The Real And Social Effects Of Credit Market Competition

Mark J. GARMAISE and Tobias J. MOSKOWITZ
(The Journal of Finance – Vol. 61, No – April 2006 – pp. 495-538)

PURPOSE
The objective of the paper is to examine the real and social impact of local credit competition. The authors point out a link between the competitiveness of the local banking market, credit conditions, urban development, and crime.

PREVIOUS STUDIES ABOUT THE TOPIC
Some papers tie credit market imperfections to reduced real economic activity and growth. Other studies connect the economic environment to property crime.

HYPOTHESIS
The main hypothesis is that reductions in bank competition may lead to economic decline and, subsequently, increase in property crime.

ANALYSIS
The authors analyze the local market effects of mergers between nonfailing banks of at least $1billion in assets to capture exogenous changes in the competitiveness of local banking markets.
It is checked the local market effects of these mergers on subsequent loan supply, interest rates, property prices, investment activity, economic and demographic variables, and crime.

LIMITATION
They only examine the relation between changes in bank competition through large mergers and changes in economic conditions and crime across neighborhoods. Bank competition cannot explain the aggregate national decline in crime nor differences in the level of crime across regions.

RESULTS
When loan competition declines (via large bank mergers), interest rate charged on loans rises and borrowers receive smaller loans. The authors show that the same bank sets different loan conditions in areas with different levels of competition.
The impact of the mergers on local competitive environment lasts about three years, and has no effect beyond that time.
During this period (three years), commercial real estate development, and new construction activity fall significantly. Local property prices decline. Unemployment rises, median income drops, and income inequality increases. Also, the income of new arrivals into the neighborhood is below that of long-term residents, suggesting an influx of lower income households.
This changing economic and demographic environment provides a channel through which crime may rise.

The effect of mergers on property crime risk varies across neighborhoods: areas that are already characterized by high banking concentration and low median income are more severely affected by a merger-induced reduction in loan competition.

The authors believe in a causal link between bank mergers and crime. However there is an alternative explanation for these results: reverse causation. It could be the case that future crime causes in some way the increase of current mergers. The authors reject this alternative idea because commercial real estate loans are only a small part of banks’ portfolios and the future price effects correlated with mergers are small. Moreover, since the study considers only large mergers, such transactions are almost surely not driven by the possibility of future neighorhood-level declines.
The authors also investigate the hypothesis that some unobservable economic variables that affect future crime motivate current mergers.

This alternative is also rejected: while bank mergers antedate economic deterioration and increases in crime, economic decline and increases in crime are not correlated with either contemporary or future bank mergers.
The authors also test mergers of failing banks, since these are the most likely associated with economic declines in the neighborhoods in which they operate. However, they find that mergers involving failing banks do not predict future crime increases.
It is also considered mergers that should not affect the local competitiveness, e.g., when outsider banks acquire insider banks or when mergers occurr within the same bank holding company (BHC). Such mergers do not predict changes in credit conditions or increases in future crime.
In sum, this collection of results suggests that the reverse causation theory is not very plausible.

Although the study is focused on the neighborhood level, the paper includes some county- and state-level analysis to obtain additional construction and crime statistics and show how that an increase in bank mergers predicts reduced construction activity and higher burglary rate (but no effect on homicide rates).
At the state level, another measure of bank competition is used: the variation in state deregulation of bank branching. Branch deregulation had the opposite effect of mergers by improving competition (potentially more branches, more banks acting in a specific area). In the years just after the deregulation, the banking concentration and the loan rates reduced, loan sizes, property prices, and burglary rates rose (but no effect on homicide). These finds at the state-level further support the link between credit conditions, economic activity, and property crime.

I – DATA AND SUMMARY STATISTICS
A. Commercial Real Estate Market
Sample: 22,642 commercial real estate transactions in 11 states of the US from 1992 to 1999.
Properties are grouped into three types: apartments, vacant lands and building.
Table I – Panel A (p. 500) shows summary statistics where we can see the that mean (median) of the sale prices (U$) is $2.4 (0.6) million, the mean capitalization rate is 10.09% and, on average, 76% of the price of properties is obtained from loans.

B. Bank Merger Activity
Table I – Panel B (p. 500) reveals that from the 769 banks in the US, 316 were involved in mergers between 1992 and 1999, being that 80 of such banks were considered large (> $ 1 billion in assets).

C. Crime data
The data about crime come from CAP Index, Inc. that computes crime scores for specific locations based on data from police reports, the FBI’s Uniform Crime Report (UCR), client loss reports, and offender and victim surveys with geographic, economic, and population data.
Such score measure the probability that a certain crime will be committed in a given location relative to national and county levels of cirme. Hence the score 1 means that the likelihood of crime in a location is equal to the national (or county) average for the year.
Crimes are classified in two types: “against persons” (homicide, rape, robery, and aggravated assault) and “against property” (burglary, larceny, and motor vehicle theft).
The paper uses Cap Index scores for three years: 1990, 1995, and 2000.
Table I – Panel C (p. 501) shows that burglary risk (one type of crime against property) is greater than homicide risk (a crime against person) and that the probability of crime in the neighborhoods studied is higher than the national and their respective county averages (the index = 1 represents the national and county averages).

D. Census and Construction Data
The paper also uses data on income distribution, median house value, unemployment, and population (from the 1990 and 2000 US Censuses)

II – BANK MERGERS AS A MEASURE OF EXOGENOUS COMPETITION CHANGES
Mergers is an exogenous measure of banking competition. It is unrelated to any of the dependent variables used in the study (demand for financing, financing terms, prices, measures of real activity, and crime).
An example of endogeneity would be: the decline in neighborhood’s property values might lead some banks to withdraw from lending activity in the area. This would lead to inconsistent coefficient estimates.

The measure for changes in bank competition is given by large nonfailing commercial bank mergers and acquisitions. This variable is exogenous because it is likely uncorrelated with subsequent neighborhood-level variation in loan rates, prices, investment, and crime.

For each property, the authors calculated, for each year, the actual concentration measure of bank loans for its local area:

where #Bankj,yr,b is the set of properties around property j (excluding the property itself) that received a loan from bank b during year yr. Bj,yr stands for the set of distinct banks that issued debts to a property around property j in year yr.

Then the authors calculate a second measure of bank loan concentration,
They do so by recomputing the prior equation assuming that all bank mergers in year yr occurred at the beginning of the year. This creates a hypothetical local bank concentration measure that treats future merged banks as a single entity in their previous deals (for example, if two banks merge during the year, all of their deals before the merger are considered as coming from the same bank in that year).
So, this value measures a hypothetical impact on banking concentration based on loans made before the merger and assuming that other banks do not change their market shares after the merger. This measure understimate the effect of the merger since the activity of the merged bank after the merger is included in the “actual measure” but not in the hypothetical one.

The predicted change in bank competitiveness from year yr1 to yr2 cause by mergers is defined as the sum of individual year diferences between the hypothetical concentration minus the actual concentration:

Note that larger differences between those measures across properties (regions) indicate a more substantial potential impact of bank mergers on local commercial real estate loan competition.

Figure 1 (pp. 507-508) presents summary statistics on BankHI measures.

III – BANK MERGER’S IMPACT ON FINANCIAL, REAL, AND SOCIAL ACTIVITY
In table II (pp. 510-512) the authors analyze how the predicted change in bank concetration is linked to actual future loan concentration, rates terms, and size of property transactions in the sample.
Panel A indicates that bank mergers make local markets less competitive (more concentrated) over the following three years. That is, bank mergers have a significantly positive effect on subsequent changes in banking market concentration. The estimate for the period 1992-1995 is positively associated with the actual data for 1996-1999.According to Panel C, mergers in the last three years (1995-1997) worsen loan conditions, but “old” mergers seem to have no effect. Panel D has the purpose of checking some deals and mergers that should hav no competitive effect and concludes that an increase in local banking market competitiveness has no effect on loan rates, frequence, and size of loans.

In table III (p. 517) the authors analyze the impact of bank competition on the local economy. In Panel A, they conclude that when bank concentration increases, oscillation of age of properties also increases and proportion of “young” (newly built) properties decreases. That is, less competition leads to less development. Panel B reveals that when bank concentration rises, local income falls at the same time that the inflow of people with lower income in the following years, unemployment, income dispersion, and vacancy and rental rates rise.

Table IV (pp. 520-521) presents an analysis about the impact of bank competition (through mergers) on crime. It can be seen from Panel A that change in property crime is positively related to bank concentration. In panel B, the percentage change in BankHI1992:1995 is defined as:

being that the numerator is the estimated change in bank concentration measured through mergers (as shown before) and the denominator is the actual change. The conclusion is that the elasticity of property crimes with relation to bank concentration is low because the latter affects the former indirectly through economic variables. Naturally, the elasticity of property crime to, e.g, unemployment and wages is much higher.
Panel C indicates that fragile neighborhoods already experiencing concentrated loan markets and loan median income are more affected by an increase in bank concentration.

IV – EXOGENEITY OF COMPETITION MEASURE
Tables II, III, and IV indicate a link between bank mergers, loans conditions, real activity, and crime. The authors interprete this link as a causal one: mergers reduce financial access, which leads to neighborhood declines and subsequent increases in crime.
However, there is an alternative explanation for such correlations: the reverse causation, whereby unobservable variables that affect future crime drive current mergers. This idea is rejected by the authors for three reasons:
- Commercial real estate loans usually comprise a small portion of a balance sheet, so it is unlikely that changes in commercial estate market drive mergers;
- The analysis uses county fixed effects. It seems unlikely that large bank mergers are motivated by neighborhood-level variation of future crime or economic declines.
- The authors also employ another measure of bank competition (state branching deregulation) and find similar results. It is unlikely that both instruments (mergers and state branching deregulation) would be contaminated by the same endogeneity problems.

Table V (pp. 525-526) tests and rejects the reverse causation theory by showing that recents changes in crime, income and home value do not predict actual future bank concentration.

Table VI (p. 528) presents an additional test of the exogeneity of merger as measure of competition. The effect of the merger per se is separated from the effect of reduced competition caused by mergers. A dummy variable is set equal to one if a merger took place in the local banking market. As a result, the authors found that it is not mergers that affect loans and crime; such variables are affected by changes in bank concentration (Panel A). Moreover, mergers with no predicted competitive effects (BankHI = 0) have no effect on loan rate, burglary, and property.

V – COUNTY- AND STATE-LEVEL REGRESSIONS
For robustness, the authors search for evidence at county- and state-level. Table VII (p. 530) reveals that, at county-level, increase in bank concentration leads to decrease in the number of constructions in the following period and in the value and the number of new constructions in the following period (Panel A). Also, bank concentration affects only burglary (positive relation), which was the same result got for neighborhood-level. When abortion and imprisonment rate are included, bank concentration keeps its positive impact on burglary risk but has no effect on homicide (a crime against person), which is the same result got for neighborhoods.

Another robustness check is shown in Table VIII (pp. 533-534) where the impact of state-level bank branching regulation on some related variables is analyzed. The results are: unrestricted branching decrease subsequent actual bank concentration, loan rates, capitalization rates, and burglary risk decrease with bank competition, and loan size reduces with bank competition. These conclusions confirm prior results for neighborhood-level.

VI – CONCLUSIONS
The paper provides evidence that neighborhoods most affected by reductions in bank competition (due to bank mergers) subsequently experience worse credit terms (less credit and higher interest rates), less development and investment, lower real estate prices, an inflow of poorer households, and greater increases in future property crime.

Such results are confirmed by their findings that state branching deregulation (more competition) leads to more credit supply and fewer future property crimes.
The findings suggest that in evaluating the impact of bank mergers, regulators should consider not only the present bank concentration of the area but also its social fragility.

Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License