## _wp15121

## Source details

**Canonical URL:** [_wp15121](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15121.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15121.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15121.pdf.json)

---

### Objectives and approach
- Investigates the effect of crime on overall economic activity over the period 1993–2012, aiming to: (i) uncover underlying causal relationships; (ii) account for the dynamic nature of such relationships; and (iii) explain heterogeneous responses across Mexican states.
- Uses panel SVAR methodology (Pedroni (2013)) that:
  - Identifies causal “shocks” driving both crime and economic activity.
  - Distinguishes supply-side and demand-side economic shocks from crime-originating shocks using a mix of short-run and long-run restrictions.
  - Defines crime shocks as any crime which induces a movement in the homicide rate subject to identifying restrictions.
- Primary analysis studies the response of homicide data and economic data to general crime shocks (not a particular category of crime shock).

### Data and identification
- Panel: 32 Mexican states, annual observations 1993–2012 on:
  - homicides (primary: INEGI recorded intentional homicides),
  - state gross domestic product (state GDP),
  - state foreign direct investment (FDI),
  - state population migrations (international net in-migration).
- VAR variables:
  - GDP: log per capita state GDP.
  - FDI: log per capita cumulative FDI state inflows (real, deflated by U.S. CPI).
  - Crime: log per capita state homicides (INEGI).
  - Migration: log per capita international net migration into the state (transformed to ensure positivity where needed).
- Identification restrictions (short-run and long-run):
  - AD shocks: restricted not to cause permanent movements in GDP (long-run exclusion).
  - AS shocks: permitted to cause permanent movements in GDP.
  - Crime shocks: allowed immediate impact on homicides (short-run restriction); GDP may be impacted immediately.
  - Migration shocks: migration can respond quickly to crime shocks; homicides change more slowly in response to migration shocks.
  - Signs fixed via economic rules (e.g., crime shock positive if increases homicides in impact period).
- Structural shocks decomposed into common (national/international) and idiosyncratic (state-specific) components, yielding eight structural shocks and 32 impulse responses per state.

### Methodology and robustness
- Panel SVAR permits heterogeneous state-specific dynamic responses; responses represented via impulse responses and importance via dynamic variance decompositions.
- Bootstrap estimation with 5,000 draws used to produce confidence bands for impulse responses, variance decompositions, and second-stage regression correlations.
- Robustness checks:
  - Five chaining methods to link state-level RGDP series into a single base year; key results qualitatively invariant.
  - Subsystems (two- and three-variable blocks) and two-variable VAR (homicides and GDP, 2003–2012) examined; primary results hold.
  - Inclusion of dummy years 2008 and 2009 had only relatively minor numerical effects.

### Key empirical findings — national and aggregate
- National-level impact:
  - Crime shocks originating at the national or international level are responsible for a quarter of a percent impact on national per capita GDP over time (reported aggregate national GDP response: roughly one-quarter percent of national GDP for up to two years, decreasing in the third year and dissipating by the fourth year).
- On average across states:
  - Median state: a typical state-specific crime shock induces a roughly one-half percent temporary decrease in per capita GDP which persists up to two years after the shock and dissipates after the third year.
- GDP-approach diagnostic statistics (second-stage / bootstrapped estimates):
  - GDP-1 based VAR estimate: 1.091.
  - GDP-2 based VAR estimate: 0.924.
  - Bootstrapped ninety-fifth percentile for GDP-1: 2.065.
  - Bootstrapped fifth percentile for GDP-1: 0.112.
  - Bootstrapped ninety-fifth percentile for GDP-2: 1.944.
  - Bootstrapped tenth percentile for GDP-2: 0.112.
  - Interpretation: "we have greater than 90 percent confidence that the relationship is positive and statistically significant" (GDP-1), and a similar but slightly wider confidence band for GDP-2.

### Key empirical findings — heterogeneity and state-level distributions
- Variance decompositions (relative importance of shocks):
  - Median percentage of GDP variation explained by state-level aggregate supply (AS) shocks: roughly 70 percent.
  - Median percentage of variation in state GDP driven by state-specific crime shocks: roughly five percent.
  - Median percentage of variation in state GDP driven by common national or international crime shocks: roughly a half a percent.
  - Seventy-fifth percentile: state-specific crime shocks account for as much as 10 percent and national/international crime shocks another one percent (combined up to 11 percent for at least a quarter of states).
- Extreme-quantile and multi-year effects:
  - Up to 5.5 percent of the variation in state-level GDP in the years following shocks can be attributed to state-specific and common crime shocks together.
  - Up to one-fourth of all states experience as much as 11 percent of the variation in their GDP attributable to crime shocks.
  - Up to a quarter of states see on average a 1.6 percent per annum decrease in GDP for the first five years following a crime shock.
- Impulse response quantiles (example interpretation):
  - Median response of log per capita GDP to a state-specific crime shock: initially slightly negative and eventually moves to approximately -0.005.
  - Twenty-fifth percentile: starts at -0.004 and drops to around -0.16 by five years after the shock (i.e., at least a quarter of states: ceteris paribus crime shock reduces real per capita GDP by 1.6 percent per year after five years).
  - For at least a quarter of states, log per capita GDP decreases in response to crime shocks (90 percent confidence); for at least a quarter of states, log per capita GDP increases.
  - For the 25 percent or more states where log per capita GDP falls: annual GDP falls by at least one-half percent, and this decrease appears to be permanent.
- Scaling examples and interpretations:
  - Median state homicide rate increases by 0.2 percent in the initial period in response to the crime shock.
    - A crime shock that induces an initial 0.2 percent increase in homicides causes a 0.5 percent decrease in GDP for the median state.
  - Median state accumulated FDI decreases by 0.1 percent by the fifth year following the crime shock.
    - A crime shock that decreases accumulated FDI by 0.1 percent over five years for the median state causes a 0.5 percent decrease in GDP for the median state.
- FDI and migration dynamics:
  - Median state variation in FDI explained by crime shocks: roughly one percent.
  - For states in the top quartile, up to eight percent of FDI variation is driven by crime shocks initially, rising to as much as 15 percent by five years after the shock.
  - For at least a fourth of states, state-specific crime shocks initially account for 10 percent of the variation in migration, rising to close to 20 percent five years after the shock.
  - FDI responses heterogeneous: for at least 25 percent of states, by the fifth year cumulative FDI has decreased by somewhere between one percent and six percent; for at least 25 percent of states, FDI increases following a crime shock.

### State characteristics associated with smaller economic responses to crime shocks (second-stage analysis)
- Estimation approach:
  - For each of 5,000 bootstrap draws: compute state-specific impulse responses and regress them on state characteristics; collect slope estimates to form sampling distribution and 90 percent confidence bands.
- Robust relationships identified:
  - Construction sector share:
    - States with a bigger proportion of economic activity devoted to construction appear to experience smaller magnitude negative impacts of crime shocks on per capita GDP (positive association: higher construction share → crime shock more likely to have a positive or less negative effect on overall state GDP).
  - Perception of public security:
    - Higher household-reported sense of security systematically associated with smaller consequences for the economy from an equal-sized crime shock and with less negative per capita accumulated FDI responses.
    - Bootstrapped median estimator for the slope linking FDI response to public security:
      - GDP-1 based VAR: 0.006.
      - GDP-2 based VAR: 0.005.
      - Bootstrapped ninety-fifth percentile (both VARs): 0.01.
      - Bootstrapped fifth percentile:
        - GDP-1: 0.002.
        - GDP-2: 0.001.
      - Interpretation: "we are 90 percent confident in both cases that the relationship is statistically significant and positive."

### Which shocks drive homicide variation
- Large cross-state spread in drivers of homicide variation.
- In top quartile states, over 10 percent of the variation in crime can be attributed to state-specific migration shocks.
- At longer horizons, a little over three percent each of crime variation can be attributed to state-specific AS shocks and AD shocks.

### Limitations, scope, and caveats
- Time coverage: sample period ends in 2012; the analysis does not account for reforms or developments since 2012.
- Measurement issues:
  - Measurement error in crime data (INEGI vs SESNSP differences; reclassification of deaths) can limit inference.
  - Homicide data cannot distinguish organized-crime-related homicides from other causes.
  - FDI statistics recorded in the home of the company's headquarters may bias common vs state-specific shock attribution.
- Stationarity implications:
  - Log per capita GDP and INEGI per capita homicides: found to follow a unit root at state level → shocks can have permanent effects.
  - SESNSP crime measures: appear stationary → shocks transitory and cannot explain long-run GDP changes.
  - Migration: per capita international net in-migration appears to follow a unit root at state level.
- Structural VAR interpretations depend on the chosen identification scheme; results robust to viable identification variations but not conclusive as causal proof.

### Policy relevance and implications (interpretive, not prescriptive)
- Perception and public confidence in institutions providing public security are potentially important in mitigating the economic impact of crime.
- Findings relevant for:
  - Ongoing judicial reform in Mexico (federal judicial reform approved by Congress in 2008; expected in all states by 2016).
  - Continuing efforts to strengthen the AML/CFT regime in Mexico (noted efforts through end-2012 and referenced reforms).
- Noted complementary developments (outside primary analysis period): Mexico reported efforts toward further strengthening of its AML/CFT regime; INEGI recorded a decrease in total number of homicides in the period since 2012; judicial reform agenda expected by 2016 — effects remain to be assessed.

### Technical appendix — rebasing and linking real GDP series (method summaries and scenarios)
- Problem: INEGI published RGDP aggregated data 2003–2012 base year 2008; RGDP 1993–2003 base 1993 not rebased to 2008. Two main linked series constructed:
  - RGDP-1: aggregate linking by computing for year 2003 the ratio (aggregate RGDP base 2008)/(aggregate RGDP base 1993) and multiply base-1993 series by this ratio.
  - RGDP-2: sector-level linking: compute ratio for each GDP sector using common years 2003–2006, estimate sector values with base 2008 for 1993–2002, then sum sectors to obtain aggregated RGDP base 2008 for 1993–2002.
- Chain-linking and rebasing approaches detailed (preserved formulas and notation in source):
  - Aggregated constant-scaling: Formula (6) and (7) (single aggregated ratio R or averaged aggregated ratios over 2003–2006).
  - Sectoral constant-scaling: Formula (5) using averaged sectoral ratios R_i across 2003–2006.
  - Geometric/gradual adjustment: Formulas (8) and (9) using geometric development over 10 years and averaging across 2003–2006.
  - Log-linear fit to account for accounting changes and time variation: Formula (10) fits log D_t = c + δ t for t=2003,2004,2005,2006 and constructs D̂_t for t=1993,...,2002; RGDP(2008,7)t = D̂_t * RGDP(1993)t.
- Implementation and robustness:
  - Two linked real GDP series used in reported empirical results:
    - GDP Approach 1: linked by formula (6).
    - GDP Approach 2: linked by formula (5).
  - Experimentation with other scenarios (3),(5)–(9) shows results "quite robust, regardless of the choice of approach to link the GDP series."
  - Visual comparisons across approaches reported in source figures.

### Key quantitative summaries (preserved numeric statements)
- National per capita GDP impact from common crime shocks: a quarter of a percent.
- Median state homicide impact scaling: initial 0.2 percent increase in homicides → 0.5 percent decrease in GDP (median state).
- Median accumulated FDI decrease by fifth year: 0.1 percent (median state).
- Median state GDP variation explained by state-specific crime shocks: roughly five percent; by common crime shocks: roughly a half a percent.
- Seventy-fifth percentile combined crime-share in GDP variation: up to 11 percent (state-specific up to 10 percent + national/international 1 percent).
- Up to 5.5 percent of the variation in state-level GDP attributable to crime shocks (state-specific and common) in years following shocks.
- For up to a quarter of states: on average a 1.6 percent per annum decrease in GDP for the first five years following a crime shock.
- Confidence and bootstrap parameters:
  - Bootstrap draws: 5,000.
  - Confidence bands reported at 90 percent (equivalent to significance at the 10 percent level).
  - GDP-1 and GDP-2 diagnostic estimates and percentiles as listed above (1.091, 0.924, 2.065, 0.112, 1.944, 0.112).
- Regulatory note from source: Mexico forbade cash payments of more than a half million pesos ($38,750) for real estate transactions (end of 2012).

*Source: _wp15121*

### Executive Summary ......................................................................................................

### Executive Summary

### Objectives and approach
- Investigates the effect of crime on overall economic activity over the period 1993–2012, aiming to: (i) uncover underlying causal relationships; (ii) account for the dynamic nature of such relationships; and (iii) explain heterogeneous responses across Mexican states.
- Uses newly published econometric techniques for panel data to identify causal “shocks” driving both crime and economic activity, distinguishing supply-side and demand-side economic shocks from crime-originating shocks based on long-run effects on per capita output and short-run immediate effects on the crime variable.
- Primary analysis studies the response of homicide data and economic data to general crime shocks (not a particular category of crime shock).

### Data and identification
- Panel comprises the 32 states of Mexico with annual observations from 1993 to 2012 on:
  - homicides,
  - state gross domestic product (state GDP),
  - state foreign direct investment (FDI),
  - state population migrations.
- Identification scheme uses homicide data in part to capture the response to crime shocks and imposes econometric restrictions to disentangle classes of shocks.
- Recognition of limitations: measurement error in data; results dependent on the econometric identification scheme and assumptions about captured shocks.

### Key empirical findings
- National-level impact:
  - Crime shocks originating at the national or international level are responsible for a quarter of a percent impact on national per capita GDP over time.
- State-level responses (reported as sample distribution quartiles due to limited precision of individual state estimates):
  - State-specific crime shocks associated with on average an initial one fifth percent increase in homicides induce on average a one-half percent temporary decrease in per capita GDP.
  - The average temporary decrease in per capita GDP persists for up to two years after the shock and dissipates after the third year.
  - For the quarter of states experiencing the biggest effect, the impact is roughly a one-half percent decrease in per capita GDP that does not dissipate but remains permanently reduced.
- Heterogeneity:
  - Responses across states are mixed in sign and magnitude; aggregate estimates can mask important state-level variation.
- International factor flows and crime:
  - Crime has been intertwined with overall economic activity, including international factor flows in the form of FDI and migrations during the sample period.
  - The study isolates the relative magnitude and importance of various shocks affecting these linkages.

### State characteristics associated with smaller economic responses to crime shocks
- Construction sector share:
  - States with a bigger proportion of economic activity devoted to construction appear to experience smaller magnitude impacts of crime shocks on per capita GDP.
- Perception of public security:
  - Higher household-reported sense of security is systematically associated with smaller consequences for the economy from an equal-sized crime shock, highlighting the potential importance of public confidence in institutions providing security.
- FDI responsiveness:
  - The size of the economic response, particularly the FDI response to a crime shock, is related to these state-specific characteristics.

### Limitations, scope, and policy context
- The paper is an econometric research study and does not advocate specific policy responses.
- Sample period ends in 2012; the analysis does not cover developments since 2012.
- Notes subsequent developments mentioned in the text outside the primary analysis:
  - Mexico reported efforts toward further strengthening of its AML/CFT regime.
  - INEGI recorded a decrease in total number of homicides in the recent period since 2012.
  - A judicial reform agenda expected to be in place in all states by 2016.
- These efforts may mitigate the impact of crime on state-level economic activity, but their effects remain to be assessed.

*Source: _wp15121 - Executive Summary*

### section III of our paper, we employ the panel SVAR methodology of Pedroni (2013), which is designed to

### _wp15121 - section III of our paper, we employ the panel SVAR methodology of Pedroni (2013), which is designed to

### Methodology and identification
- Employs the panel SVAR methodology of Pedroni (2013), designed to accommodate panels comprised of relatively short and possibly noisy data.
- Uses a structurally identified VAR approach subject to an econometric identification scheme to study the role of shocks to crime in general rather than relying on narrowly defined proxies.
- Recorded homicide data are used as one of the endogenous variables, but in the VAR analysis "crime shocks" are defined as any crime which induces a movement in the homicide rate subject to the identifying restrictions placed on the other economic variables in the system.
- Section III of the paper describes the empirical methodology and strategy for identifying crime shocks, other key structural shocks in a panel VAR context, and limitations of the analysis.

### Key empirical findings and quantitative results
- On average, state-specific and common national and international crime shocks together are responsible for up to 5.5 percent of the variation in state-level GDP in the years following the shocks.
- Considerable heterogeneity across states:
  - Up to one-fourth of all states experience as much as 11 percent of the variation in their GDP attributable to crime shocks.
  - Up to a quarter of all states see on average a 1.6 percent per annum decrease in GDP for the first five years following a crime shock.
- Crime shocks have a considerable but very heterogeneous relationship to FDI flows at the state level.
- Two state-level characteristics associated with the magnitude of GDP and FDI responses to crime shocks:
  - The relative importance of the construction sector as a percentage of the state economy is on average associated with a diminished negative impact of crime on overall economic activity.
  - A sense of insecurity, as reflected in public survey data, is on average associated with an increased negative impact of crime on economic activity and FDI in particular.

### Policy relevance and implications
- Findings are potentially important for:
  - The ongoing process of judicial reform in Mexico.
  - Continuing efforts toward a further strengthening of the Anti-Money Laundering/Countering the Financing of Terrorism (AML/CFT) regime in Mexico.
- Note on temporal coverage:
  - The study focuses on the period from 1993 to 2012 and therefore does not account for reforms in Mexico since 2012.

### Paper structure and additional materials
- The remainder of the paper (as described in the source):
  - Presents the empirical framework for examining the effect of crime on the economy.
  - Describes empirical methodology and identification strategy (section III).
  - Describes the data (section IV).
  - Presents and discusses empirical results (section V).
  - Summarizes and concludes (section VI).
  - Includes a technical appendix with a description of the methodology used for constructing a sufficiently long series for real GDP using a single base year.

### Related literature (brief)
- Cites theoretical and empirical literature linking crime to growth and development and quantifying economic and social costs of organized crime (examples include Bourguignon, 2001; Fajnzylber et al., 2002; Mauro and Carmeci, 2007; Fajnzylber et al., 2002c; Buvinic and Morrison, 1999; Glaeser, 1999; International Centre for the Preventing of Crime, 1998; Londono and Guerrero, 2000; Rios, 2011).
- Notes empirical work on transmission channels through which crime impacts economic growth (e.g., Goulas and Zervoyianni, 2013; Detotto and Otranto, 2010; Detotto and Vannini, 2009; Czabanski, 2008; Brand and Price, 2000; Anderson, 1999).
- Highlights that empirical studies have not yet produced a definitive consensus.

*Source: _wp15121 (section III and surrounding text) — content as provided.*

### conclusion regarding the impact of crime on economic growth. A way to measure the crowding-out effect

### _wp15121 - conclusion regarding the impact of crime on economic growth. A way to measure the crowding-out effect

### Approaches to measuring the crowding-out effect of crime
- Two broad empirical approaches:
  - Cross-sectional macro comparisons of countries/regions with high versus low crime, controlling for other variables (originating from Barro, 1996).
    - Example findings: Peri (2004) — annual per capita income growth negatively affected by murders (1951–1991). Gaibulloev and Sandler (2008) — impact of domestic and transactional terrorism on income per capita growth for 1971–2004 in a panel of 18 Western Europe countries.
  - Univariate and multivariate time series / panel time-series methods that consider crime jointly with macro variables.
    - Examples: Detrotto and Otranto (2010) for Italy; Chen (2009) VAR for Taiwan; Narayan and Smyth (2004) ARDL Granger causality for Australia; Kumar (2013) reduced form + instrumental variables for India; Pinotti (2011) synthetic control for mafia-affected regions in southern Italy.

### Quantified economic effects of organized crime (literature synthesis)
- Primary mechanisms and empirical findings reported:
  - Organized crime primarily increases costs of doing business (Hallward-Driemeier and Stewart, 2004; Daniele and Marani, 2011).
  - Adverse channels include regional security risk assessments (Kotabe, 2005), security budgets (Spich and Grosse, 2005; Czinkota et al., 2010), extortion (Brock, 1998; Daniele and Marani, 2011), kidnapping of workers (Clegg and Gray, 2002), supply chain disruptions (Barnes and Oloruntoba, 2005; Czinkota, 2005; Globerman and Storer, 2009; Branzei and Abdelnour, 2010), and decreases in local demand via emigration, business relocation, and firm closures (Greenbaum et al., 2007).
  - Foreign direct investment (FDI) associations:
    - Brock (1998) — relatively higher FDI in Russian regions with lower crime.
    - Madrazo Rojas (2009) — negative association between violent organized crime and FDI in Mexican states.
    - Daniele and Marani (2011) — negative relationship between total regional crime and FDI in Italy.
    - Ashby and Ramos (2013) — organized crime deters FDI in financial services, commerce, and agricultural sectors, but not oil and mining (where increased crime associated with increased investment).
  - Threshold effects: Robles et al. (2013) — below a crime threshold firms/individuals internalize security costs; above the threshold firms/individuals modify investment, production, labor participation, and employment decisions with negative impacts on economic activity.
- Common crime indicator in literature: number of recorded intentional homicides.

### Novel empirical approach used in this paper
- Objective: investigate the effect of crime on economic performance in Mexican states while systematically accounting for dynamic endogeneity and heterogeneity across states.
- Key methodological choices:
  - Avoid conventional dynamic panel methods that assume homogeneous dynamics across states.
  - Employ panel structural VAR (panel SVAR) methodology that allows heterogeneous state-specific dynamic responses to structural shocks (method developed in Pedroni (2013)).
  - Decompose unobserved structural shocks into common (national/international) and idiosyncratic (state-specific) components; permit the relative importance of common vs idiosyncratic shocks to differ by state.
  - Represent responses via impulse responses and importance via dynamic variance decompositions, producing state-specific impulse responses and variance decompositions for each of the 32 Mexican states.
  - Use bootstrap estimation to produce confidence bands for impulse responses, variance decompositions, and for subsequent analysis relating state characteristics to response patterns.
  - Panel SVAR references and prior applications: Pedroni and Verdugo (2011); Mishra et al. (2014).

### Estimation and identification strategy (structure and restrictions)
- Variables of interest and economic reasoning:
  - Primary variables: GDP and crime (homicides), plus international factor movements: FDI and migration.
  - Four broad structural shock categories:
    - Aggregate supply shocks (AS): conceptualized as shocks that permanently increase or decrease total economic activity.
    - Aggregate demand shocks (AD): conceptualized as shocks that have only transitory effects on total economic activity.
    - Crime shocks: perturbations in supply or demand for criminal activity for a given level of economic activity (conceptualized as shocks to equilibrium levels of crime).
    - Migration shocks: shocks to either supply or demand for net migration for a given level of economic activity.
- Variable definitions used in the VAR:
  - GDP: log per capita state GDP.
  - FDI: log per capita cumulative FDI state inflows.
  - Crime: log per capita state homicides.
  - Migration: log per capita international net migration into the state.
- Identification restrictions (mix of short-run and long-run):
  - AD shocks: restricted not to cause permanent movements in GDP (long-run exclusion).
  - AS shocks: permitted to cause permanent movements in GDP.
  - Crime shocks: allowed to have an immediate impact on homicides (short-run restriction); economic activity (GDP) may be impacted immediately as well.
  - Migration shocks: migration can respond quickly to crime shocks, while homicides change more slowly in response to migration shocks (short-run restriction).
  - Short-run and long-run exclusions identify shocks up to sign; signs are fixed via economically motivated rules:
    - Crime shock positive if it increases homicides in the impact period.
    - Migration shock positive if it increases net international in-migration in the impact period.
    - AS shock positive if it permanently increases GDP in the long-run steady state.
    - AD shock positive if it increases GDP in the short-run impact period.
- Identification specifics:
  - The VAR system includes six exclusion restrictions on short-run or long-run responses and four sign restrictions.
  - Structural shocks are assumed orthogonal.
  - In the panel framework, each structural shock has both a common and idiosyncratic component (total of eight structural shocks), yielding for each state 32 impulse responses (four variables × eight shocks), and for all 32 states a rich sample distribution of state-specific responses.
  - Robustness: key findings checked against viable variations in identification; blocks of the system can be investigated separately (e.g., bivariate structural VAR of homicides and GDP).

### Data sources, measurement choices, and sample period
- Geographic and temporal coverage:
  - Mexican states (32 states), sample period generally 1993–2012 for state-level analysis.
- Real GDP (RGDP) construction challenges and two approaches (for 1993–2012 with base year 2008):
  - INEGI published RGDP aggregated data for 2003–2012 with base year 2008; INEGI had aggregated RGDP 1993–2003 with base year 1993 but not rebased to 2008.
  - RGDP-1: link aggregated RGDP series (1993–2003 base 1993) and (2003–2012 base 2008) by computing for year 2003 the ratio (aggregate RGDP base 2008)/(aggregate RGDP base 1993) and multiply the base-1993 series by this ratio to obtain 1993–2002 series with base 2008; does not account for heterogeneous ratios among RGDP sectors.
  - RGDP-2: apply the above linking method at the GDP sector level (compute ratio for each GDP sector), estimate sector values with base 2008 for 1993–2002, then sum sector values to obtain aggregated RGDP with base 2008 for 1993–2002. Technical Appendix 1 contains methodological description and robustness checks. For sectoral analyses in second stage, attention restricted to 2003–2012.
- Crime measures and rationale for primary crime variable:
  - Primary crime variable: number of recorded intentional homicides (chosen for highest reliability among crime variables).
  - Homicide rates computed as reported intentional homicides (INEGI) by state and year divided by population and multiplied by 100,000 (population at mid-year).
  - Sources for homicide data:
    - INEGI: public health records (death certificates) — compiled death by homicide; used to construct the primary INEGI homicide series for 32 states during 1993–2012.
    - SESNSP (Executive Secretary of the National System for Public Security): law enforcement homicide data from criminal investigations (available monthly in recent years).
  - Measurement caveats:
    - INEGI and SESNSP differ noticeably due to timing and methodology of classification; series are closely correlated but distinct.
    - Homicide data cannot distinguish whether a case is due to organized crime versus other causes (social conflict, demographic phenomena, etc.).
    - Some evidence suggests sudden increases in homicides in Mexico (2007–2012) may be attributable largely to organized crime groups, but there is no formal legal definition for “drug-related homicide” in Mexican criminal law.
    - INEGI homicide series found to follow a unit root at the state level (see stationarity discussion below); other SESNSP crime measures (homicides, violent crimes, organized crime, crimes against health) appear stationary.
  - Additional crime incidence measures available (SESNSP) for analyses of heterogeneity: incidence of crimes against health (drugs), Federal Law Against Organized Crime (LFCDO), kidnapping, extortion, aggravated assault, robbery, and classification by fuero comun (local) vs fuero federal (federal).
- Migration data:
  - Source: Consejo Nacional de Poblacion (CONAPO) state-level demographic data 1990–2010 and forecasts 2011–2030; includes population, internal and international immigration/emigration and related demographic measures.
  - Per capita international net in-migration treated carefully: series can be positive or negative, converted to logs by adding (largest negative value + 1.0) to all values to ensure positivity prior to log transform; this affects fixed effects but not dynamic impulse responses or variance decompositions.
- Foreign direct investment (FDI):
  - Source: Mexico’s Secretary of Economy — nominal state-level FDI by sector, origin, and investment type (currency in U.S. dollars).
  - Three investment types: new investments, reinvestment of profits, intercompany accounts.
  - Real FDI constructed by deflating nominal FDI with U.S. consumer price index.
  - For econometric treatment, flows found stationary; analysis uses accumulated net FDI (per capita stock) and logs constructed similarly to migration where necessary.
  - Caveat: FDI statistics are recorded in the home of the company’s headquarters, which may differ from the state where investment occurred; this can bias common vs state-specific shock attribution.
- State-specific characteristics used to explain heterogeneous responses (variables and rationales):
  - Average schooling years (INEGI Statistical Yearbook By State, 2003–2012).
  - Unemployment rates (INEGI ENOE, 2003–2012).
  - Bank deposits (Comisión Nacional Bancaria y de Valores, CNBS, 2000–2010) — size of deposits as proxy related to laundering/financial capacity.
  - Marginalization indices (CONAPO) — five-year periods since 1990, based on literacy, dwellings without drainage/toilet, households without power/piped water/overcrowding, percentage with income up to two minimum wages.
  - Perception of lack of public security (INEGI victimization/perception surveys: ENSI 2005 and 2009; ENVIPE 2010–2012) — percentage reporting “secure” used as measure.
  - Government spending on security (Mora 2009) — spending in years 2007, 2008, 2009 across prevention, law enforcement, social reintegration, judiciary, ombudsman, and Fund for Public Safety (FASP) contributions.
  - Informality (ENOE) — occupation rate in the informal sector.
  - Disaggregated GDP sectors and disaggregated FDI sectors (sector lists summarized in appendix Table 2 and Secretaría de Economía (2014) respectively) used for sectoral analyses (where second-stage analyses restrict attention to 2003–2012).
- Data period and specific series decisions:
  - Main analysis period: 1993–2012 for state-level panel.
  - Primary crime measure: INEGI recorded intentional homicides (1993–2012).
  - Migration: log per capita international net in-migration (transformed to handle negative values).
  - FDI: log per capita accumulated net FDI (real, deflated by U.S. CPI).

### Stationarity and variable selection implications for identification and interpretation
- Panel unit root testing (Im, Pesaran, and Shin (2003) tests):
  - Log per capita GDP: unable to reject unit root for each Mexican state — GDP likely follows a unit root process at the state level (implication: shocks can permanently lower or raise state-level GDP).
  - INEGI per capita homicides: also found to follow a unit root process at the state level — suggests some shocks can have long-run effects on homicide rates and thus can be the channel for long-run relationships with GDP.
  - Other crime measures (SESNSP homicides, violent crimes, organized crime, crimes against health): appear stationary — shocks to these series are transitory and cannot explain long-run changes in GDP that follows a unit root.
    - Implication: if there are long-term causal effects between crime and economic activity measured by GDP, they are likely reflected in INEGI homicide data rather than the stationary SESNSP series.
  - Migration: most internal migration measures stationary; per capita international net in-migration appears to follow a unit root at state level (so possible long-run effects).
  - FDI: flow variables stationary; accumulated net FDI (stock) used and logged for analysis.
- Consequent model specification:
  - Four-variable VAR specification uses log per capita INEGI recorded homicides, log per capita international net in-migration, log per capita GDP, and log per capita accumulated net FDI.
  - Four structural shock categories: crime shocks, migration shocks, AS shocks, AD shocks.
  - Identification and interpretation of long-run vs short-run effects depends on stationarity properties above and on the chosen short-run and long-run restrictions.

### Limitations and robustness considerations (as noted by authors)
- Measurement error in data (especially crime data) can limit inference.
- Structural VAR interpretations depend on the chosen identification scheme and on using homicide data in part to capture crime shocks; results are not intended to be conclusive.
- Sensitivity to identification: authors confirm key findings are robust to viable variations in identification scheme and analyze subsystems (e.g., bivariate homicides–GDP) for robustness.
- Caveats regarding FDI recording at company headquarters and potential upward bias in common shock responses at expense of state-specific shocks.
- Homicide series classification issues in Mexico: some events initially recorded as intentional homicides may later be reclassified (suicide, accident, suspicious death, natural death), a source of measurement error.
- SESNSP incidence data reflect initiated investigations and may not capture judicial outcomes; broad categorical groupings (e.g., “other crimes”) limit specific attribution.

_Italic source: _wp15121 - conclusion regarding the impact of crime on economic growth. A way to measure the crowding-out effect_

### section III.

### section III

### Variance decompositions — relative importance of shocks
- Median percentage of GDP variation explained by state level aggregate supply (AS) shocks is roughly 70 percent.
- Median percentage of variation in state GDP driven by state-specific crime shocks is roughly five percent, with another half a percent driven by common national or international crime shocks.
- Seventy-fifth percentile: state-specific crime shocks account for as much as 10 percent and national or international crime shocks another one percent (combined up to 11 percent for at least a quarter of states).
- Comparison: these crime-shock shares are roughly comparable to the percentage variation in state GDP attributable to economic aggregate demand (AD) shocks (private sector demand changes and government spending).
- Migration:
  - For at least a fourth of states, state-specific crime shocks initially account for 10 percent of the variation in migration, rising to close to 20 percent five years after the shock.
- FDI:
  - Median state variation explained by crime shocks is roughly one percent.
  - For states in the top quartile, up to eight percent of the variation is driven by crime shocks in the initial period, rising to as much as 15 percent by five years after the shock.

### Which shocks drive homicide variation
- Large cross-state spread in drivers of homicide variation.
- In top quartile states, over 10 percent of the variation in crime can be attributed to state-specific migration shocks.
- At longer horizons, a little over three percent each of crime variation can be attributed to state-specific AS shocks and AD shocks.

### Impulse response analysis — signs, magnitudes, duration
- Quantile impulse response estimates computed for a general four-variable VAR; figures reported for idiosyncratic and common shocks.
- Robustness checks:
  - Five methods examined for chaining state-level GDP estimates into a single base year (1993–2012); key results qualitatively invariant to chaining method.
  - Subsystems (two- and three-variable blocks) examined; primary results hold across these specifications.
  - Two-variable VAR (log per capita homicides and log per capita GDP) estimated for 2003–2012 as additional robustness check; results representative of both 1993–2012 and 2003–2012.
  - Inclusion of dummy years 2008 and 2009 (financial crisis) had only relatively minor numerical effects on reported impulse responses (with one qualitative effect on homicides response to migration shocks).
- Example: quantile responses of log per capita GDP to state-specific crime shocks (idiosyncratic):
  - Median point estimate initially slightly negative and eventually moves to approximately -0.005.
  - Twenty-fifth percentile starts at -0.004 and drops to around -0.16 by five years after the shock.
  - Interpretation: for at least a quarter of Mexican states, a ceteris paribus crime shock reduces real per capita GDP by 1.6 percent per year after five years following the shock.

### Scaling shocks and interpretation
- Median state homicide rate increases by 0.2 percent in the initial period in response to the crime shock.
  - A crime shock that induces an initial 0.2 percent increase in homicides causes a 0.5 percent decrease in GDP for the median state.
- Median state accumulated FDI decreases by 0.1 percent by the fifth year following the crime shock.
  - A crime shock that decreases accumulated FDI by 0.1 percent over five years for the median state causes a 0.5 percent decrease in GDP for the median state.
- Clarification: crime shocks are identified within the structural VAR using homicides (among other variables); homicide rates are not a proxy equating to crime shocks.

### Statistical significance and heterogeneous responses
- Bootstrapped 90 percent confidence bands computed using 5,000 bootstrap draws to assess heterogeneity significance.
- Example findings with 90 percent confidence (equivalent to significance at the 10 percent level or better):
  - For at least a quarter of states, log per capita GDP decreases in response to crime shocks; for at least a quarter of states, log per capita GDP increases.
  - For the 25 percent or more states where log per capita GDP falls in response to a crime shock: annual GDP falls by at least half a percentage point, and this decrease appears to be permanent.
  - For at least 25 percent of states where GDP rises, the increase is small and not necessarily greater than around a fifth of a percent of GDP.
- FDI responses:
  - Median state cumulative FDI response point estimate is slightly negative.
  - For at least 25 percent of states, by the fifth year following the crime shock cumulative FDI has decreased by somewhere between one percent and six percent.
  - For at least 25 percent of states, FDI increases following a crime shock.
- Prevailing pattern: regardless of median sign, in each response there exists a group of states with statistically significant positive or negative responses at the 10 percent level or better.

### Common (national/international) shocks — magnitudes
- Median state-level responses to common national/international shocks during the sample period appear smaller in magnitude:
  - Most negatively affected quantile of states: around a one-third percent decrease in per capita GDP.
  - Median state: around a one-quarter percent decrease in per capita GDP.
- Implied aggregate national GDP response to a similar common shock:
  - Roughly one-quarter percent of national GDP for up to two years, decreasing in the third year, and dissipating by the fourth year.
- Note: magnitudes scale with the realized size of shocks; aggregation of state-level responses to state-level crime shocks has a larger overall effect on national GDP.

### Second-stage analysis — linking heterogeneity to state characteristics
- Estimation approach:
  - Use a bootstrapped estimator for the regression correlation between state-specific impulse responses and measured state characteristics.
  - For each of 5,000 bootstrap draws: compute impulse responses and regress them on state characteristics; collect 5,000 slope estimates to form sampling distribution and 90 percent confidence bands for the regression correlation estimator.
- State characteristics considered (described in section IV):
  - Average schooling years, unemployment rates, bank deposits, marginalization indices, perceptions of public security, government spending on security, measure of economic informality.
  - Concentration of different economic sectors as a percentage of state GDP, concentration of sectors to which FDI flows, various crime and migration measures.
  - When multiple observations over time exist for a state, the sample average is used; economic quantities used in log per capita form.
- Robust and consistent findings across specifications:
  - Two particularly robust relationships identified; chief among them:
    - Association between the response of GDP to crime shocks and the percentage of GDP devoted to construction.
      - Scatterplot interpretation: vertical axis = percentage of state GDP devoted to construction; horizontal axis = response of log per capita GDP in the second year following the crime shock, scaled by the first period homicide response to the crime shock.
      - Positive association: in states where construction is more prominent, a crime shock is more likely to have a positive (or less negative) effect on overall state-level GDP.
- Notes and context:
  - Construction GDP classification reference provided in original text (not reproduced here).
  - Literature mentions potential vulnerability of real estate sector to misuse by criminal organizations; referenced reports and summaries described in source text.

*Source: section III.*

### 1.091 for the GDP-1 based VAR and 0.924 for the GDP-2 based VAR. The bootstrapped ninety-fifth

### VI.   SUMMARY AND CONCLUSIONS

### Key empirical findings: VAR first-stage and bootstrapped estimates
- GDP-1 based VAR estimate: 1.091.
- GDP-2 based VAR estimate: 0.924.
- Bootstrapped ninety-fifth percentile for GDP-1 based VAR: 2.065.
- Bootstrapped fifth percentile for GDP-1 based VAR: 0.112.
- Bootstrapped ninety-fifth percentile for GDP-2 based VAR: 1.944.
- Bootstrapped tenth percentile for GDP-2 based VAR: 0.112.
- Interpretation: "we have greater than 90 percent confidence that the relationship is positive and statistically significant" (GDP-1), and a similar but slightly wider confidence band for GDP-2.

### FDI responsiveness and public security (second-stage relationships)
- Scatterplot analysis links the second period response of FDI to a crime shock (scaled by the effect of the crime shock on homicides) with the measure of the sense of public security at the state level.
- Relationship interpretation:
  - Greater sense of security → per capita accumulated FDI response is less negative to a crime shock.
  - Higher sense of insecurity → crime shock more likely to induce a larger cumulative decrease in FDI.
- Bootstrapped median estimator for the slope:
  - GDP-1 based VAR: 0.006.
  - GDP-2 based VAR: 0.005.
- Bootstrapped ninety-fifth percentile for the slope:
  - Both GDP-1 and GDP-2 based VARs: 0.01.
- Bootstrapped fifth percentile values for the slope:
  - GDP-1: 0.002.
  - GDP-2: 0.001.
- Interpretation: "we are 90 percent confident in both cases that the relationship is statistically significant and positive."

### Robustness checks and auxiliary statistics
- Inclusion of dummies for states with major international seaports or with borders to the U.S.:
  - Relationship between GDP response to crime shocks and prevalence of construction: raw unbootstrapped t-statistic changed from 1.837 to 1.891 (slightly more significant).
  - Relationship between FDI response to crime shocks and sense of public security: raw unbootstrapped t-statistic changed from 2.096 to 2.094 (virtually unchanged).
- Regulatory note: "at the end of 2012, on the basis of the Bill for a Federal Law for the Prevention and Identification of Transactions with Criminal Proceeds, Mexico forbade cash payments of more than a half million pesos ($38,750) for real estate transactions."

### Aggregate and heterogeneity results across states (1993–2012 sample)
- Sample period: 1993–2012.
- Overall conclusion: crime is a relevant factor for many state economies in Mexico; relationships with economic activity (GDP, FDI, migration) are inter-related but heterogeneous across states.
- Median-state quantitative result:
  - A typical state-specific crime shock induces a roughly one-half percent temporary decrease in per capita GDP which persists up to two years after the shock before gradually dissipating after the third year.
- Most negatively affected quantile:
  - "for the most negatively affected quantile of states, we can say with 90 percent confidence that a typical crime shock induces at least one-half percent decrease in GDP that is permanent and does not dissipate."
- Factor-flow dynamics:
  - International migration and FDI responses to crime shocks vary substantially across states, with a substantial number of states experiencing factor flows in opposing directions.

### State characteristics associated with heterogeneous responses
- Construction share (percentage of GDP devoted to construction):
  - Associated with the size and magnitude of the economic response to crime.
- Sense of public security (survey measure):
  - Associated with the responsiveness of FDI to crime shocks; higher perceived security reduces the negative impact of crime shocks on FDI and overall economic activity.

### Policy-relevant implications and reforms
- Perception and public confidence in institutions that provide public security are potentially important in mitigating the economic impact of crime.
- Institutional reform context cited:
  - Judicial reform at the federal level in Mexico was approved by Congress in 2008 and "it is expected that this reform will be in place in all states by 2016."
- AML/CFT context:
  - "current efforts toward a further strengthening of the AML/CFT regime, particularly in areas of high ML/TF related risks, might also help to further improve future potential economic growth prospects for Mexico."

### Limitations and scope
- Time coverage: analysis covers 1993 through 2012 and "does not account for the reforms that have occurred in the period since then."
- Geographic focus: analysis is primarily intended to focus on the state level.

*Source: VI. SUMMARY AND CONCLUSIONS, paper covering 1993–2012.*

### 2013. Justice in Mexico Project: Department of Political Science and International Relations (San Diego:

### _wp15121 - 2013. Justice in Mexico Project: Department of Political Science and International Relations (San Diego:

### Literature and data sources cited
- Mexico has rebased the real GDP at the national and state levels in 1993, 2003, and 2008.
- Bibliographic and data sources listed include (verbatim titles and years as cited):
  - Im K, Pesaran H, Shin Y., 2003, “Testing for unit roots in heterogeneous panels,” Journal of Econometrics 115: 53–74.
  - International Monetary Fund, 2001, “Financial System Abuse, Financial Crime and Money Laundering—Background Paper” (Washington, International Monetary Fund).
  - International Monetary Fund, 2011a, “Anti-Money Laundering and Combating the Financing of Terrorism (AML/CFT)—Report on the Review of the Effectiveness of the Program,” http://www.imf.org/external/np/pp/eng/2011/051111.pdf (Washington, International Monetary Fund).
  - International Monetary Fund, 2011b, “Acting Chair’s Summing Up, Anti-Money Laundering and Combating the Financing of Terrorism (AML/CFT)—Report on the Review of the Effectiveness of the Program,” https://www.imf.org/external/pubs/ft/sd/index.asp?decision=EBM/11/55 (Washington, International Monetary Fund).
  - International Monetary Fund, 2013, Anti-Money Laundering and Combating the Financing of Terrorism—Inclusion in Surveillance and Financial Stability Assessments—Guidance Note, http://www.imf.org/external/np/pp/eng/2012/121412a.pdf (Washington, International Monetary Fund).
  - Instituto Nacional de Estadística y Geografía, 1993–2000, 2003, 2006, 2007–2011, 2008, 2012, 2013, Anuario de Estadisticas por entidad federativa Datos estadísticos (Mexico City).
  - Instituto Nacional de Estadística y Geografía, 2006, Sistema de Cuentas Nacionales de Mexico. Producto Interno Bruto por Entidad Federativa. Metodologia.
  - Instituto Nacional de Estadística y Geografía, 2010, Series del PIB Trimestral a Precios Constantes, Retropolacion para el periodo 1993–2002, Ano base 2003.
  - Instituto Nacional de Estadística y Geografía, 2011, Marco Concepctual del Censo de Poblacion y Vivienda 2010.
  - Instituto Nacional de Estadística y Geografía, 2011, 2012, ENVIPE-Encuesta Nacional de Victimizacion y Percepcion sobre Seguridad Publica.
  - Instituto Nacional de Estadística y Geografía, 2013a, Sistema de Cuentas Nacionales de Mexico, Cuentas de Corto Plazo y Regionales: Fuentes y Metodologias.
  - Instituto Nacional de Estadística y Geografía, 2013b, Modernizacion de las Cuentas Nacionales Cambio de Base de 2003 a 2008.
  - Instituto Nacional de Estadística y Geografía, 2013c, Sintesis metodológica del Sistema de Clasificacion Industrial de America del Norte, Mexico, SCIAN.
  - INEGI and STPS (Ministry of Labour and Social Welfare), 2005, 2007, 2009–2014, Encuesta Nacional de Ocupación y Empleo. Available via the Internet: http://www.inegi.org.mx/est/contenidos/Proyectos/encuestas/hogares/regulares/enoe/
  - Institute for Economics and Peace, Mexico Peace Index. Available via the Internet: http://www.visionofhumanity.org/#/page/indexes/mexico-peace-index
  - Koopman, Siem J., Neil Shephard, and Jurgen A. Doornik, 1999, “Statistical Algorithms for Models in State Space Using SsfPack 2.2,” Econometrics Journal, Vol. 2, No. 1, pp. 107–160.
  - Kotabe, M., 2005, “Global security risks and international competitiveness,” Journal of International Management, Vol. 11, pp. 453–455.
  - Kumar, S., 2013, Crime and Economic Growth: Evidence from India. Available via the Internet: http://mpra.ub.uni-muenchen.de/id/eprint/48794
  - La Jornada, 2008 and 2012 (specific articles cited with URLs).
  - Lin, M.J, 2009, “More Police, Less Crime: Evidence from U.S. State Data,” International Review of Law and Economics, Vol. 29, No. 2, pp. 73–80.
  - Lochner, Lance and E. Moretti, 2004, "The Effect of Education on Crime: Evidence from Prison Inmates, Arrests, and Self-Reports," American Economic Review, Vol. 94, March, No. 1, pp. 155–189.
  - Londoño, Juan Luis and Rodrigo Guerrero, 1999, “Violencia en America Latina: epidemiología y costos,” Working Paper R-375 (Washington, D.C., United States: Banco Interamericano del Desarrollo).
  - Lopez, R. and S. Velarde, “Una aproximación a los patrones de migración interregional en Mexico, 1990–2010,” La situación demográfica de México 2013. Available via the Internet: http://www.conapo.gob.mx/es/CONAPO/Una_aproximacion_a_los_patrones_de_migracion_interregional_en_Mexico_1990-2010
  - Madrazo Rojas, F.M., 2009, “The effect of violent crime on FDI: the case of Mexico 1998–2006,” unpublished Master’s thesis, Georgetown Public Policy Institute, Georgetown University, Washington, D.C.
  - Mauro, L. and G. Carmeci, 2007, “A poverty trap of crime and unemployment,” Review of Development Economics, Vol. 11, August, No. 3, pp. 450–462.
  - McCollister, K., M. French, and H. Fang, 2010, “The cost of crime to society: New crime-specific estimates for policy and program evaluation,” Drug and Alcohol Dependence, Vol. 108, Nos. 1–2, pp. 98–109.
  - McDonald, J., 2005, “The narcoeconomy and small-town, rural Mexico,” Human Organization, Vol. 64, No. 2, pp. 115-125.
  - McLennan, W., 1998, “Introduction of Chain Volume Measures in the Australian National Accounts, Information Paper 5248.0 (Canberra, Australia: Australian Bureau of Statistics).
  - Mejia, D. and P. Restrepo, 2010, “Crime and conspicuous consumption,” Working Paper CEDE No. 2010-32 (Bogota, Colombia).
  - Merino, José, 2011, “Los operativos conjuntos y la tasa de homicidios: Una medicion,” Nexos, June 1. Available via the Internet: http://www.nexos.com.mx/?p=14319
  - Miguel, Edward, Shanker Satyanath, and Ernest Sergenti, 2004, “Economic Shocks and Civil Conflict: An Instrumental Variables Approach,” Journal of Political Economy, Vol. 112, No. 4, pp. 725–753.
  - Millo, Giovanni and Gianfranco Piras, 2012, “splm: Spatial Panel Data Models in R,” Journal of Statistical Software, Vol. 47, April, No. 1, pp. 1–38.
  - Mishra, Pratchi. Peter Montiel, Peter Pedroni and Antonio Spilimbergo, 2014, “Monetary Policy and Bank Lending Rates in Low-Income Countries: Heterogeneous Panel Estimates,” Journal of Development Economics, 111, 117-131. (IMF working paper version, available via the Internet: http://www.prachimishra.net/mmps_draft_march29-final-withappendix.pdf)
  - Molzahn, Cory, Octavio Rodriguez Ferreira, and David A. Shirk, 2013, Drug Violence in Mexico Data and Analysis Through 2012 (San Diego, California: University of San Diego).
  - Mora, Maestro Carlos Mendoza, 2009, Encuestas Nacionales sobre Inseguridad, “El costo de la inseguridad en México,” Cuadernos del ICESI no.5. Available via the Internet: http://www.seguridadcondemocracia.org/encuestas/estadisticas/cuadernos-del-icesi-5.-el-costo-de-la-inseguridad-en-mexico.html
  - Moran, Jon, 2011, Crime and Corruption in New Democracies (Basingstoke: Palgrave Macmillan).
  - Morris, Stephen D., 2012, “Corruption, Drug Traffcking, and Violence in Mexico,” Brown Journal of World Affairs, Vol. 18, No. 2, pp. 29–43.
  - Narayan, Paresh Kumar and Smyth, Russell, “Dead Man Walking: An Empirical Reassessment of the Deterrent Effect of Capital Punishment Using the Bounds Testing Approach to Cointegration,” No 1028, American Law & Economics Association Annual Meetings.
  - National Drug Intelligence Center, 2010, Drug Trafficking Organizations. Available via the Internet: http://www.justice.gov/ndic/pubs38/38661/dtos.htm#Top.
  - Organisation of Economic Co-operation and Development, 2000, “Basic Principles and Practices in Rebasing and Linking National Accounts Series,” paper presented at Joint ADB/ESCAP Workshop on Rebasing and Linking of National Accounts Series, Bangkok, Thailand, March 21–24 2000.
  - Osorio, Javier, 2012 and 2013 (papers and mimeo cited).
  - Pedroni, Peter, 2013, “Structural Panel VARs,” Econometrics Vol. 2, pp. 180–206.
  - Pedroni, Peter and Concepcion Verdugo, 2011, “The Relationship between Illicit Coca Production and Formal Economic Activity in Peru,” IMF Working Papers 11/182 (Washington, International Monetary Fund).
  - Peri, G., 2004, “Socio-cultural variables and economic success: evidence from Italian provinces 1951–1991,” Topics in Macroeconomics, Vol. 4, pp. 1–34.
  - Pinotti, P, 2011, “The economic consequences of organized crime: Evidence from Southern Italy,” Working Paper (Rome, Italy: Banco de Italia).
  - Poiré, Alejandro, 2011; Poiré, Alejandro and Maria Teresa Martínez, 2011 (Nexos articles cited).
  - Prasad, Kislaya, 2012, “Economic Liberalization and Violent Crime”, Journal of Law and Economics, vol. 55, issue 4, pages 925–948.
  - Ramirez, T. and D. Aguado, 2013, “Determinantes de la migracion de retorno en Mexico, 2007–2009,” La situación demográfica de México 2013.
  - Revista Proceso, 2013, “Zacatecas, tierra de nadie...y de cadáveres,” January.
  - Rios, Viridiana, multiple works 2008–2012 cited (working papers, articles, dissertation).
  - Ríos, Viridiana and David A. Shirk, 2011, Drug Violence in Mexico: Data and Analysis Through 2010 (San Diego, California: University of San Diego).
  - Robles, Gustavo, Gabriela Calderon, and Beatriz Magaloni, 2013, “The economic consequences of drug violence in Mexico,” Inter-American Development Bank Working Paper No. 83734 (Stanford, California: Inter-American Development Bank).
  - Romo Viramontes, Raúl, Yolanda Téllez Vázquez, and Jorge López Ramírez, 2013, “Tendencias de la migración interna en Mexico en el Periodo Reciente.”
  - Rosas, G., 2011, “Los números de la guerra: ¿error de cálculo político o error de cálculo estadistico?” Nexos.
  - Ross, Marc H., and Elizabeth Homer, 1976, “Galton’s Problem in Cross-National Research,” World Politics, Vol. 29, No. 1, pp. 1–28.
  - Schelling, Thomas, 1967 and 1971 (works cited).
  - Secretaría de Economía, 2014, Estadística oficial de los flujos de IED hacia México.
  - Secretariado Ejecutivo del Sistema Nacional de Seguridad Publica, 2014, “Incidencia Delictiva del Fuero Federal (1997-2012).” and “Incidencia Delictiva del Fuero Comun (1997-2012).”
  - Sheridan, M. B., 2010, “Mexico’s Calderon tells Congress he needs U.S. help in fighting drug wars,” Washington Post, May 21.
  - Skaperdas, Stergios, 2002, “Warlord Competition,” Journal of Peace Research, Vol. 39, No. 4, pp. 435–446.
  - Shirk, David A, 2010, Drug Violence in Mexico Data and Analysis from 2001–2009 (San Diego, California: University of San Diego).
  - Shirk, David A. and Alejandra Rios Cazares, 2007, Reforming the Administration of Justice in Mexico (Notre Dame, Indiana: Notre Dame University Press).
  - Signoret, P., 2011, “Captura de jefes del narcotráfico y niveles de violencia en México,” Tesis de Licenciatura (Mexico City, Mexico: Instituto Tecnológico Autónomo de México).
  - Sistema de Cuentas Nacionales de Mexico, 2001 and 2006 (Producto Interno Bruto por Eutidad Federativa series).
  - Soares, R., 2009, “Welfare cost of crime and common violence: A critical review,” The Cost of Violence (Washington: World Bank).
  - Stanger, Michael V., 2007, “Empalme del PIB y de los componentes del Gasto: Series Anuales y Trimestrales 1986–2002, Base 2003,” Studies in Economic Statistics, No. 55.
  - Stratfor, 2012, “Polarization and sustained violence in Mexico’s cartel war,” January 24.
  - Unger, Brigitte et al., 2006, “The Amounts and the Effects of Money Laundering,” Report for the Ministry of Finance, Utrecht School of Economics, February 26, 2006.
  - United Nations Statistics Division, Review of Country Practices on Rebasing and Linking National Accounts Series.
  - U.S. Department of Justice, 2011, National Drug Threat Assessment, 2011, U.S. Department of Justice National Drug Intelligence Center, Washington, D.C.
  - United Nations Office on Drugs and Crime, 2014, Global Study on Homicide on 2013, Vienna.
  - Villarreal, A., 2002, “Political competition and violence in Mexico: Hierarchical social control in local patronage structures,” American Sociological Review, Vol. 67, No. 4, pp. 477–498.
  - Waters, H., A. Hyder, Y. Rajkotia, S. Basu, and J.A. Rehwinkel, 2004, The Economic Dimensions of Interpersonal Violence (Geneva, Switzerland: World Health Organization).

### Technical appendix — chain linking two real GDP series in different base years
- Purpose and context:
  - Mexico has real and nominal data, implicit prices and volumes for GDP sectors, and totals for years 1993–2006 using the 1993 base year and the 1993 accounting system for all the Mexican states.
  - Mexico also has the same data for years 2003–2012 using the 2008 base year and the 2008 accounting system for all the states.
  - Although the 2008 accounting system contains more subsectors than the 1993 accounting system, both accounting systems have three sectors in common: primarias (Sector 1), secundarias (Sector 2), and terciarias (Sector 3).
  - Overlapping years for both base years and both accounting systems are 2003-2006.
  - To compare the real GDPs with different bases and different accounting systems, the series need to be chain linked.

- Notational definitions and identities (preserved verbatim from source):
  - For t=1993,...,2006, let )(93 tQ i be the real value of sector i at year t for each Mexican State, using the 1993 accounting system and the base year 1993.
  - Let )(93 tP i be the implicit price of sector i at year t, using base year 1993 and the 1993 accounting system.
  - Note that 100)1993( 93 = iP for all sectors i since the base year is 1993.
  - The i-th sector component of the real GDP:
    - RGDP )( )1993( 93 t = Σ_{i=1}^{3} 93 )( i tQ ,    t=1993,...,2006,
      - where the superscript denotes the 1993 accounting system and the subscript denotes the base year.
  - The nominal GDP using the 1993 accounting system and base year 1993 is:
    - NGDP )( )1993( 93 t = Σ_{i=1}^{3} 100 ),()( 93 i i tP tQ    t=1993,...,2006.
  - In this expression, 100/)()(
    9393 tPtQ ii is the i-th sector component in the NGDP and 100/()( 93 tP i is the i-th sector's relative price at year t with respect to that in year 1993.
  - The above equation relates sectoral RGDP to the aggregated NGDP: )(93 tQ i acts as quantity in calculating the aggregated volume NGDP and the relative price 100/()( 93 tP i as the corresponding price (ABS(1998, page 7)).
  - Re-basing to 2003-base NGDP using the same accounting system:
    - As usual, 100)2003( 03 = iP.
    - Thus )2003( )(100 )(93 93 03 i i i P tP tP = for t=1993,1994 ,..., (text continues into derivation).

- Methodological note:
  - The approach follows McLennan (1998) for chaining GDP series (explicit reference in source).
  - The appendix sets up the formulas needed to re-base and chain-link series that share three aggregate sectors across accounting systems and base years.

*Italic source attribution line: _wp15121 - 2013. Justice in Mexico Project: Department of Political Science and International Relations (San Diego:_*

### 2006.  Applying the Laspeyres method (ABS(1998, page 8)), the 2003-based NGDP for years 1993–2006 in

### _wp15121 - 2006.  Applying the Laspeyres method (ABS(1998, page 8)), the 2003-based NGDP for years 1993–2006 in

### Re-basing method and formulas
- Applying the Laspeyres method yields the 2003-based NGDP for t=1993,...,2006 under the 1993 accounting system:
  - Formula (1): NGDP(1993)(03)t = 100 * sum_{i=1}^{93} Q_i^{1993} * (P_i^{t}/P_i^{1993}), t=1993,...,2006.  
  - Key point: change in rebasing replaces the relative price at year t over 1993 with the relative price at year t over 2003; quantities remain the same when rebasing NGDP under the same accounting system.
- Re-basing 2008-based NGDP to 2003 base under the 2008 accounting system:
  - Formula (2): NGDP(2008)(03)t = 100 * sum_{i=1}^{08} Q_i^{2008} * (P_i^{t}/P_i^{2008}), t=2003,...,2012.
- Using common year 2003 to bridge accounting systems:
  - For t=2003, NGDP(1993)(03) and NGDP(2008)(03) are expressed in terms of 2003 quantities Q_i^{2003}.
  - Define R_i = Q_i^{2003(1993)}/Q_i^{2003(2008)} (notation in source: R = )2003(93)Q / )2003(08)Q).
  - Natural extension of NGDP(2008)(03)t to t=1993,...,2002:
    - NGDP(2008)(03)t = 100 * sum_{i=1}^{93} (P_i^{t} * Q_i^{2003(2008)} * R_i), t=1993,...,2002.

### Converting to 2008 base and deriving RGDP
- After re-basing from 2003 to 2008 under the 2008 accounting system:
  - NGDP(2008)(08)t = 100 * sum_{i=1}^{93} (P_i^{t} * Q_i^{2003(2008)} * R_i), t=1993,...,2002 (same expression given for NGDP in 2008 base and 2008 accounting system).
- Real GDP with 2008 base and 2008 accounting system:
  - Formula (3): RGDP(2008,1)t = sum_{i=1}^{93} (Q_i^{t} * R_i), t=1993,...,2003.

### Averaging across common years (2003–2006) for sectoral ratios
- Use four common years 2003–2006 to average ratios of real sectorial values:
  - Formula (4): R_i = (1/4) * (Q_i^{2003(1993)}/Q_i^{2003(2008)} + Q_i^{2004(1993)}/Q_i^{2004(2008)} + Q_i^{2005(1993)}/Q_i^{2005(2008)} + Q_i^{2006(1993)}/Q_i^{2006(2008)}).
- Using average ratios, obtain RGDP in 2008 base and 2008 accounting system for t=1993,...,2002:
  - Formula (5): RGDP(2008,2)t = sum_{i=1}^{93} R_i * Q_i^{t}, t=1993,...,2002.

### Aggregated-volume scaling approaches (no sectoral data)
- Simplified aggregated scaling: define R = RGDP(2008)/RGDP(1993) for 2003, then:
  - Formula (6): RGDP(2008,3)t = R * RGDP(1993)t, for t=1993,...,2002.
- Averaged aggregated ratios over 2003–2006:
  - Define R = (1/4) * (RGDP(2003,2008)/RGDP(2003,1993) + RGDP(2004,2008)/RGDP(2004,1993) + RGDP(2005,2008)/RGDP(2005,1993) + RGDP(2006,2008)/RGDP(2006,1993)).
  - Formula (7): RGDP(2008,4)t = R * RGDP(1993)t, t=1993,...,2002.

### Gradually adjusted/geometric rebasing approaches
- Assume ratio D = RGDP(2008)(2003)/RGDP(1993)(2003) is result of geometric development over 10 years (1993–2003):
  - Annual rebasing ratio = D^{1/10}.
  - Formula (8): RGDP(2008,5)t = D^{(t-1993)/10} * RGDP(1993)t, t=1993,...,2002 (notation in source: (2008,5)(1993)/10 ...).
- Capitalizing on all four common years 2003–2006 yields:
  - Formula (9): RGDP(2008,6)t = (1/4)*(D_{2003}^{(t-1993)/10} + D_{2004}^{(t-1993)/11} + D_{2005}^{(t-1993)/12} + D_{2006}^{(t-1993)/13}) * RGDP(1993)t, t=1993,...,2002.
  - Where D_{2004}, D_{2005}, D_{2006} are respective year ratios: D_{2004} = RGDP^{2008}(2004)/RGDP^{1993}(2004), etc.

### Log-linear fit to account for accounting changes and combined formula
- Fit log D_t = c + δ t + ε_t using t=2003, 2004, 2005, 2006 to estimate ĉ and δ̂.
- Construct D̂_t for t=1993,...,2002 using log D̂_t = ĉ + δ̂ t.
- Combined accounting-change and geometric rebasing formula:
  - Formula (10): RGDP(2008,7)t = D̂_t * RGDP(1993)t, t=1993,...,2002 (source notation: (10) t=1993,...,2002, RGDP(2008,7)(t) = D̂_t * RGDP(1993)(t)).

### Scenarios and estimation strategies summarized
- Sectoral constant-scaling: scenarios (3) and (5) — sectoral scaling ratios constant over time.
- Aggregated constant-scaling: scenarios (6) and (7) — aggregated scaling ratios constant.
- Geometric/gradual adjustment: scenarios (8), (9), and (10) — time-varying rebasing ratios using geometric progression and log-linear fits.
- Key temporal ranges used across methods:
  - Common years for linking: 2003–2006.
  - Target extension years: t=1993,...,2002 (to extend 2008-based series backward).
  - Available NGDP/RGDP original spans noted in data sources (see Data Sources).

### Robustness checks and empirical implementation
- Two linked real GDP series used in reported empirical results:
  - GDP Approach 1: linked by formula (6).
  - GDP Approach 2: linked by formula (5).
- Experimentation with other scenarios (3),(5)–(9) shows results are "quite robust, regardless of the choice of approach to link the GDP series."
- Authors applied all above real aggregated GDP approaches for each Mexican state in (3), (5)–(9) and found "very similar" results across scenarios.
- Visual comparisons across approaches are presented in Figure 9 (plots for Mexican states) in the source.

### Data, figures, and tables referenced
- Figures and illustrations in the source include:
  - Figure 9: Approaches for Linking Real GDP Series for Mexican States (plots by state, 1995–2010 timeline shown).
  - Figure 10: State Level Quantiles for Impulse Responses to Idiosyncratic and Common Shocks (Four variable system, 1993-2012, GDP approach 2).
  - Figure 11: State Level Quantiles for Variance Decompositions for Idiosyncratic and Common Shocks (Four variable system, 1993-2012, GDP approach 2).
- Tables referenced:
  - Table 1: GDP Sectors for years 1993 and 2008 (Source: INEGI 2013b and 2013c).
  - Table 2: Names of Mexican States (lists 32 state names and abbreviations including "National Nat").
  - Table 3: Data Sources (variables, ranges, and sources; includes details such as Aggregated Real GDP Base year=1993 1993-2006 INEGI; Base year=2008 2003-2012 INEGI; Rebasing to base=2008 1993-2012 See Appendix).
- Notes on data transformations:
  - Most data used in log per capita form.
  - For first stage VAR estimation, Homicides, GDP and Net In-migration were used in log per capita growth rates, and subsequent impulse responses were accumulated.

*Source: _wp15121 - 2006. Applying the Laspeyres method (ABS(1998, page 8)), the 2003-based NGDP for years 1993–2006 in (source PDF).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2015/_wp15121.pdf_