Barr invshock v2_slides

233 views
155 views

Published on

Published in: Technology, Education
0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total views
233
On SlideShare
0
From Embeds
0
Number of Embeds
2
Actions
Shares
0
Downloads
1
Comments
0
Likes
0
Embeds 0
No embeds

No notes for slide

Barr invshock v2_slides

  1. 1. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . . Permanent and Transitory Shocks in Capital Structure and Their Relation to Investment, Leverage, and Speed of Adjustment . Stephen J. Barr stephen.barr@simon.rochester.edu University of Rochester May 15, 2012 . . . . . .
  2. 2. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Introduction - Welcome to my presentation Motivation: Better understand the profitability shock process and its industry-level variation, and its effect on capital structure Economic Findings: Permanent shocks, although relatively small in magnitude, have a large impact on leverage and investment decisions . . . . . .
  3. 3. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Firms: . . . . . .
  4. 4. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Firms: Q: What types of uncertainty to profitability do firms face, and how does it affect their capital structure? . . . . . .
  5. 5. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Firms: Q: What types of uncertainty to profitability do firms face, and how does it affect their capital structure? Q: Is all uncertainty faced by a firm simply be summarized by volatility and autocorrelation of profitability, or is there more to it? . . . . . .
  6. 6. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Firms: Q: What types of uncertainty to profitability do firms face, and how does it affect their capital structure? Q: Is all uncertainty faced by a firm simply be summarized by volatility and autocorrelation of profitability, or is there more to it? Literature: . . . . . .
  7. 7. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Firms: Q: What types of uncertainty to profitability do firms face, and how does it affect their capital structure? Q: Is all uncertainty faced by a firm simply be summarized by volatility and autocorrelation of profitability, or is there more to it? Literature: Uncertainty in profitability usually modeled as strictly transient process . . . . . .
  8. 8. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Intuition - There are many different types of shocks to profitability - legislation, technology, labor disputes, etc. Expectations differ. . . . . . .
  9. 9. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Intuition - There are many different types of shocks to profitability - legislation, technology, labor disputes, etc. Expectations differ. Examples (Transitory): E. coli. scares in spinach, peanut butter (Gorbenko and Strebulaev 2010). Mad cow in beef. Recall of a competitors product Labor issues (union strikes every few years) . . . . . .
  10. 10. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation - Transient Shock - Mad Cow . . . . . .
  11. 11. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Intuition - There are many different types of shocks to profitability - legislation, technology, labor disputes, etc. Expectations differ. Examples (Transitory): E. coli. scares in spinach, peanut butter (Gorbenko and Strebulaev 2010). Mad cow in beef. Recall of a competitors product Labor issues (union strikes every few years) . . . . . .
  12. 12. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation Intuition - There are many different types of shocks to profitability - legislation, technology, labor disputes, etc. Expectations differ. Examples (Transitory): E. coli. scares in spinach, Examples (Permanent): peanut butter (Gorbenko Changes in legislation and Strebulaev 2010). Technological innovations Mad cow in beef. Patents expiring Recall of a competitors Trade agreements product Labor issues (union strikes every few years) . . . . . .
  13. 13. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . Motivation - Permanent Shock - Legislation (Tobacco Tax) “CHIPRA (Children’s Health Insurance Program Reauthorization Act) substantially raised rates on cigarettes, roll-your-own tobacco, and small cigars, but did not raise taxes on pipe tobacco to equivalent rates.”1. . . . . . .
  14. 14. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Motivation - Summary Why do we care about permanent shocks? . In reality, not all shocks are permanent. 1 . In reality, not all shocks are temporary. 2 . . . . . .
  15. 15. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . 1. Introduction 2. Related Literature . 3 Model 4. Comparative Statics and Identification Leverage Investment Identification 5. Data - Full Sample 6. Estimation Results Where I can improve 7. Conclusion 8. Data - Industry Subsets . 9 Estimation Per Industry . 10 Appendices Detrending The Model Proof: Model is a Contraction Mapping . . . . . .
  16. 16. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Related Literature Dynamic Trade-off Models - DeAngelo, DeAngelo, and Whited (2011), Hennesey and Whited (2005, 2007) Permanent and Transitory Shocks in Investment Gourio (2008) Permanent and Transitory Shocks in Capital Structure - Gorbenko and Strebulaev 2010 Permanent and Transitory - Macro - Hall and Mishkin (1982), Flavin 1984, Blundell and Preston (1998) . . . . . .
  17. 17. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Economics of Productivity Shock . . . . . .
  18. 18. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Results Preview . . . . . .
  19. 19. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Results Preview . Transient only (zT ∗ ) 1 with ρ∗ 0.7 or . . . . . .
  20. 20. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Results Preview . Transient only (zT ∗ ) 1 with ρ∗ 0.7 or . Transient plus permanent (zT + zP ) 2 . . . . . .
  21. 21. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Results Preview . Transient only (zT ∗ ) 1 with ρ∗ 0.7 or . Transient plus permanent (zT + zP ) 2 . ρ 1 0.3 to 0.5 and σ P 0.03 is also plausible . . . . . .
  22. 22. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Intuition Firms will react, in general, more quickly to a permanent shock WHY?: There is no concept of “waiting out” a permanent shock - the marginal cost of adjusting is quickly swamped by the marginal benefit to adjusting, as the change is expected to be permanent IMPLICATION: Change induced by a transient shock can be induced by a much smaller permanent shock. This framework can match many interesting moments with a much lower autocorrelation than in the transient-only case . . . . . .
  23. 23. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Model Modeling Objective: Create a dynamic capital structure model Incorporate permanent and transitory shocks Firm controls debt and capital Firm’s Purpose: Maximize Present Value of Firm . . . . . .
  24. 24. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = . . . . . .
  25. 25. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity . . . . . .
  26. 26. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs . . . . . .
  27. 27. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) . . . . . .
  28. 28. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ( ) 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. . . . . . .
  29. 29. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , zP ) t t kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , zt ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ( ) where β = 1 1+r , dΓ(zt , zt+1 ) where zt ≡ zT + zP t t . . . . . .
  30. 30. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Payout / Equity Issuance FN: e(Kt , Kt+1 , Bt , Bt+1 , zT , t P) t Economics: e(·) < 0 ⇒ Firm needs financing ⇒ Firm issues equity e(·) > 0 ⇒ Firm has excess cash ⇒ Firm makes distribution to shareholders e(·) = (1 − τc )πt (zt , Kt ) Production / Profits − δKt τc Depreciation Tax Shield + It (Kt , Kt+1 ) Investment − A(Kt , Kt+1 ) Capital Adjustment Costs + Dt Net Debt . . . . . .
  31. 31. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Profits Profits occur according to a Cobb-Douglas production function π = z ∗ Kθ z profitability shock, K capital stock, θ production curvature e(·) = (1 − τc )πt (zt , Kt ) Production / Profits − δKt τc Depreciation Tax Shield + It (Kt , Kt+1 ) Investment − A(Kt , Kt+1 ) Capital Adjustment Costs + Dt Net Debt . . . . . .
  32. 32. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Profitability Shock Process Transitory Shock log zT = ρ ∗ log zT + t+1 t T t , T t ∼ N(0, σT ) , ρ ∈ (0, 1) . . . . . .
  33. 33. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Profitability Shock Process Transitory Shock log zT = ρ ∗ log zT + t+1 t T t , T t ∼ N(0, σT ) , ρ ∈ (0, 1) Permanent Shock log zP = log zP + t+1 t P t , P t ∼ N(0, σP ) . . . . . .
  34. 34. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Profitability Shock Process Transitory Shock log zT = ρ ∗ log zT + t+1 t T t , T t ∼ N(0, σT ) , ρ ∈ (0, 1) Permanent Shock log zP = log zP + t+1 t P t , P t ∼ N(0, σP ) Total Shock Process log zt = log zP + log zT t t . . . . . .
  35. 35. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Taxes τc corporate tax rate, 35% Profits taxed at this rate Capital Depreciation is not taxed (depreciation tax shield) Corporate debt also serves as a tax shield e(·) = (1 − τc )πt (zt , Kt ) Production / Profits − δKt τc Depreciation Tax Shield + It (Kt , Kt+1 ) Investment − A(Kt , Kt+1 ) Capital Adjustment Costs + Dt Net Debt . . . . . .
  36. 36. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Investment Investment - Law of Motion It+1 (Kt , Kt+1 ) ≡ Kt+1 − (1 − δ) ∗ Kt . e(·) = (1 − τc )πt (zt , Kt ) Production / Profits − δKt τc Depreciation Tax Shield + It (Kt , Kt+1 ) Investment − A(Kt , Kt+1 ) Capital Adjustment Costs + Dt Net Debt . . . . . .
  37. 37. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Capital Adjustment Costs Convex Adjustment Cost Function ( )2 a Kt+1 − (1 − δ)Kt A(Kt , Kt+1 ) = γKt Φi + Kt 2 Kt where γ, a are constants. Φi indicates investment. Cooper and Haltiwanger 2006, DDW 2011, HW 2005, 2007 e(·) = (1 − τc )πt (zt , Kt ) Production / Profits − δKt τc Depreciation Tax Shield + It (Kt , Kt+1 ) Investment − A(Kt , Kt+1 ) Capital Adjustment Costs + Dt Net Debt . . . . . .
  38. 38. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Debt Modeling debt as 1-period debt Negative Debt ≡ Cash ¯ Bt ∈ [B, B] ⊂ R Dt+1 ≡ Bt+1 − (1 + r(1 − τc ))Bt e(·) = (1 − τc )πt (zt , Kt ) Production / Profits − δKt τc Depreciation Tax Shield + It (Kt , Kt+1 ) Investment − A(Kt , Kt+1 ) Capital Adjustment Costs + Dt Net Debt . . . . . .
  39. 39. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ) ( 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. . . . . . .
  40. 40. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ) ( 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. . . . . . .
  41. 41. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Equity Issuance II, Issuance Costs Firms pay a cost to issue equity U-shaped cost curve Altınkılıç and Hansen (2000) More capital beyond some amount entails rising costs of underwriter certification, monitoring, and marketing, which increase the spread. ( ) 1 φ(e(Kt , Bt , Kt+1 , Bt+1 , z∗ )) = Φe ∗ λ1 e(·) − λ2 e(·)2 t 2 where λi ≥ 0, i = 1, 2, and Φe indicates equity issuance (e(· < 0)) . . . . . .
  42. 42. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ) ( 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. . . . . . .
  43. 43. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ) ( 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. . . . . . .
  44. 44. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ) ( 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. . . . . . .
  45. 45. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Maximizing the Present Value of Equity V(kt ,bt , zt ) = [ max e(kt , kt+1 , bt , bt+1 , zT , t P t) kt+1 ,bt+1 ∈K×B Payout/Equity ] ( ) T P + φ e(kt , kt+1 , bt , bt+1 , zt , t ) Equity Issuance Costs ∫ +β ∗ V(kt+1 , bt+1 , zt+1 ) dΓ(zt , zt+1 ) Continuation Value (Dynamic Model) ) ( 1 where β = 1+r , dΓ(zt , zt+1 ) where zt is the sum of a permanent shock and a transitory shock. Use Value Iteration to solve. . . . . . .
  46. 46. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Note: Detrending As written, the model is non-stationary and thus cannot be solved using standard techniques To solve, detrend the model. .. See Details Detrending refers to the idea that, with permanent shocks, firm size can grow without bound Main Idea: f (..., zT , t zP t ) Non-stationary Detail: Proof of sufficient conditions to solve .. See Proof . . . . . .
  47. 47. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Note: Detrending As written, the model is non-stationary and thus cannot be solved using standard techniques To solve, detrend the model. .. See Details Detrending refers to the idea that, with permanent shocks, firm size can grow without bound Main Idea: f (..., zT , t zP t ) ⇒ ˆ (..., zT , f t P t ) Non-stationary Stationary Detail: Proof of sufficient conditions to solve .. See Proof . . . . . .
  48. 48. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . Comparative Statics . . . . . .
  49. 49. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Leverage P σt ↑⇒ Mean Leverage ↓ Slope is as expected. Economics: More volatility ⇒ less leverage . . . . . .
  50. 50. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Leverage Economics: Small magnitude shocks have large effects P σt ↑⇒ Mean Leverage ↓ Slope is as expected. Economics: More volatility ⇒ less leverage . . . . . .
  51. 51. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Leverage - Quantiles Many go to zero For a few at the highest levels, they increase their leverage when σ P ↑ . . . . . .
  52. 52. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .For Comparison: Transient Shock and Mean Leverage In the transient only case, mean leverage acts as expected, but σ T has to increase substantially to show this drop in mean leverage When permanent shocks are added, the affect of changes in σ T is muted . . . . . .
  53. 53. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Investment Why does investment increase? . . . . . .
  54. 54. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Investment P σt ↑⇒ Mean Investment ↑, ceteris paribus . . . . . .
  55. 55. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Investment Economics: What is driving this behavior? P σt ↑⇒ Mean Investment ↑, ceteris paribus . . . . . .
  56. 56. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock and Mean Investment - Quantiles The 75% percentile is downward sloping “Mean Investment” is being pulled up by the 90% percentile Economics: Intuition is right for MOST firms (σ P ↑⇒ higher volatility ⇒ less investment). For a small number of firms, it is better to increase investment even in the face of additional uncertainty . . . . . .
  57. 57. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .For Comparison: Transient Shock and Mean Investment T σt ↑⇒ Mean Investment ↓, but only slightly This is consistent with Gourio 2008 - “investment reacts more the a permanent shock” . . . . . .
  58. 58. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .For Comparison: Transient Shock and Mean Investment Economics: Firm invests more when it expects productivity to be higher in the future T σt ↑⇒ Mean Investment ↓, but only slightly This is consistent with Gourio 2008 - “investment reacts more the a permanent shock” . . . . . .
  59. 59. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Comparative Static Analysis Summary . Permanent shocks, for their magnitude, have a large impact 1 on moments of interest (investment and leverage) relative to transient shocks of similar magnitude . When dealing with simulated firms that can experience 2 permanent shocks, the behavior of the extremes can affect the sign if the partial derivative of that moment . . . . . .
  60. 60. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock Identification Strategy Using established moments PLUS two additional moments . . . . . .
  61. 61. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock Identification Strategy Using established moments PLUS two additional moments . Speed of Adjustment to Target Leverage - Regression-based 1 SOA Justification: Firms react much more quickly to permanent shocks . . . . . .
  62. 62. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Permanent Shock Identification Strategy Using established moments PLUS two additional moments . Speed of Adjustment to Target Leverage - Regression-based 1 SOA Justification: Firms react much more quickly to permanent shocks . Covariance of Long-Run Growth of Firm size and Lagged 2 Profitability Justification: In the long run, what differentiates a permanent shock from a temporary shock is the size of the firm ( ) Kit πi,t−3 Cov log , Ki,t−3 Ki,t−3 . . . . . .
  63. 63. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Speed of Adjustment to Target Leverage -Regression-based SOA Interpretation: Increasing permanent shock variance (σ P ) does, to a point, increase SOA However, this affect eventually gets swamped by other factors . . . . . .
  64. 64. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Covariance of Investment and Profitability Moment: ( ) Kit πi,t−3 Cov log , Ki,t−3 Ki,t−3 Interpretation: As σ P increases, the shock becomes more permanent. Permanent shocks lead to long-lasting changes in firm size. . . . . . .
  65. 65. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Identification - Moments 1 and 2 Note that for any given σ P , there is a unique pair: ( ( )) ⇒ σ P is identified π SOAβ, Cov log KKit , Ki,t−3 i,t−3 i,t−3 . . . . . . .
  66. 66. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Full Sample - Summary Statistics Table: Summary Statistics for Full Sample Panel A: Moment Means and Variances Statistic Mean Var 3rd Central Leverage 0.2608 0.0349 Investment 0.1161 0.0053 0.0130 Equity Issuance 0.0173 0.0019 Tobin’s Q 2.5702 Operating Income 0.1628 0.0068 Ser. Cor. of OpInc 0.7877 Var ( Innov to OpInc of ) 0.0028 Kit π Cov log , i,t−3 Ki,t−3 Ki,t−3 0.0063 SOA βLEVt−1 0.8814 Num. firm-year Obs 4769 . . . . . .
  67. 67. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Full Sample - Summary Statistics Table: Summary Statistics for Full Sample Panel A: Moment Means and Variances Statistic Mean Var 3rd Central Leverage 0.2608 0.0349 Investment 0.1161 0.0053 0.0130 Equity Issuance 0.0173 0.0019 Tobin’s Q 2.5702 Operating Income 0.1628 0.0068 Ser. Cor. of OpInc 0.7877 Var ( Innov to OpInc of ) 0.0028 Kit π Cov log , i,t−3 Ki,t−3 Ki,t−3 0.0063 SOA βLEVt−1 0.8814 Num. firm-year Obs 4769 . . . . . .
  68. 68. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . Table: Compustat Variables and Moments Moment Compustat Variables ( Investment CAPX (Capital Expenditures) − ) SPPE (Sale of Property) / PPEGT (Property, Plant, and Equipment - Total (Gross) ( Leverage DLTT (Long-Term Debt - Total) + ) DLC (Debt in Current Liabilities - Total) / AT (Assets - Total) Operating Income OIBDP (Operating Income Before Depreciation) / AT (Assets - Total) Equity Issuance SSTK (Sale of Common and Preferred Stock) / AT (Assets - Total) Cash Flow CHE (Cash and Short-Term Investments) / AT (Assets - Total) ( Market Value CSHO (Common Shares of Stock Outstanding) ∗ ) PRCCF Price Close - Annual (Fiscal Year) ( Market-to-Book DLTT (Long-Term Debt - Total) + DLC (Debt in Current Liabilities - Total) + PRSTKC (Purchase of Command and Preferred Stock) + ) Market Value / AT (Assets - Total) Book Equity SEQ (Stockholders’ Equity - Total) + TXDITC (Deferred Taxes and Investment Tax Credit) − PSTK (Preferred stock) Book Debt AT (Assets - Total) − ( Book Equity Tobin’s Q Market Value + Book Debt + ) ACT (Current Assets - Total) / PPEGT (Property, Plant, and Equipment - Total (Gross) Definitions taken from the documentation for the Compustat Annual Data - Industrial documentation. . . . . . .
  69. 69. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Full Sample - Estimation Table: Full Sample Estimations for Barr and DDW Models DDW S.E BARR S.E. Parameter Estimate Estimate Autocorrelation ρ 0.585518 0.297351 Std Dev σT 0.298915 0.206295 Agency s 0.078320 0.181950 Fixed γ 0.002986 0.002928 Convex a 0.153295 0.859097 Equity - Fixed λ1 0.099852 0.102322 Equity - Convex λ2 0.003255 0.003502 Curvature θ 0.837924 0.790740 Permanent Shock σP 0.026600 . . . . . .
  70. 70. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Full Sample Estimation - Discussion The estimation seems to indicate that a combination of permanent and transitory shock can also match the data The introduction of even a relatively small permanent shock (σ P = 0.0266) takes away a large portion of the autocorrelation of the transient shock ρT = 0.5855 ↓ ρT+P = 0.2974 . . . . . .
  71. 71. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Another Estimation Table: Full Sample Estimations With and Without Permanent Shocks Trans. Only Both Shocks Parameter Estimate Estimate Autocorrelation ρ 0.585518 0.517682 Std Dev σT 0.298915 0.0794019 Agency s 0.078320 1.204867 Fixed γ 0.002986 0.012505 Convex a 0.153295 0.226857 Equity - Fixed λ1 0.099852 0.024780 Equity - Convex λ2 0.003255 0.006924 Curvature θ 0.837924 0.815899 Permanent Shock σP 0.030197 . . . . . .
  72. 72. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Moment Match Moment Diff ≡ Simulated − Data Mean Investment -0.1357 0.2378 0.101983 Investment Variance -0.1867 0.192035 0.00525221 3rd Central Moment of Inv. -0.0740 0.0744569 0.000374579 Mean Leverage 0.1792 0.100575 0.279821 Leverage Variance 0.0246 0.0175781 0.0421828 Mean Opr. Income -0.0805 0.231223 0.150638 Ser. Cor. of Op. Inc. 0.0275 0.760189 0.78767 Var. Opr. Inc. Innov 0.0020 0.000780283 0.00276893 Mean Tobin’s Q -3.7123 5.82386 2.11158 Mean Equity Issuance -0.3352 0.349156 0.0139093 Var Equity Issuance -0.1111 0.112791 0.00180158 Mean Cash Balances 0.0718 4.06533e-17 0.0718038 Cov(Asset Growth, Profit) 4.39693e-05 0.00622209 0.00626606 Table: An SMM estimation - Moments Fit . . . . . .
  73. 73. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . One possible reason the estimations are struggling is that I am calculating my moments without trimming the extremes. Recalling the plot where mean investment was increasing in σ P , and this was being driven by a very small amount of firms. However, from the perspective of the optimizer, ∂[Mean Investment] >0 ∂σ P This is going to lead to perverse matches. It is possible that there may be a similar but reversed affect for some other parameter that affects investment. This, the matching is being pulled in opposite directions. . . . . . .
  74. 74. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Conclusion Permanent shocks matter, disproportionate to their size It is possible to match many interesting moments with a different shock process This shock process seems economically plausible - in reality, both and permanent and transitory shocks exist More work needed to properly estimate . . . . . .
  75. 75. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Data - Industry Subsets - Summary Statistics Moment Full Manuf. Mining Retail Svcs. Transport Mean Inv. 0.1020 0.0948 0.1195 0.1071 0.1138 0.1037 Var Inv. 0.0053 0.0044 0.0078 0.0054 0.0061 0.0063 Inv. 3rd 0.0004 0.0003 0.0007 0.0002 0.0005 0.0004 Mean Lev. 0.2798 0.2529 0.2941 0.2845 0.3370 0.3413 Var Lev. 0.0422 0.0334 0.0372 0.0522 0.0649 0.0516 Mean OpInc. 0.1506 0.1528 0.1403 0.1498 0.1677 0.1521 Ser Cor OpInc. 0.7877 0.8269 0.4935 0.8757 0.8809 0.8283 Var OpInc Innov 0.0028 0.0026 0.0079 0.0015 0.0017 0.0015 Mean Tobin’s Q 2.1116 2.4069 0.9768 1.5688 3.6498 1.5691 Mean Eq. Iss. 0.0139 0.0115 0.0311 0.0090 0.0217 0.0135 Var Eq. Iss. 0.0018 0.0012 0.0044 0.0008 0.0038 0.0024 Mean Cash Bal. 0.0718 0.0763 0.0519 0.0656 0.0952 0.0689 SOA Beta Lev.t−1 0.8815 0.8733 0.7314 0.9226 0.8708 0.9158 Cov(Size,Profit) 0.0063 0.0072 0.0139 0.0078 0.0044 0.0010 N Firm-Year 4769 2362 462 793 385 552 . . . . . .
  76. 76. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ . Table: SMM Estimation By Industry MODEL Industry ρ σT σP BARR ALL 0.297351 0.20629564 0.026608675 DDW ALL 0.585518 0.29891501 BARR Manuf 0.594036 0.088540518 1.7910021e-3 DDW Manuf 0.445501 0.094870450 BARR Mining 0.607389 0.039622347 1.2838319e-3 DDW Mining 0.394985 0.090336292 BARR Transport 0.613267 0.093294045 3.3792821e-3 DDW Transport BARR Retail 0.319490 0.15415480 1.1743024e-3 DDW Retail 0.410377 0.091564558 BARR Services 0.975731 0.027947906 1.1740852e-3 DDW Services 0.926724 0.039731578 . . . . . .
  77. 77. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrending the Model The permanent shock introduces a challenge: non-stationarity. Economic issues: Add content here Solving issues: the state space of permanent shocks is infinite . . . . . .
  78. 78. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrending - Intuition The permanent shock process can be detrended. . . . . . .
  79. 79. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrending - Intuition Normal Model: ( ) f levels: zP , Kt , Bt , ... t Detrended Model: ( ) ˆ f innovations: P, kt , bt , ... t Notation: UPPERCASE implies normal model (Kt , Bt ...) lowercase implies detrended model (kt , bt ...) . . . . . .
  80. 80. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrending - Primitives Permanent shock log zP = log zP + t t−1 P t + zP + µ , 0 P t ∼ N(0, σP ) zP t P = exp( t + zP + µ) 0 zP t−1 Capital Stock (1/(1−α)) Kt = kt ∗ zP t−1 Debt dt = bt+1 ∗ exp( P ) − (1 + r (1 − τc ))bt . t . . . . . .
  81. 81. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrended Equity Issuance e(kt , kt+1 , bt ,bt+1 , zT t , P t)= zT exp(µ t + P + zP )kα t 0 t + δ ∗ k t τc Profit Depreciation Tax Shield   kt+1 − ( ) − (1 − δ)kt  exp 1 P + zP ) α−1 (µ + t 0 Investment ( ( )2 ) a kt+1 − γkΦi + − (1 − δ) exp( p )1/(1−α) 2 kt Adjustment Costs P + bt+1 ∗ exp( t) − (1 + r(1 − τc ))bt Debt . . . . . .
  82. 82. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrended Equity Issuance e(kt , kt+1 , bt ,bt+1 , zT t , P t) = P ( ( zT t exp(µ ((( P (((+ t + z0 ) kα t + δ ∗ k t τc Profit Depreciation Tax Shield   kt+1 − ( ( ( − (1 − δ)kt  () P( exp (((((( + zP ) 1 (( α−1 (µ + t 0 Investment ( ( ) ) a kt+1 2 − γkΦi + − (1 − δ) exp( p )1/(1−α) 2 kt Adjustment Costs + bt+1 ∗ ) − (1 + r(1 − τc ))bt exp( P t Debt . . . . . .
  83. 83. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Detrended Equity Issuance e(kt , kt+1 , bt ,bt+1 , zT t , P t) = zT t exp(µ + P t + zP ) kα 0 t + δ ∗ k t τc Profit Depreciation Tax Shield   kt+1 − ( ) − (1 − δ)kt  exp 1 P + zP ) α−1 (µ + t 0 Investment ( ( )) ) a kt+1 2 − γkΦi + − (1 − δ)exp( p )1/(1−α) 2 kt Adjustment Costs P + bt+1 ∗ exp( t )−(1 + r(1 − τc ))bt Debt . . . . . .
  84. 84. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Proof: Model is a Contraction Mapping To prove that the model is a contraction mapping, it is sufficient to satisfy Blackwell’s sufficient conditions. . 1 Monotonicity . Discounting 2 . . . . . .
  85. 85. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Condition 1: Monotonicity Write model as T(V) = max e(a) + φ(e(a)) + βE[V(a)] a∈A Then, take V1 V2 under the sup norm. T(V1 ) = max [e(a) + φ(e(a)) + βE[V1 (a)]] a∈A ≤ max [e(a) + φ(e(a)) + βE[V1 (a)]] a∈A . . . . . .
  86. 86. Introduction Related Literature Model Comparative Statics and Identification Data - Full Sample Estimation Results Conclusion ............ .Condition 2: Discounting Let β β 1. [ ] T(V + m) = max e(a) + φ(e(a)) + βE[V(a) + m] a∈A [ ] = max e(a) + φ(e(a)) + βE[V(a)] + βm a∈A [ ] ≤ max e(a) + φ(e(a))βE[V(a)] + β m a∈A .. Back to Main . . . . . .

×