – Multiple Indicators, Multiple Causes MIMIC approach

 

MIMIC is a special type of structural equation modeling (SEM) based on the statistical theory of unobserved variables developed in the 1970s by Zellner (1970) and Joreskog and Goldberg (1975). It provides a set of exogeneous causal variables on the latent variable (SE), as well as the effect of SE on macro indicator variables.

Literature provides a MIMIC model to measure the size of the SE, Schneider et al. (2010), Hassan et al. (2016), Buehn et al. (2009):

1)      Modeling the SE as an unobservable variable

2)      Structural Model – SE as a result of a vector of causal variables. Causal variables:

a.      Share of direct taxation

b.      Share of indirect taxation

c.       Share of social security burden

d.      Burden of state regulation

e.      Quality of state institutions

f.        Tax morale

g.      Unemployment quota

h.      GDP per capita

3)      Measurement Model – Indicators as a result of the SE

a.      Employment quota

b.      Change of local currency

c.       Average working time

 

The Author discusses limitations of the MIMIC approach, as well as literature related to it, and describes a new hybrid-model using CDA and MIMIC models as something promising in the research landscape.

 

The Article provides then SE estimates for 158 countries using the MIMIC approach. It also provides a discussion for alternative ways to avoid misspecification resulting from endogeneity on causal variables. For instance, it describes the usage of night lights intensity as a proxy for GDP.

 

Predictive Mean Matching PMM:

Treats it as a missing data problem: for several countries, we have survey-based estimates of the size of the shadow economy, but for other countries it’s missing. PMM may work as a matching tool to run a non-parametric approach, matching countries that lack data to countries that have data, based on their similarity, where PSM (propensity score matching) is a promising candidate

 

 

Section 4 (A Comparison of MIMIC with Micro Survey Results) –

a)      MIMIC vs SNA – MIMIC is considerably larger for most countries.

b)      MIMIC vs Micro Survey methods – adjusted MIMIC comes close to the micro survey estimates.

c)      Macro vs Micro methods –

 

  

Section 5 (Summary)

-          Macro approaches provide upper bound estimates as they include crime activities and others

-          MIMIC estimates depend on the starting values if they are taken from macro estimates (bias)

-          Promising approach: Dybka et al. (2017) with hybrid CDA and MIMIC – however they produce much lower sized SE estimates

-          Provide SE estimates for 158 countries from 1991 to 2015 using different methods, including

o   MIMIC with the light intensity approach instead of GDP

o   PMM methodology – robust results that confirms those of MIMIC

-          Overall, SE declines over time (from 1991 to 2015) except for 2008 with the financial crisis

adADDITIONAL INFORMATION

C1A- CCESIFO working paper

Do different estimation methods lead to implausible differences in the size of NOE or shadow economies? A preliminary answer (Friedrich Schneider)

c

df