## 1. Introduction (wp17234)

## Source details

**Canonical URL:** [1. Introduction (wp17234)](https://www.imf.org/-/media/files/publications/wp/2017/wp17234.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17234.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17234.pdf.json)

---

### Overview and motivation
- Over the past few decades, the importance of intangible capital in corporate balance sheets has steadily increased.
- Intangible assets are more firm-specific and more difficult to verify, price, and liquidate than tangible assets, and therefore offer less collateral value.
- The rise in intangible capital may increase intermediation frictions: borrowers with large shares of intangible assets may find it difficult to obtain funding through secured lending.
- Existing literature links the rise in intangible capital to lower corporate debt capacity and greater reliance on internal financing; however, how this friction affects the banking sector is unexplored in prior work.

### Main hypotheses
- Empirical hypothesis 1: Higher corporate intangible assets reduce the equilibrium volume of bank C&I lending, all else equal.
- Empirical hypothesis 2: Higher corporate intangible assets increase the equilibrium volume of bank non-C&I assets, reduce the equilibrium size of bank balance sheets, or lead to a combination of the two.

### Empirical approach and scope
- Match firms and banks based on headquarters’ Metropolitan Statistical Area (MSA) to exploit geographical variation in corporate intangible capital and bank portfolio allocations.
- Use a large bank-level panel over 1977-2010 to examine how banks’ commercial loans and total assets respond to growth in intangible capital for firms located in the same MSA.
- Focus on bank non-C&I asset allocations, especially growth rates of real estate loans and liquid assets.

### High-level main findings
- The rise in corporate intangible capital reduces commercial loan growth, controlling for bank fixed effects, time-varying bank characteristics, and MSA-level loan demand variables.
- A one standard deviation increase in intangible capital is accompanied by a contraction in bank C&I lending by 2 to 3.5 percentage points in the main specification.
- Up to 20% of the secular decline in the share of commercial loans in total bank lending since the mid-1970s can be attributed to the increase in corporate intangible capital.
- Banks exposed to higher growth of MSA-level corporate intangible capital do not reduce their size; instead, banks reallocate toward higher growth of non-C&I assets, especially residential real estate loans and liquid assets.
- Interpretation: banks reallocate lending capacity to other asset types rather than contracting balance sheets in response to reduced commercial lending opportunities.

---

### 3.1 Firm- and MSA-level measurement of intangible capital

### Firm-level construction
- Intangible capital constructed from past investments following Falato et al. (2013) using Compustat accounting data for 1977–2010.
- Sample exclusions: financial firms (SIC codes 6000-6999), regulated utilities (SIC codes 4900-4999), firms with missing or non-positive book values of assets or sales in any given year.
- Final sample: 176,877 firm-year observations for 18,535 unique firms.
- Three capitalized components:
  - Knowledge capital: capitalize past R&D with depreciation rate of 15%; initial stock set equal to the expenditure in the first year divided by the respective depreciation rate; interpolate missing R&D following Hall (1993); set R&D to zero for firms not reporting R&D.
  - Organizational capital: capitalize past SG&A with depreciation rate of 20%; constructed stock weighed by 0.2 following Corrado et al. (2005).
  - Informational capital: capitalize expenditures on computerized information and software with depreciation rate of 31% using (2-SIC) industry BEA Fixed Reproducible Tangible Wealth data.
- Normalization: intangible capital measure normalized by total tangible assets (total balance sheet assets excluding cash).
- Result: average intangible-to-tangible capital ratio at the firm level in mid-2000s is about 120%.

### MSA-level construction and cross-sectional facts
- Firms’ headquarters mapped to MSAs using State/County FIPS merged with Metropolitan Areas and Components data as of 2010; historical headquarters backfilled using Compustat tapes for 1977–1989 and annual data over 1990–2010.
- MSA-level intangible capital defined as the simple (unweighted) average of the intangible capital to total asset ratio of firms headquartered in that MSA.
- Cross-sectional MSA statistics:
  - Average MSA has 73 firms.
  - Average MSA intangible capital ratio (to book assets net of cash): 67.8%.
  - Average growth rate of the intangible capital ratio in the cross-section of MSAs: 14%.
  - Top 5 MSAs (≥10 firms) with average annual intangible capital growth rate ≥ 20% during 1977-2010: Reno-Sparks (NV), Boulder (CO), Riverside-San Bernardino (CA), Denver-Aurora (CO), Salt Lake City (UT).
  - Bottom 5 MSAs with average intangible capital growth < 10%: Grand Rapids-Wyoming (MI), Richmond (VA), Louisville-Jefferson (KY), Akron (OH), Greater Hartford (CT).
- Distributional notes:
  - About 10% of MSA-year observations experience negative growth rates of firms’ intangible capital ratio.
  - Cross-sectional variation in intangible capital growth is large; distribution remains fairly stable over time.

### Matching caveat
- Some firms may borrow from banks located in other MSAs and banks may diversify C&I lending beyond headquarters MSA, especially post-mid-1990s due to deregulation and consolidation; this imperfect matching introduces measurement error.

---

### Empirical identification, robustness, and heterogeneity

### Identification and robustness strategies
- Control for MSA-level non-C&I loan demand variables: population growth, per capita household income growth, and house price growth.
- Use subsamples of MSAs where intangible capital growth is least correlated with non-C&I loan demand measures and MSAs without a house price boom.
- Include MSA×year fixed effects in alternative specifications to absorb all time-varying local economic conditions; focus on interactions of intangible asset growth with bank size and capital when MSA×year FE are used.
- Exploit staggered interstate bank branching deregulation (late 1970s to early 1990s) in differences-in-differences specifications:
  - Compare regulated banks (restricted geographically) with deregulated banks.
  - Regulated banks exhibit greater contraction of commercial lending for a given increase in local intangible capital; effects are significantly smaller for deregulated banks.

### Robustness tests and alternative specifications
- Excluding capitalized IT expenditure from intangible measure and isolating R&D or SG&A components yields results consistent with main specification.
- Adding MSA-level share of firms with a bond market rating and patent measures (unweighted and citation-weighted) preserves the relation between intangible assets and bank portfolio allocations; precision declines in some liquid-asset estimates due to sample size.
- Alternative level specifications, state-level matching, and various clustering schemes for standard errors largely preserve the main patterns.
- Restricting to MSAs where intangible capital growth is negatively correlated, near-zero correlated (< 0.25), or Granger-noncausal with demand proxies (house prices, per capita income, total household income) yields similar significant coefficients on intangible capital growth.

### Key heterogeneity results
- Intangible asset growth affects commercial loan growth primarily for banks with low capital.
- Banks with high capital reallocate assets toward real estate loans; banks with low capital reallocate toward liquid assets.
- Effects concentrated in large and less well capitalized banks, implying bank consolidation and capital erosion since the 1990s may have amplified the impact.

---

### 5.3 Effects on bank profitability

- Test logic: if intangible capital increases demand for non-C&I loans, net interest margins should increase; if frictions in C&I lending dominate, profitability should decline.
- Findings (Table 9):
  - Column (1): Intangible capital growth = -0.0142***
  - Column (2): Intangible capital growth = -0.0110***
  - Column (3): Intangible capital growth = -0.0036*
- Interpretation: higher intangible asset growth is associated with lower bank profitability (net interest margins), consistent with the frictions-in-C&I-lending channel and inconsistent with a pure demand-channel explanation.

---

### 6 Robustness highlights and identification via branching deregulation

### Branching deregulation (difference-in-differences)
- Use interstate branching deregulation as quasi-natural experiment to refine the MSA-level bank-firm link.
- Findings (Table 10, single-state MSAs):
  - IK growth*Regulated (interaction) shows stronger negative relationship with C&I loans in regulated states:
    - Column (1): IK growth*Regulated = -0.0444***
    - Column (2): IK growth*Regulated = -0.0308*
  - IK growth main effects are weaker but often statistically significant in deregulated states, indicating a present but smaller link.

### Additional robustness
- Excluding IT from IK, using R&D- or SG&A-based dependent variables, and adding controls for rated firms and patents do not overturn core results.
- Alternative clustering (bank & year; MSA; MSA & year) leaves main coefficients on IK growth statistically significant at conventional levels in many specifications.

---

### 7 Conclusions and macro implications

### Core conclusion
- Increased corporate intangible capital leads banks to reallocate portfolios away from commercial loans towards non-C&I assets, primarily real estate loans and liquid assets, consistent with intangible assets having lower collateral values.

### Quantitative and aggregate implications
- Estimates suggest the rise in corporate intangible assets since the late 1970s may account for up to 20% of bank portfolio reallocation from commercial loans to real estate loans and liquid assets over the same period.
- Greater intermediation frictions from intangible assets could raise demand for safe assets as corporate savings increase and banks deprived of corporate lending increase demand for MBS and sovereign bonds.
- The channel complements other explanations for the secular decline in safe yields and relates to debates on secular stagnation.

### Differential and policy-relevant patterns
- Impact concentrated in large and less well capitalized banks, suggesting bank consolidation and capital erosion since the 1990s may have amplified the effect of corporate intangible assets on bank portfolios.
- Banks respond to reduced C&I lending opportunities by reallocating to other asset types rather than uniformly shrinking balance sheets.

---

### Key reported estimates and statistics (selected exact figures)
- Baseline IK growth coefficients on bank-level C&I loan growth (Table 2):
  - Column (1): IK growth = -0.0490***
  - Column (2): IK growth = -0.0318***
  - Column (3): IK growth = -0.0204***
- IK growth effects on non-C&I assets (Table 2 and Table 3):
  - Column (4): IK growth = 0.0085***
  - Column (5): IK growth = 0.0161***
  - Residential real estate loans: IK growth = 0.0111**
  - Commercial real estate loans: IK growth = 0.0113*
  - Consumer loans: IK growth = -0.0083*
- Heterogeneity (Table 4):
  - IK growth>p50 on C&I loans: -0.0254***
  - IK growth>p50 on real estate loans: 0.0144***
  - IK growth>p50 on liquid assets: 0.0119**
  - IK growth<p50 on C&I loans: 0.0349**
- Interaction with bank size and capital (Table 5 selected):
  - IK growth*Large bank on C&I loans: -0.0269***
  - IK growth*Low capital on C&I loans: -0.0585***
  - IK growth*Low capital on liquid assets: 0.0470***
- Profitability (Table 9):
  - Net interest margins — Column (1): Intangible capital growth = -0.0142***
  - Column (2): Intangible capital growth = -0.0110***
  - Column (3): Intangible capital growth = -0.0036*
- Time-series and distributional moments (Table 1, Panel A & B):
  - IK growth: No. obs 109,710; Mean 14.7%; St. Dev. 13.9%; Min -31.8%; p25 8.3%; Median 13.8%; p75 19.6%; Max 102.3%
  - C&I loan growth: No. obs 100,115; Mean 6.6%; St. Dev. 24.2%; Min -40.4%; p25 -9.9%; Median 3.5%; p75 19.4%; Max 91.8%
  - Real estate loan growth: No. obs 101,855; Mean 8.7%; St. Dev. 16.6%; Min -20.3%; p25 -2.9%; Median 6.0%; p75 17.2%; Max 70.0%
  - Bank asset growth: No. obs 109,710; Mean 5.1%; St. Dev. 9.0%; Min -10.6%; p25 -1.3%; Median 3.6%; p75 10.0%; Max 33.6%
  - Net interest margin growth: No. obs 76,610; Mean -2.5%; St. Dev. 8.1%; Min -20.2%; p25 -8.2%; Median -3.0%; p75 2.6%; Max 20.8%
  - IK (levels, % assets): No. obs 109,710; Mean 67.8%; St. Dev. 67.3%; Min 0.8%; p25 34.9%; Median 50.4%; p75 76.5%; Max 1428.3%
  - C&I loans (levels, % assets): No. obs 109,710; Mean 12.1%; St. Dev. 8.8%; Min 0.0%; p25 5.8%; Median 10.2%; p75 16.4%; Max 100.0%

*Source: wp17234 (IMF Working Paper excerpts).*

### 1. Introduction  3

### 1. Introduction

### Overview and motivation
- Over the past few decades, the importance of intangible capital in corporate balance sheets has steadily increased.
- Intangible assets are more firm-specific and more difficult to verify, price, and liquidate than tangible assets, and therefore offer less collateral value.
- The rise in intangible capital may increase intermediation frictions: borrowers with large shares of intangible assets may find it difficult to obtain funding through secured lending.
- Existing literature links the rise in intangible capital to lower corporate debt capacity and greater reliance on internal financing; however, how this friction affects the banking sector is unexplored in prior work.

### Key descriptive patterns (aggregate-level)
- For U.S. corporates, the share of intangible assets relative to firms’ tangible assets rose from roughly 40% in the 1970s to over 100% in the 2000s.
- Over the same period, the share of commercial loans in banks’ total assets fell by a third, while the share of real estate loans almost doubled.
- Figures in the source document illustrate these trends for 1977-2010 (e.g., Corporate intangible assets and bank C&I loans, 1977-2010; Corporate intangible assets, and bank loan portfolios, 1977-2010).

### Main hypotheses
- Empirical hypothesis 1: Higher corporate intangible assets reduce the equilibrium volume of bank C&I lending, all else equal.
- Empirical hypothesis 2: Higher corporate intangible assets increase the equilibrium volume of bank non-C&I assets, reduce the equilibrium size of bank balance sheets, or lead to a combination of the two.

### Empirical approach and scope
- Exploit geographical variation in corporate intangible capital and bank portfolio allocations by matching firms and banks based on their headquarters’ Metropolitan Statistical Area (MSA).
- Use a large bank-level panel over 1977-2010 to examine how banks’ commercial loans and total assets respond to growth in intangible capital for firms located in the same MSA as the bank.
- Examine bank non-C&I asset allocations, focusing on growth rates of real estate loans and liquid assets.
- Matching banks to local firms relies on the assumption banks cannot easily diversify away intangible-asset-related shocks geographically; measurement error is likely smaller in the first half of the sample when banks had limited cross-state scope.

### Main empirical findings
- The rise in corporate intangible capital reduces commercial loan growth, controlling for bank fixed effects, time-varying bank characteristics, and MSA-level loan demand variables.
- In the main specification, coefficient estimates indicate that a one standard deviation increase in intangible capital is accompanied by a contraction in bank C&I lending by 2 to 3.5 percentage points.
- This effect is statistically and economically significant and implies that up to 20% of the secular decline in the share of commercial loans in total bank lending since the mid-1970s can be attributed to the increase in corporate intangible capital over the same period.
- Banks exposed to higher growth of MSA-level corporate intangible capital do not reduce their size; rather, bank balance sheets show higher growth of non-C&I assets, especially residential real estate loans and liquid assets.
- Interpretation: banks reallocate lending capacity to other asset types rather than contracting balance sheets in response to reduced commercial lending opportunities.

### Identification strategy and robustness checks
- Address potential alternative channels whereby intangible capital could affect bank lending via increased local demand for non-C&I loans (e.g., due to different workforce composition) through:
  - Controlling for MSA-level non-C&I loan demand variables: population growth, per capita household income growth, and house price growth.
  - Confirming results in subsamples of MSAs where intangible capital growth is least correlated with measures of non-C&I loan demand, and in MSAs without a house price boom.
  - Including interacted MSA×year fixed effects to control for all unobserved time-varying MSA-level macroeconomic factors.
- Exploit staggered deregulation of interstate bank branching (late 1970s to early 1990s) in differences-in-differences specifications:
  - Compare banks facing geographic restrictions (“regulated”) with those not facing restrictions (“deregulated”).
  - For a given increase in local intangible capital, regulated banks (less geographically diversified) exhibit a greater contraction of commercial lending.
  - Point estimates for regulated banks are sizable and greater than baseline; effects are significantly smaller for deregulated banks, consistent with the intangible-capital shock reducing firms’ borrowing capacity rather than unobserved increases in non-C&I loan demand.

### Broader implications and contributions
- The rise in corporate intangible capital is a quantitatively important contributor to the long-run reallocation of bank lending away from commercial loans toward real estate loans and liquidity holdings.
- This channel complements existing explanations for the shift in bank lending (e.g., rise in house prices, securitization, development of securities markets).
- Implication: even if real estate demand subsides, supply pressures in real estate lending and banks’ demand for liquid assets—and the downward pressure on interest rates—may persist because they partly stem from the long-run transformation of the corporate sector toward intangible assets.
- The study is the first to provide empirical evidence linking aggregate increases in corporate intangible capital to bank portfolio reallocation patterns.

*Source: wp17234 - 1. Introduction (excerpts) from the provided IMF Working Paper content.*

### 3.1    Firm and MSA-level measures of intangible capital

### 3.1    Firm and MSA-level measures of intangible capital

### Measurement of firm-level intangible capital
- Intangible capital is constructed from past investments in intangible assets following Falato et al. (2013).
- Sample: Compustat accounting data between 1977 and 2010.
  - Exclusions: financial firms (SIC codes 6000-6999), regulated utilities (SIC codes 4900-4999), firms with missing or non-positive book values of assets or sales in any given year.
  - Final sample: 176,877 firm-year observations for 18,535 unique firms.
- Intangible capital for each firm-year = sum of capital accumulated through three types:
  - Knowledge capital: capitalize past R&D expenditures using the perpetual inventory method with depreciation rate of 15% (Hall et al., 2001).
    - Initial stock set equal to the expenditure in the first year divided by the respective depreciation rate.
    - For R&D, interpolate missing values following Hall (1993); set R&D to zero for firms that do not report R&D.
  - Organizational capital: capitalize past SG&A expenses with depreciation rate of 20% (Lev and Radhakrishnan, 2005; Eisfeldt and Papanikolaou, 2013).
    - Constructed stock of organizational capital is weighed by 0.2 following Corrado et al. (2005) to account for SG&A including many non-investment expenses.
  - Informational capital: capitalize expenditures on computerized information and software with depreciation rate of 31% following BEA, using (2-SIC) industry BEA Fixed Reproducible Tangible Wealth data.
- Normalization: intangible capital measure normalized by total tangible assets (total balance sheet assets excluding cash).
- Resulting estimate: average intangible-to-tangible capital ratio at the firm level in mid-2000s is about 120%, comparable to Corrado et al. (2005).

### Construction of MSA-level intangible capital
- Firms’ headquarters state and county from Compustat are mapped to MSAs by merging State/County FIPS code with Metropolitan Areas and Components data defined by the Office of Management and Budget as of 2010.
  - Note on Compustat location data: Compustat reports only current state and county of headquarters. Physical Compustat tapes were used to collect historical headquarters information annually over 1990-2010 and backfill 1977-1989 (see Falato et al. (2013)).
- MSA-level intangible capital defined as the simple (unweighted) average of the intangible capital to total asset ratio of firms headquartered in that MSA.
- Cross-sectional MSA statistics:
  - The average MSA in the sample has 73 firms.
  - Average MSA intangible capital ratio (to book assets net of cash): 67.8%.
  - Average growth rate of the intangible capital ratio in the cross-section of MSAs: 14%.
  - Top 5 more populated MSAs (with at least 10 firms) with average annual intangible capital growth rate ≥ 20% during 1977-2010: Reno-Sparks (NV), Boulder (CO), Riverside-San Bernardino (CA), Denver-Aurora (CO), Salt Lake City (UT).
  - Bottom 5 MSAs where corporate intangible capital grew on average at below 10%: Grand Rapids-Wyoming (MI), Richmond (VA), Louisville-Jefferson (KY), Akron (OH), Greater Hartford (CT).
- Distributional notes:
  - About 10% of MSA-year observations experience negative growth rates of firms’ intangible capital ratio.
  - Cross-sectional (cross-MSA) variation in intangible capital growth is large; the distribution of growth rates remains fairly stable over time.

### Matching firms to banks at the MSA level (empirical context)
- Empirical approach examines effect of intangible capital for firms in a given MSA on asset allocations for banks headquartered in the same MSA.
- Caveat: some firms may borrow from banks located in other MSAs and some banks may diversify C&I lending beyond headquarters MSA, especially since mid-1990s due to deregulation and consolidation. This imperfect matching introduces measurement error (addressed in Section 6.1).

*Source: wp17234 - 3.1    Firm and MSA-level measures of intangible capital*

### 5.1    Control for unobserved macro factors

### 5.1    Control for unobserved macro factors

### MSA×year fixed effects versus MSA macro controls
- Specification: bank size and capital interactions estimated using MSA×year fixed effects instead of macro controls.
- Rationale: MSA×year fixed effects absorb all time-varying local economic conditions, including unobserved factors driving demand for non-C&I loans; intangible asset growth varies at MSA-year level so its standalone coefficient cannot be estimated with MSA×year fixed effects.
- Focus: interaction of intangible asset growth with bank size and capital to test consistency with Table 5.
- Key findings:
  - Coefficients on intangible capital growth interacted with the dummy for large bank size (holding low bank size as the base) are insignificant throughout (columns 1, 4, 7, 10), consistent with Table 5 where coefficients for large and small banks are not statistically different based on the p-value of the F-test.
  - Coefficients on intangible capital growth interacted with the dummy for low bank capital (holding high bank capital as base) are significant (columns 2, 5, 8, 11), consistent with Table 5 where high- and low-capital banks differ statistically (p-value of the F-test).
  - Interactions with indicators for large banks with low capital, large banks with high capital, and small banks with low capital (holding small banks with high capital as the base) yield coefficients that are statistically significant when they differed from small banks with high capital in Table 5 (columns 3, 6, 9, 12).
  - Both tables indicate that:
    - intangible asset growth affects commercial loan growth primarily for banks with low capital (columns 3 in both tables);
    - it leads to a contraction of assets of larger magnitude in all groups of banks compared to the contraction (or expansion) of assets in small banks with high capital (columns 6 of both tables);
    - in response to corporate intangible asset growth banks with high capital reallocate their assets towards real estate loans whereas banks with low capital reallocate to liquid assets (columns 6 and 12 in both tables).
  - Point estimates in Table 6 are almost identical to the difference between the point estimates of the respective coefficients in Table 5.
- Interpretation: Macro controls in baseline specifications appear to capture the local macroeconomic environment (including demand-channel factors), making an omitted-variable “demand channel” explanation unlikely to drive the main results.

### Footnote and additional control
- In all specifications we further control for interactions of house price growth with indicators of bank size and capital quantiles to allow for the demand channel to vary with bank conditions (see, e.g., Chakraborty et al. (2016)).

### Conclusion from this subsection
- Estimated coefficients on the interaction of intangible assets and bank characteristics are virtually identical in statistical significance and economic magnitudes when using macro controls versus MSA×year fixed effects, supporting the baseline control strategy.

---

### 5.2    Exploit correlation of intangible assets and macro variables

### Strategy to reduce demand-channel influence
- Identify subset of MSAs where corporate intangible capital growth is least related to proxies of non-C&I loan demand.
- Demand proxies used:
  - (a) house prices,
  - (b) per capita household income,
  - (c) total household income (to account for changes in population).
- Empirical observation: Histograms of simple correlation coefficients between MSA-level corporate intangible capital growth and each proxy have negative mean and median (reported in Panel C of Table 1), casting doubt on a positive association between intangible capital growth and demand for non-C&I loans.

### Formal selection criteria for “least relation”
- Conditions for an MSA to be considered “least related” to a demand proxy:
  - (i) negative correlation between corporate intangible capital growth and the proxy,
  - (ii) near-zero correlation (lower than 0.25 in absolute value),
  - (iii) a rejected hypothesis of Granger causality from corporate intangible capital growth to the proxy (at the 90% level of confidence).
- Each condition leads to dropping between one third and two thirds of the sample.

### Results
- Table 7 (summary): coefficients on corporate intangible capital growth remain significant in subsamples defined by the above conditions, with point estimates close to the baseline (Table 2).
- Interpretation: Baseline findings are not driven by MSAs with high correlation between corporate intangible shocks and non-C&I loan demand (e.g., housing); instead results stem mainly from MSAs where such correlation is low or negative.

### Housing-boom robustness
- Table 8: restrict to MSAs that did not experience a house price boom in 2000-2007 (columns 1-3) and 1990-2007 (columns 4-6).
- Definition of “housing boom” MSAs: house price growth above the 50th (Panel A), 75th (Panel B), and 90th (Panel C) percentile across all MSAs.
- Finding: In all MSAs without a housing boom, coefficients on intangible capital growth remain statistically significant with point estimates close to baseline (Table 2), suggesting results are not driven by contemporaneous housing booms.

---

### 5.3    Effects on bank profitability

### Test logic
- If corporate intangible capital increases demand for non-C&I loans, bank profitability (net interest margins) should increase.
- If higher frictions in C&I lending prevail, bank profitability should decline.

### Specification and controls
- Table 9: presentation mirrors Table 2 — column 1: no controls; column 2: adds macro controls; column 3: adds bank and year fixed effects.
- Macro controls enter with expected signs: better local conditions associated with higher bank profitability.
- Additional correlates: larger banks and more rapidly growing banks are less profitable; bank capitalization is positively related to profitability.

### Findings
- Across specifications, higher intangible asset growth is associated with lower bank profitability (net interest margins).
- Interpretation: Results are consistent with the frictions-in-C&I-lending channel and inconsistent with a pure demand-channel explanation.

---

### 6    Robustness Tests

### 6.1 The effects of bank branching deregulation
- Identification: exploit interstate bank branching deregulation (mid-1970s to mid-1990s) as a quasi-natural experiment to refine the bank-firm link at the MSA level.
- Regulated branching used as a proxy for a more precise MSA-level bank-firm link (banks less able to lend outside headquarters MSA/state).
- Difference-in-differences results (Table 10) for single-state MSAs:
  - Interaction “IK growth×Regulated” consistent with a stronger relationship between corporate intangible assets and bank portfolio allocations in states with regulated branching.
  - Coefficients for MSAs in deregulated states (term “IK growth”) are weaker but statistically significant in 5 of 6 specifications (columns 1-6), indicating a weaker yet present link.
- Interpretative test: If corporate intangible assets operated via increased demand for non-C&I loans, regulated MSAs (tighter funding) would shift relatively less to non-C&I assets; instead, they shifted more to non-C&I assets — inconsistent with tighter funding constraints and consistent with a lower ability to geographically reallocate C&I lending.

### 6.2 Alternative specifications and further refinements
- Table A1:
  - Uses measure of corporate intangible capital that excludes capitalized IT expenditure (columns 1-4).
  - Also uses dependent variables capitalized R&D expenditure (columns 5-8) and capitalized SG&A expenditure (columns 9-12).
  - Results fully consistent with main specification.
- Table A2:
  - Includes share of firms with a bond market rating at MSA-year level (columns 1-4) to account for access to alternative financing.
  - Adds MSA-level control for number of patents, unweighted and citation-weighted (columns 5-12).
  - Relationship between corporate intangible assets and bank portfolio allocations continues to hold; precision declines for liquid assets (columns 8, 12) likely due to sample size drop, but magnitudes remain close to baseline (Table 2, column 12).
- Table A3:
  - Captures dynamics of bank portfolio allocations in levels instead of growth rates (columns 1-4).
  - Limits sample to MSAs with at least 3 Compustat firms (median number is 5) in columns 5-8.
  - Matches banks and firms on state instead of MSA in columns 9-12.
  - Results largely robust.
- Table A4:
  - Alternative clustering of standard errors: cluster on bank and year (columns 1-4), on MSA (columns 5-8), and on MSA and year (columns 9-12).
  - Main coefficients on intangible capital growth remain statistically significant at conventional levels; for liquid assets coefficients are statistically significant at the 15% level in these cases (columns 4, 8, 12).

---

### 7    Conclusions

- Main result: Increased corporate intangible capital leads banks to reallocate portfolios away from commercial loans towards non-C&I assets, primarily real estate loans and liquid assets, consistent with intangible assets having lower collateral values.
- Multiple empirical tests rule out a demand-driven reallocation in favor of a supply-side frictions channel in C&I lending.
- Novelty: First paper, to the authors' knowledge, studying how the rising share of intangible assets on corporate balance sheets affects bank balance-sheet structure.
- Aggregate magnitudes:
  - Estimates suggest the rise in corporate intangible assets since the late 1970s may account for up to 20% of bank portfolio reallocation from commercial loans to real estate loans and liquid assets over the same period.
- Macro implications:
  - Greater intermediation frictions from corporate intangible assets could raise demand for safe assets as corporate savings increase and banks deprived of corporate lending increase demand for MBS and sovereign bonds.
  - Findings may complement explanations for the secular decline in safe yields and debates on secular stagnation.
- Differential effects:
  - Impact concentrated in large and less well capitalized banks, implying that bank consolidation and capital erosion since the 1990s may have amplified the effect of corporate intangible assets on bank portfolios.
- Overall: The shift in corporate asset composition has produced a significant shift in banking-sector asset composition over the last four decades.

*Source: IMF Working Paper section "5.1    Control for unobserved macro factors" (wp17234).*

### References

### References

### Cited literature
- List of cited works includes academic articles, speeches, technical reports, and working papers by authors such as Alessandri, P. and Haldane, A. G. (2009); Bates, T. W., Kahle, K. M., and Stulz, R. M. (2009); Berger, A. N. and Mester, L. J. (1997); Bester, H. (1985); Brown, J. R., Fazzari, S. M., and Petersen, B. C. (2009); Corrado, C., Hulten, C., and Sichel, D. (2005, 2009); Falato, A., Kadyrzhanova, D., and Sim, J. (2013); Gorton, G. B. and Ordonez, G. (2013); Hall, B. H., Jaffe, A. B., and Trajtenberg, M. (2001); Hart, O. and Moore, J. (1994); Rajan, R. and Winton, A. (1995); Shin, H. S. (2012); Stein, J. C. (2002); Summers, L. H. (2015); and others cited in the source.

### Empirical findings — high-level summary of regression results (preserving reported estimates exactly)
- Baseline relationships (Table 2)
  - IK growth coefficients on bank-level C&I loan growth:
    - Column (1): IK growth = -0.0490***
    - Column (2): IK growth = -0.0318***
    - Column (3): IK growth = -0.0204***
  - IK growth coefficients on total bank asset growth and liquid assets (selected columns):
    - Column (4): IK growth = 0.0085***
    - Column (5): IK growth = 0.0161***
    - Column (11): IK growth = 0.0148***
    - Column (12): IK growth = 0.0092**
  - House price (HP) growth and population growth are consistently positive and significant correlates of real estate and bank asset measures (e.g., HP growth = 0.1752*** in column (3); Population growth = 0.6212*** in column (3)).
  - Bank asset growth exhibits large positive coefficients across specifications (e.g., 0.6945***, 0.6851***, 0.6107*** in various columns).

- Decomposition of non-C&I loans (Table 3)
  - IK growth effects on components:
    - Residential real estate loans: IK growth = 0.0111**
    - Commercial real estate loans: IK growth = 0.0113*
    - Consumer loans: IK growth = -0.0083*

- Heterogeneity by magnitude of intangible capital change (Table 4)
  - For IK growth > p50:
    - C&I loans: IK growth>p50 [1] = -0.0254***
    - Real estate loans: IK growth>p50 [1] = 0.0144***
    - Liquid assets: IK growth>p50 [1] = 0.0119**
  - For IK growth < p50:
    - C&I loans: IK growth<p50 [2] = 0.0349**

- Heterogeneity by bank size and capital (Tables 5 and 6) — selected interaction estimates
  - IK growth*Large bank [1]:
    - C&I loans (Table 5 col 1): -0.0269***
    - Bank assets (Table 5 col 2): 0.0163***
    - Real estate loans (Table 5 col 3): 0.0151***
    - Liquid assets (Table 5 col 4): 0.0122**
  - IK growth*Low capital [3] (Table 5):
    - C&I loans: -0.0585***
    - Bank assets: 0.0103***
    - Liquid assets: 0.0470***
  - Alternative specification with MSA*Year FE (Table 6) highlights similar patterns for IK growth*Low capital:
    - C&I loans: -0.0525***
    - Liquid assets: 0.0660***

- Ruling out loan-demand channels (Tables 7 and 8)
  - Restricted samples where intangible capital growth is negatively correlated with or uncorrelated to macro variables still show negative IK growth coefficients on C&I loans:
    - Example (Table 7 Panel A, Negative correlation sample): C&I loans coefficient estimate = -0.0259*** (SE (0.009))
  - Samples excluding housing booms (Table 8):
    - No housing boom 2000-2007, HP increase below median: C&I loans coefficient estimate = -0.0296*** (SE (0.010))

- Bank profitability (Table 9)
  - Net interest margins — IK growth:
    - Column (1): Intangible capital growth = -0.0142***
    - Column (2): Intangible capital growth = -0.0110***
    - Column (3): Intangible capital growth = -0.0036*
  - House price growth and Pc income growth are positive predictors in some specifications (e.g., House price growth = 0.1532***; Pc income growth = 0.0720*** in column (3)).

- Identification using interstate bank branching deregulation (Table 10)
  - Interaction IK growth*Regulated (pre-deregulation states) shows negative and significant effects on C&I loans in some specifications:
    - Column (1): IK growth*Regulated = -0.0444***
    - Column (2): IK growth*Regulated = -0.0308*
  - IK growth main effects vary in sign and significance across columns (e.g., IK growth = 0.0140** in column (1); IK growth = -0.0151*** in column (4)).

- Robustness and components (Appendix Tables A1–A4) — selected results
  - Excluding IT from intangible capital, and isolating R&D or SG&A components, yields variation in IK growth coefficients but generally preserves the negative association with C&I lending in many specifications (e.g., A1 column (1) IK growth = -0.0192***; columns (9)–(12) also report significant coefficients).
  - Additional controls (Table A2) such as % rated firms and patent measures do not overturn core patterns; IK growth estimates remain negative and significant in many columns (e.g., A2 column (1) IK growth = -0.0197***; column (5) IK growth = -0.0415***).
  - Alternative approaches and clustering (Tables A3, A4) show the results are sensitive to specification choices but retain the main patterns: negative associations between IK growth and C&I lending and positive associations between IK growth and real estate/liquid asset allocations in several specifications (e.g., Table A4 IK growth = -0.0204** with various clustering schemes).

### Key statistics and data notes (from figures and Table 1)
- Growth and level statistics (Table 1, Panel A & B — exact reported moments)
  - IK growth: No. obs 109,710; Mean 14.7%; St. Dev. 13.9%; Min -31.8%; p25 8.3%; Median 13.8%; p75 19.6%; Max 102.3%
  - C&I loan growth: No. obs 100,115; Mean 6.6%; St. Dev. 24.2%; Min -40.4%; p25 -9.9%; Median 3.5%; p75 19.4%; Max 91.8%
  - Real estate loan growth: No. obs 101,855; Mean 8.7%; St. Dev. 16.6%; Min -20.3%; p25 -2.9%; Median 6.0%; p75 17.2%; Max 70.0%
  - Bank asset growth: No. obs 109,710; Mean 5.1%; St. Dev. 9.0%; Min -10.6%; p25 -1.3%; Median 3.6%; p75 10.0%; Max 33.6%
  - Net interest margin growth: No. obs 76,610; Mean -2.5%; St. Dev. 8.1%; Min -20.2%; p25 -8.2%; Median -3.0%; p75 2.6%; Max 20.8%
  - IK (levels, % assets): No. obs 109,710; Mean 67.8%; St. Dev. 67.3%; Min 0.8%; p25 34.9%; Median 50.4%; p75 76.5%; Max 1428.3%
  - C&I loans (levels, % assets): No. obs 109,710; Mean 12.1%; St. Dev. 8.8%; Min 0.0%; p25 5.8%; Median 10.2%; p75 16.4%; Max 100.0%
- Figures depict time-series and distributional patterns for 1977-2010:
  - Figure 1: Corporate intangible assets and bank C&I loans, 1977-2010 (series plotted as average share of corporate intangible capital in total assets for US Compustat firms and average share of C&I loans in total loans for US commercial banks).
  - Figure 2: Corporate intangible assets and bank loan portfolios, 1977-2010 (C&I loans, RE loans, corporate intangible capital).
  - Figure 3: Growth rate of corporate intangible assets, 1977-2010 (histogram and MSA-year time-series distribution).
  - Figure 4: Histograms of simple correlation coefficients between intangible asset growth and house price growth, per capita household income growth, and total MSA household income growth (computed per MSA over 1977-2010 conditional on at least 10 consecutive observations).

*Source: wp17234 - References, wp17234 - References (PDF). *

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17234.pdf_