The historic rise of financial profits in the U.S. economy

Abstract

The ratio of financial to nonfinancial profits in the U.S. economy has risen greatly during the last four decades, a period often called the financialization of capitalism, but the reasons for the increase are not well understood. This article tackles the issue by developing a model that incorporates the relationships between financial and nonfinancial capitalists that are characteristic of the period of financialization. It is shown that the ratio of financial to nonfinancial profits depends positively on the net interest margin and the noninterest income of banks, but negatively on the general rate of profit, the noninterest expenses of banks, and the ratio of the capital stock to interest-earning assets. Empirical estimation for the United States strongly supports the model and reveals that financial profits have varied mainly with respect to the net interest margin, although non-interest income has also been important.

Keywords:

• Financial profits
• financialization
• U.S. economy
• political economy of finance

JEL CLASSIFICATION:

• E10
• E44
• G20

Notes

1 See Greenwood and Scharfstein (2013) and Philippon and Reshef (2013) for recent studies compiling evidence for the United States and other developed countries.

2 Ideally the measure of financial profits should also include profits made by other economic agents through financial activities (e.g., profits made by households through trading in financial assets) as well as profits made by nonfinancial corporations through engaging in purely financial activities (e.g., through share transactions; see Lapavitsas 2013). Note that the financial profits measured by the U.S. Bureau of Economic Analysis, and used to construct Figures 1 and 2, include profits made by Federal Reserve banks, commercial banks, savings institutions, other depository credit intermediaries, credit unions, federally sponsored credit agencies, nondepository credit intermediaries, securities, commodity contracts and investments, insurance agencies and brokerages, real estate investment trusts, regulated investment companies and other financial vehicles, bank holding companies and other holding companies; see Chapter 13: Corporate Profits in Bureau of Economic Analysis (BEA) (2017). Because of data limitations, however, the econometric analysis carried out in Section 3 considered only profits made by insured commercial banks, which are measured by the Federal Deposit Insurance Corporation. Commercial banks’ profits are a proxy for financial profits; see also Footnote 16.

3 For the relevant theoretical discussion, see Lapavitsas (2013).

4 For useful surveys, see Caverzasi and Godin (2015) and Nikiforos and Zezza (2017).

5 Financialization has brought an increasing penetration of household income and wealth by formal finance, and a fuller model would include a household sector to capture the extraction of financial profit out of household income as “financial expropriation” (Lapavitsas 2013). However, for the purposes of explaining the division of total profit the added complexity of a household sector would offer little extra insight. The main conclusions regarding the ratio of financial to nonfinancial profits can still be derived by simply assuming a financial and a nonfinancial enterprise sector.

6 This approach to analyzing finance in a capitalist economy draws on the Uno School of Japanese Marxism; for further detail, see Itoh and Lapavitsas (1999, Chapters 3 and 4).

7 This is the reason why do not consider it necessary to define a stock revaluation matrix.

8 In a more complex model, it would also be possible to include further services relating to trust accounts and long-run certificates of deposit, annuity contracts, securities brokerage, underwriting, and other fee-earning activities. Note also that some of the “financial services” recorded in the National Income and Product Accounts (NIPA) data might actually include interest as, for instance, when late fees and penalties are charged on credit cards and take the form of a sharp rise in interest charged. We owe this point to one of the reviewers of the article. A more comprehensive examination of financial profits would also take account of such data issues. We leave these issues for future research.

9 Similarly, if a household sector had been included, it would have been possible also to incorporate the relevant aggregate flows without altering the thrust of the analysis. Note that Assa (2016) stresses that the revenues and costs of the financial system, far from being an addition to the value of the annual output of the economy, represent a net cost. Assa subtracts the “value added” of finance to derive a measure of “Final GDP”.

10 This is a fundamental condition of Marxist analysis of finance (especially banking); see Itoh and Lapavitsas (1999, pp. 95–96) and Shaikh (2016, pp. 449–452).

11 Since wage-labor has been left out of account, we do not explicitly consider cost conditions.

12 Note that the assets, $\mathrm{L},$ of the nonfinancial sector do not generate the average rate of profit but earn interest. Therefore, the financial sector measures its profitability with respect to its own capital advanced, i.e., $\mathrm{G},$ which is then equalized against the rate of profit for the entire capital advanced, $\mathrm{K}.$ There is, indeed, an asymmetry as $\mathrm{K}$ is an asset but $\mathrm{G}$ is a liability. The asymmetry captures the difference between the nature of the two sectors: productive versus intermediary.

13 Note that the assets-to-equity ratio or leverage ratio of the non-financial sector in this simplified two-sector model would be $(\mathrm{K}+\mathrm{B})/\mathrm{S}.$

14 It is worth noting that there might well be financial profit hiding in these non-interest expenses, particularly in the form of salaries and bonuses. Quantifying this additional source of financial profit is also left for future research.

15 By definition, any theoretical model is a simplification of reality, and it would be a mistake to consider that the model deployed in the previous section could be deterministic. The model has deployed simplifying assumptions to isolate the most relevant effects accounting for the rise of financial profits relative to total profits. Econometric analysis is required to test the relevance and predictions of the theoretical model and to avoid incorrect conclusions.

16 Econometric analysis could have also considered the profits made by U.S. shadow banks in the course of financialization by including, for example, a broader range of financial profits, such as the measure provided by the BEA (which was used to construct Figures 1 and 2). However, there is no official macroeconomic data that would allow for the construction of a reliable dataset corresponding to several of the relevant theoretical concepts discussed in the section on theorizing financial profit. For the purposes of this article, the best estimate of financial profits is given by the profits of financial institutions, that is, mostly banks. This is evidently an underestimate of the total, but there is little doubt that commercial banks remain the pivot of the financial system as well as one of the main agents of financialization. Their profits are a reasonable proxy for financial profits.

17 The FDIC calculates noninterest income as the annualized income from bank services and sources other than interest-bearing assets; see Part II: CAMELS, Section 5.1: Earnings in FDIC (2018). Noninterest income is, thus, largely of a fee nature: service charges on deposits, trust department income, mortgage servicing fees, certain types of loan and commitment fees, and the results of trading operations and a variety of miscellaneous transactions. To the best of our knowledge, the only table published by the FDIC in which the components of non-interest income are shown is Table CB07: Noninterest Income and Noninterest Expense, which shows only two categories associated with noninterest income: Service Charges on Deposit Accounts and Other Noninterest Income. For the period 1955–2014, the former represents approximately 24.1% of total noninterest income, whereas the latter represents the remainder (75.9% of total noninterest income).

18 It is worth mentioning that the models were also tested for cointegration in the context of vector autoregressive models using the methodology developed by Johansen (1991, 1995). These results are available on request. In brief, Johansen’s cointegration tests find no evidence of cointegration during the period 1955–2014; and show the presence of two cointegrating equations during the period 1974–2014. However, there are no theoretical foundations to provide guidance for a clear-cut identification strategy using the vector autoregressive and vector error correction methodologies. Without such foundations, it is not possible to provide meaningful impulse-response functions and variance decompositions analyses. In the same vein, as documented by Pesaran and Shin (1999), the small sample properties of the bounds testing approach are superior to that of the traditional Johansen (1991, 1995) cointegration approach, which typically requires a large sample size for the results to be valid. Because of these reasons, it is more appropriate to follow the single equation settings shown in Equations (8a)(8a) $${\mathit{FP}}_t={\beta }_{1,0}+{\beta }_{1,1}{\mathit{NIM}}_t+{\beta }_{1,2}{\delta }_t+{\beta }_{1,3}{r}_t+{\beta }_{1,4}{\mathit{KL}}_t+{\beta }_{1,5}{\gamma }_t+u_{1,t}$$(8a) and (8b)(8b) $${\mathit{FP}}_t={\beta }_{2,0}+{\beta }_{2,1}{\mathit{NIM}}_t+{\beta }_{2,2}{\delta }_t+{\beta }_{2,3}{\mathit{PL}}_t+{\beta }_{2,4}{\gamma }_t+u_{2,t}$$(8b) .

19 It is also possible to corroborate the presence of statistically significant breakpoints if other dates after 1974 are considered. As a robustness check, we corroborated the presence of a structural break in 1985 and we estimated the models for the period 1985–2014. The estimation results remained qualitatively the same and are available on request.

20 Pesaran et al. (2001) pointed out that the asymptotic theory of the bounds tests is not affected by the inclusion of such “one-off” dummy variables as long as the fraction of periods in which the dummy variables are nonzero tends to zero with the sample size $T.$

21 The estimation results presented in Tables 6 through to Table 8 correspond to error correction models with restricted intercept and no deterministic trend term because the latter was found to be statistically non-significant in all cases.

22 The meaning of the difference related to this short-run effect is not clear, but it does not affect the overall balance of empirical analysis.

Additional information

Notes on contributors

Costas Lapavitsas

Costas Lapavitsas, Department of Economics, SOAS University of London.

Ivan Mendieta-Muñoz

Ivan Mendieta-Muñoz, Department of Economics, University of Utah.