1 Consumption Risk Sharing over the Business Cycle: the Role of Small Firms Access to Credit Markets Mathias Hoffmann & Iryna Shcherbakova-Stewen University of Zurich PhD Topics Lecture, 22 Sep 2014 www.econ.uzh.ch/itf
Background The effects of financial deregulation on growth, business cycles and risk sharing Financial frictions vary over the business cycle (Financial accelerator models) The role of HH-level and firm heterogeneity for the cost of business cycles: Small firms may find it hard to borrow in recessions We exploit inter-state and time variation in the regulatory framework for banks to study the role of small firms in the US-wide pooling of state-specific risks.
Our story Interstate consumption risk sharing (CRS) in the U.S. increases in booms and decreases in recessions one percentage point increase in aggregate GDP growth increases CRS by 4 percentage points. In the trough of the average NBER recession, interstate consumption risk sharing was 17 percentage points below its long-run mean. CRS fluctuates more strongly over the aggregate cycle in states where small businesses are important. Intrastate bank branching deregulation (i.e. better local access to finance) has substantially weakened this dependence. So far unexplored macroeconomic benefit from banking deregulation: deregulation makes HH s and firms access to finance much less sensitive to the state of the business cycle.
A laboratory case of financial integration: U.S. state level banking deregulation Until the 1980s, the U.S. had a highly segmented, localized banking systems. (state right to levy licensing fees) The 1956 Douglas Amendment to the BHC act effectively gave states the right to bar entry by out-of-state banks. During the 1970s and 1980s many states gradually abolished intra- and inter-state branching and de-novo banking restrictions. The (federal) Riegle-Neal Act finally liberalized interstate branching and banking in the early 1990s This experiment has created wide cross-state variation in both intra- and interstate regulation: intra-state: banks allowed to branch or merge across county-borders inter-state: out-of-state banks allowed to branch or merge with local banks
Related Literature (I) The effects of state-level bank branching deregulation on...... state level output growth Jayaratne and Strahan (QJE 1996)... Business cycle comovement between U.S. federal states Morgan, Rime and Strahan (QJE 2004)...on the level of risk sharing Demyanik, Ostergaard, Sørensen, J. Finance (2007)... state-level sectoral portfolio composition Imbs et al. (2006)
Related Literature (II) Financial frictions vary over the business cycle...... models of the financial accelerator place collateral restrictions centre stage (Kyotaki and Moore (JPE,1997))... housing / mortgage collateral scarcity restrains risk sharing among U.S. regions (Lustig and van Nieuwerburgh, 2006)
Related Literature (III) Small firms and proprietary business activity... constitute source of non-insurable idiosyncratic risk for their owners (Heaton and Lucas (2000, J. Finance, EJ))... which is important for asset pricing (Heaton and Lucas (2000, J. Finance), Hoffmann (2006))...are particularly exposed to credit market restrictions (Gertler and Gilchrist (QJE, 1994))...affect the amount of interstate risk sharing (Agronin (2003))... seem to have played a key role in improving risk sharing after banking deregulation (Demyanik et al. (2007, J. Finance))
Data Sample period 1963-2005 construct or obtain gross state product, state personal income, disposable income at state level from regional economic accounts consumption at state level: retail sales in state, rescaled with share of aggregate retail sales in U.S. consumption. Business cycle indicators: annual GDP growth, NBER recession dates Small firm importance (µ): share of proprietors income in state personal income (µ 1 ), small business employment (µ 2 ) (only after 1977) Deregulation data: data on intra- and interstate deregulation from Kroszner and Strahan and Demyanik et al. (2007)
Fact Aggregate Risk Sharing increases in booms and decreases in recessions. [ ] ct k ct = τ t + β U (t) gspt k gspt + ut k 1 Figure 1: Degree of Uninsured Risk and Business Cycle 0.1 Degree of Uninsured Risk 0.5 0 0.05 0 Real GDP Growth 0.5 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 0.05
Fact This effect is stronger in federal states where small businesses are important 1.5 Figure 2A: High Share of Proprietary Income 0.08 2 Figure 2B: Low Share of Proprietary Income 0.1 1 0.06 Degree of Uninsured Risk 0.5 0 0.04 0.02 Real GDP Growth Degree of Uninsured Risk 0 0 Real GDP Growth 0.5 0 1 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 0.02 2 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 0.1 FIGURE 3: Risk sharing over the business cycle for states with above median (left panel (A)) and below median (right panel (B)) small business importance. In each panel, the blue, solid line is the coefficient β U (t) of the sequence of cross-ssectional regressions c t k = β U (t) gsp k t + τ t + ε k t. The red, dashed line is US GDP growth. Vertical lines indicate NBER business cycle troughs.
Fact This cyclical pattern in risk sharing vanishes after a state has deregulated: 0.4 Degree of Uninsured Risk around NBER Trough 0.35 Degree of Uninsured Risk 0.3 0.25 0.2 0.15 1 0 1 Years to NBER Trough FIGURE 2: Burns-Mitchell diagram of the fraction of unshared risk β k U (t) around NBER recession troughs distinguishing between states that have not yet (blue, solid line) and those that have already deregulated(red, dashed line).
General framework Whole point of our exercise is that these β U varies over time and region. Our baseline specification is so that c k t = a U gsp k t + b U gdp t gsp k t + d k Ut 1 + εk Ut (1) β U (t) = a U + b U gdp t can be interpreted as the fraction of unshared risk that varies with GDP growth. We expect: b U < 0 before banking deregulation and zero thereafter b U should be more negative for states with lots of small businesses
Two ways of pursuing these hypotheses: 1 split sample according to whether a state has lots of small businesses, has deregulated early or late etc... 2 More formally, let β U to depend on a vector of time and state-varying characteristics z k t : [ ] β k U (t) = gdp t z k t b U + z k t a U Then our typical regression has the form: c k t = gdp t [ z k t b U ] gsp k t + z k t a U gsp k t + y k t c U + d k Ut 1 + εk U,t
Examples To check whether risk sharing is more cyclical in states with lots of small businesses: [ ] β k U (t) = a U0 + b U0 gdp t + b U1 µ k µ gdp t + a 0 µ k where µ k is a measure of the importance of small businesses (proprietary income, small business emplyoment etc.) To examine what banking deregulation has done to the cyclicality in risk sharing: c k t = [ a U0 + b U0 gdp t + b U1 gdp t SD k t + a U1 SD k t ] gsp k t + c U SD k t + d k t 1 + ε k U where SDt k is a dummy that becomes one from the date when state k deregulated its banking sector.
Taking stock Risk sharing increases in booms and decreases in recessions but only so in the first half of the sample This pattern is stronger in states with lots of small businesses and robust to a range of controls (industrial structure, endogeneity) But: would also like to know HOW (i.e. through which channels) risk is shared
Channels of risk sharing A fraction 1 β U of the variance of a typical state-specific shocks gets laid off through various channels. The question is: which channels? Here follow Asdrubali, Sorensen and Yosha (1996) and consider 3 channels: income smoothing through factor income from out-of-state fiscal transfers consumption smoothing by asset cumulation or decumulation
How to measure the contribution of these channels? [ gsp k t = gsp k t si k ] t + [ si k t dsi k t ] + Take covariance with gsp k t on both sides and rearrange to get where β I + β F + β C = 1 β U [ dsi k ] t c t k + c t k β I = cov( gsp k t si k t, gsp k t )/var( gsp k t ) income smoothing β F = cov( si k t dsi k t, gsp k t )/var( gsp k t ) fiscal smoothing β C = cov( dsi k t c k t, gsp k t )/var( gsp k t ) consumption smoothing β U = cov( c k t, gsp k t )/var( gsp k t ) unsmoothed component (we had that already!)
These β s can be obtained as regression coefficients gsp k t si k t = α I + β I gsp k t + δ k I + ε k I,t (2) si k t dsi k t = α F + β F gsp k t + δ k F + εk F,t dsi k t c k t c k t = α C + β C gsp k t + δ k C + εk C,t = α U + β U gsp k t + δ k U + εk U,t and they can be made to vary across state and time in the same way as β U : [ ] β k X (t) = gdp t z k t b X + z k t a X where X = I, F, C, U in turn.
Channels of risk sharing and the business cycle Table 1: Risk Sharing and the Business Cycle Baseline specification: channels over the business cycle (I ) (F ) (C ) (U) Panel A: β X (t) = a X 0 + b X 0 gdp t 1964-1984 a X 0 0.54** 0.02 0.19** 0.25** (9.38) (0.39) (2.11) (3.35) b X 0-2.91** 0.49 6.13** -3.70** (-2.66) (1.15) (2.33) (-2.12) 1985-2005 a X 0 0.67** 0.21** -0.07 0.19** (5.23) (2.12) (-0.64) (2.07) b X 0 0.57-3.42 3.83-0.98 (0.25) (-1.11) (1.07) (-0.49)
Role of small businesses Table 2: Risk Sharing and Small Business Importance Panel A reports the results of the panel OLS regression c k,t = β U (t) gsp k t + d k Ut 1 + εk Ut for two periods: pre-1984 and post-1984. β U(t) is defined as β U (t) = a U + b U gdp t. The states are split into groups according to the importance of small businesses ( low, middle, high ) µ k. Panel A µ k = shapi (share of proprietary income 1964-1975) 1964-1984 1985-2005 low middle high low middle high a U 0.16* 0.36** 0.29** 0.09 0.17 0.16 (1.92) (3.55) (3.97) (0.57) (1.53) (1.00) b U -0.70-1.52-6.39** 2.37 0.87-3.41 (-0.43) (-0.67) (-2.56) (0.82) (0.25) (-0.84) µ k = SBE (Small Business Employment in 1977) a U 0.49** 0.20** 0.15* 0.36 0.31 0.08 (5.47) (2.08) (1.95) (1.27) (1.64) (1.60) b U -4.65 1.05-3.40-2.00-3.60-0.29 (-1.44) (0.62) (-1.62) (-0.23) (-0.69) (-0.14)
Banking deregulation & small businesses Table 5: Risk Sharing, Banking Deregulation and Small Businesses The table reports the results of the panel OLS regression for the period 1964-1984 c k,t β U (t) = a U + b U gdp t. = β U (t) gsp k t + d k Ut 1 + εk Ut, where µ k = shapi (share of proprietary income 1964-1975) early deregulation late deregulation low µ k high µ k low µ k high µ k a U 0.14* 0.37** 0.42** 0.26** (1.82) (2.44) (4.58) (3.63) b U -0.82-6.69-2.42-6.38** (-0.53) (-1.55) (-1.11) (-2.91) µ k = SBE (Small Business Employment in 1977) early deregulation late deregulation low µ k high µ k low µ k high µ k a U 0.47** 0.12* 0.30** 0.23** (4.47) (1.85) (3.04) (3.19) b U -2.15-1.97-3.03-5.41** (-0.46) (-0.89) (-1.58) (-2.30)
Impact of BD during booms and recessions Table 4: Risk Sharing, Banking Deregulation and the Business Cycle The table reports the results of the panel OLS regressions x t = β k X (t) gspk t + c X SDt k + d k Xt 1 + ε k Xt with xt = gspk t si k t, si k t dsi k t, dsi k t c t k, ck t for X = I, F, C, U respectively. β X (t) is defined as indicated in the panel heading. P tand T tare NBER peak and trough indicators. (I ) (F ) (C ) (U) β k X (t) = b X 0P t + b X 1 T t + b X 2 P t SD k t + b X 3 T t SD k t + a X 0 + a X 1 SD k t b X 0-0.15-0.00 0.29** -0.13 (-1.47) (-0.16) (2.03) (-1.59) b X 1 0.24** -0.00-0.53** 0.30** (2.76) (-0.23) (-3.65) (3.46) b X 2 0.16-0.21-0.04 0.09 (0.82) (-1.29) (-0.21) (0.53) b X 3-0.29** 0.01 0.53** -0.25** (-2.74) (0.29) (3.02) (-2.40) a X 0 0.41** 0.10** 0.35** 0.14** (7.95) (9.07) (3.84) (2.93) a X 1 0.24** -0.03-0.24** 0.03 (3.26) (-1.10) (-3.10) (0.54)
Conclusions Interstate risk sharing in the U.S. increases in booms, decreases in recessions. Small firm s access to finance (credit markets) seems key in this: cyclical dependence more pronounced where SMB are important. State-level banking deregulation has made SMB access to finance less dependent on the state of the business cycle. A novel and unexplored benefit from banking deregulation. Findings could inform the literature on the welfare costs of business cycles and monetary policy. Obvious extensions to financial liberalization and development in the international context.