Combatting collusion: early lessons from Madagascar Customs

Transcription

1 Combatting collusion: early lessons from Madagascar Customs Cyril Chalendard Ana Fernandes Aaditya Mattoo Gael Raballand Bob Rijkers PICARD Tunisia 2017

2 Introduction Although corruption (in customs) is widely discussed, practitioners and academics still have poor information on these bad practices. In particular, Little is known about the frequency and the severity of corruption agreements No estimates of related tax-losses are available =>This study aims to fill this gap Madagascar customs: an ideal case study as very admirable efforts to fight corruption are made Main challenges Corruption agreements typically «fly under the radar» The «true» transaction value is unknown => proxies 2

3 II. Definition of proxies II.A. Corruption indicators II.B. Tax losses indicators DGD - Sept

4 A. Corruption indicators (1/2) Theoretical background: (frontline) inspections = a market Declarations registered by a specific economic operator (broker or importer) = one specific market Actors: frontline inspectors For each market, we are able to calculate and analyze The market share of each inspector Percentage of declarations inspected by a specific frontline inspector. The market concentration - Minimum concentration: all inspectors have a similar market share - Maximum concentration: declarations are only inspected by one inspector DGD - Sept

5 A. Corruption indicators (2/2) 1. Measure of market concentration (Herfindahl-Hirschmann index HHi) -Calculated at the broker level, on a semester basis. -A broker with a high degree of concentration (HHi closer to 1) => his declarations are inspected by a relatively small number of inspectors => risk of collusion 2. Excess Market Share (EMS) -Calculated at the inspector-broker level, on a semester basis. -If inspector I inspects relatively more frequently declarations registered by broker B => risk of collusion between I and B 3. Date of registration/assessment during non-working hours (Extra) -At the declaration level. -When an action is taken during non-working hours, then collusion agreements are easier to reach=> risk of collusion DGD - Sept

6 A. Descriptive statistics Corruption indicators EMS(0.05)=1 if EMS>Average market share ; 0 otherwise EMS(0.1)=1 if EMS>Average market share ; 0 otherwise Extra=1 if registration/assessment during non-working hours; 0 otherwise Table 1 EMS (0.05) EMS (0.1) Extra # declarations 68,067 68,067 68,067 Table 2 =1 9,475 6,711 4,792 Frequency 14% 10% 7% Mean Median p25 p75 Min Max Concentra tion (HHi) DGD - Sept

7 B. Tax losses indicators Tax losses = Value gap x Import taxation rate Value gap= Reference value Final customs value 3 (imperfect) reference values 1. Hypothetical External Reference Value = expected export value (calculated using UN COMTRADE database) HERV gap 2. Hypothetical Internal Reference Value =import value typically declared (mean or median) HIRV gap 3. Valitrade value (advice given by Gasynet-SGS) Valitrade gap 7

8 III. The assignment of declarations to frontline inspectors 8

9 Did assignments follow assignment rules? 2-step procedure 1. For each broker, we simulate the concentration based on assignment rules (random assignment) 2. Then, we compare the observed (=real) distribution with the simulated distribution (random assignment) Distribution des indicateurs de concentration par CAD Concentration Observee Simulee (selon procedure) Distributions are statistically different => assignments did not follow rules 9

10 IV. Risk of collusion, risk of fraud and (frontline) inspections 10

11 Main facts 1. Declarations with a risk of collusion (i.e. corruption indicator=1) are declarations with the biggest stakes 1. Larger initial declared value 2. Larger taxation rate 3. More items 2. Declarations with a risk of collusion are also more risky 3. Undervaluation is estimated more important 4. There is a positive and continuous relationship between the size of collusion and the amount of taxes at risk DGD - Sept

12 The risk of fraud increases with EMS 12

13 The risk of fraud increases with EMS 13

14 Undervaluation increases with EMS 14

15 Undervaluation increases with EMS 15

16 Inspections are less efficient A risk of collusion is associated with => Lower frequency of fraud records Lower frequency of scanning Reduced clearance time Frequency of fraud records EPDM 0.10 No EPDM EPDM Fraud record (%) Note: red channel only 16

17 Inspections are less efficient A risk of collusion is associated with => Lower frequency of fraud records Lower frequency of scanning Reduced clearance time Frequency of fraud records EPDM 0.10 Normal market share No EPDM Excess market share EPDM Fraud record (%) Note: red channel only 17

18 V. Risk of collusion and tax losses DGD - Sept

19 Tax losses based on the HERV gap If a declaration is handled by a broker with a relatively high concentration (> median), then tax losses are estimated higher HERV gap= Expected export value - Final customs value 19

20 Tax losses based on the HERV gap If a declaration is handled by a broker with a relatively high concentration (> median), then tax losses are estimated higher HERV gap= Expected export value - Final customs value Low concentration USD 1,700 High concentration 20

21 Tax losses based on the HIRV gap If a declaration is handled by a broker with a relatively high concentration (> median), then tax losses are estimated higher HIRV gap= Import value typically declared - Final customs value (Median) tax losses per declaration-hirv According to the degree of concentration HHi<median HHi>median [HERV-Dec value]*tax rate Note: currency: USD 21

22 Tax losses based on the HIRV gap If a declaration is handled by a broker with a relatively high concentration (> median), then tax losses are estimated higher HIRV gap= Import value typically declared - Final customs value (Median) tax losses per declaration-hirv According to the degree of concentration Low concentration HHi<median USD 240 High concentration HHi>median [HERV-Dec value]*tax rate Note: currency: USD 22

23 First estimates Declarations with a risk of collusion are declarations with the biggest stakes (selection bias) => need to rely on advanced methods (econometrics) Tax losses= [$ 8 million ; $ 93 million] (Toamasina Port, 2016) Tax losses (% total revenues) Regression I (permissive) Regression II (conservative) Hypothetical Internal Reference Value 42.9% 3.4% Valitrade Value 10.2% 4.7% 23

24 VI. First lessons 24

25 First lessons The exploitation of customs databases may reveal collusion practices Declarations with a risk of collusion are, in Madagascar, -highly sensitive (large taxation rate, high initial value etc) -associated with important tax losses More broadly, this study -provides information on the scope of collusion practices -quantifies tax losses due to collusion -supports anti-corruption reforms (in Madagascar customs) 25