Determinants of De Novo Bank Formation

From 2010 to 2024, the average number of new, or de novo, banks chartered each year was less than seven-the lowest average since the 1960s. The near absence of de novo banks in the wake of the Great Recession stands in stark contrast to previous business downturns, when new charters surged as economic growth returned. The recent sharp decline in de novo banks may have important implications for the banking industry. De novo banks promote vitality and competition in their local markets and are also typically community banks, which serve as an essential source of credit for small businesses and agricultural borrowers.

Stephen Jones, Forest Myers, and Jim Wilkinson examine key drivers behind the recent decline in de novo bank formation. They find that while economic conditions and regulatory burden have continued to influence de novo bank formation over time, technological changes appear to have primarily driven the recent decline in de novo bank formation. Although economic and regulatory policy changes may be able to increase the number of de novo banks, they are unlikely to be sufficient to return this number to its pre-recession level.

Article Citation

Jones, Stephen, Forest Myers, and Jim Wilkinson. 2026. "Determinants of De Novo Bank Formation." Federal Reserve Bank of Kansas City, Economic Review, vol. 111, no. 4. Available at External Linkhttps://doi.org/10.18651/ER/v111n4JonesMyersWilkinson

Introduction

Since the Great Recession in 2007-09, very few de novo banks-new banks-have been chartered. The near absence of de novo banks stands in stark contrast to past bank charter behavior. Previously, every business downturn was followed by a resurgence of new charters as economic growth returned. For example, from 1960 to 2006, 185 de novo banks were chartered each year, on average. During the 2010-24 period, however, the average annual number of de novo banks was less than seven. In no other period since the 1960s has the number of de novo banks chartered been this low.

The sharp decline in de novo banks may have important implications for the banking industry. De novo banks promote vitality and competition in their local markets and provide access to banking services for underserved communities and groups (Jones, Myers, and Wilkinson 2022). De novo banks are also typically community banks, which serve as an essential source of credit for small businesses and agricultural borrowers. Understanding the reasons behind the drop in de novo bank formation may help policymakers support growth and competition in the banking sector and improve access to banking services.

In this article, we examine key drivers behind the recent decline in de novo bank formation. We find that while economic conditions and regulatory burden have continued to influence de novo bank formation over time, technological changes appear to have primarily driven the recent decline in de novo bank formation. Although economic and regulatory policy changes may be able to increase the number of de novo banks, they are unlikely to be sufficient to return this number to its pre-recession level.

Section I documents patterns of de novo bank formation since 1960 and identifies factors that may have contributed to the recent decline in new bank charters. Section II describes our data and introduces a model for de novo bank formation. Section III presents our empirical findings and regression results and identifies the most salient factors dampening recent de novo formation.

I. De Novo Bank Formation

De novo banks are an important feature of the U.S. banking system. First, the de novo process remains the primary vehicle for establishing new community banks despite a growing subset of specialized or digital-native charters. Second, de novo entry into local banking markets helps maintain banking competition (Adams and Gramlich 2014). And third, de novo banks help provide financial and credit services to small businesses as well as underserved communities with limited access to banking products (Bowman 2021). Relative to larger banks, de novo banks are more likely to rely on relationship banking, offering a more personalized touch in their customer dealings and incorporating qualitative or "soft" data into their credit reviews, which can expand credit access for some customers.^_

From 1960 through 2023, over 9,000 de novo banks were chartered in the United States, although the number chartered each year varied widely during this period. Panel A of Chart 1 shows that patterns in total numbers of de novo bank charters generally track the U.S. business cycle and consolidation in the banking industry, with rapidly rising peaks followed by equally rapid declines. Since 1983, however, the number of de novo banks has rebounded to successively smaller peaks, suggesting new bank charters had begun to wane long before the current charter decline began in 2009. Scaled charter data (Panel B) show a consistent pattern of decline and recovery as well. The ratio of new banks to total banks peaks each cycle around 2.5 percent until the Great Recession, when the pattern disappears.

Note: Gray bars denote National Bureau of Economic Research (NBER)-defined recessions.

Sources: NBER, FDIC, and FFIEC Call Reports (Board of Governors of the Federal Reserve System).

During the Great Recession in 2007-09, the number of de novo bank charters dropped significantly. Commentators at the time attributed the scarcity of new banks to the economic downturn, the worst since the Great Depression. They also pointed to a persistent low-interest-rate environment and a post-recession regulatory increase as culprits in the dearth of new charters. For example, Adams and Gramlich (2014) examine county data from 1976 to 2013 and find that as much as 75-80 percent of the decline in new charters could be traced to the weak economy and the low-interest-rate environment. They also find that laws and regulation could be a contributing factor.

Over the last decade, however, the number of de novo banks has remained low despite a strengthening economy and higher interest rates. The U.S. economy has grown by an average annual rate of 2.3 percent over the past 10 years. Although short-term rates were near zero for many years, the Federal Reserve began raising its policy rate in 2015 to a high of 2.5 percent in 2018, before dropping rates back to zero in 2020. In addition, to combat the run-up in inflation following the COVID-19 pandemic, the Federal Reserve increased the target range for the federal funds rate from near zero to a peak of 5.25-5.50 percent between March 2022 and July 2023. As Chart 2 shows, a higher federal funds rate resulted in a marked increase in operating income at community banks.^_ However, the increase in interest rates was not enough to encourage significant charter formation.

Note: Gray bars denote National Bureau of Economic Research (NBER)-defined recessions.

Sources: FFIEC Call Reports (Board of Governors of the Federal Reserve System) and NBER.

Regulatory burden may have depressed de novo formation to some degree, as Adams and Gramlich (2014) suggest.^_ The regulatory environment affects incentives to open a de novo bank in several ways. Some regulations impose costly requirements on banks, reducing profitability and thereby the attractiveness of opening a new bank.^_ Other regulatory changes, such as geographic and product deregulation, may create new opportunities for banks and thereby make opening a new bank more attractive.^_ Whether these regulations make operations easier or more difficult for banks, all new banking laws create implementation costs for bankers. The burden of implementing new regulatory changes is likely to be greater for new and small banks that do not have economies of scale in staffing. In this way, new and changed regulations may decrease investor interest in de novo banks.

One possible indicator of regulatory change is increasing overhead costs, or noninterest expense, as banks may need to hire additional personnel (including some with specialized skillsets), invest in additional technology and infrastructure, or pay for consulting or auditing fees to achieve compliance. Noninterest expense is an imperfect measure of regulatory burden as it encompasses more than just regulatory costs. Furthermore, regulatory burden may manifest in other ways, such as needing to exit certain products or services or needing to reprioritize resources. Nevertheless, contrary to the prevailing narrative regarding regulatory pressure, Chart 3 shows that noninterest expense relative to assets for community banks fell after the Great Recession. Subsequently, overhead increased but remained below historic levels. In addition, the breakdown by asset size in the chart illustrates the economies of scale at play, whereby smaller institutions, like de novo banks early in their lifecycle, bear a disproportionate burden of overhead costs.

Source: Board of Governors of the Federal Reserve System.

Although economic conditions, interest rates, and regulatory burden play a role in de novo bank formation, they may not be the only factors that explain the recent decline in new charters. Longer-term trends such as advances in technology and demographic shifts may also be at work.

Banking makes intensive use of information technology (IT), and IT has undergone incredible transformations over the last 60 years. Computing power has increased at exponential rates, and the internet has changed how banks conduct business. New technologies make banking more efficient, lower costs, and allow banks to offer new and improved banking services. Technologies also extend banks' abilities to reach new customers.

However, keeping pace with technology requires purchase and implementation costs. Small banks may be at a disadvantage relative to larger banks that have better scale economies in adapting and implementing new technologies. Furthermore, because technology enables more efficient management of large organizations, large banks can compound their advantages. Online and mobile banking are examples of technologies that large banks can use more effectively than small banks to provide better service to customers over a wider geographic area. The proliferation of these technologies may therefore discourage the formation of new banks.

Technology also has another potential and possibly more important effect on new bank formations: Communication and computer technologies reduce the costs of covering distance to serve customers. Previously, banks used brick-and-mortar buildings with accompanying staff to reach customers, an expensive proposition that afforded existing banks a degree of competitive protection. Today, banks can serve customers virtually and cost-effectively over greater distances, reducing the incentive to charter a new bank in areas that previously might have supported one. Additionally, brand recognition and established infrastructure provide legacy banks with lower costs to customer acquisition and economies of scale that newly formed digital banks struggle to match in their early years.

Recent demographic changes may also have affected both demand for banking services and the attractiveness of opening a new bank. The populations of many rural and midwestern towns have gradually declined through migration to urban areas, particularly in areas reliant on agriculture and mining (Field 2025). Labor-saving mechanization and changes in demand for products struck a blow to many communities that relied on these and other industries for economic support. The demographic shift to urban areas has been especially significant for banking in midwestern states, which had many small banks due to their history of unit banking, where banks were restricted to a single office or limited geographic areas, such as a county.^_

Declining area populations decrease demand for banking services and the number of viable locations for new banks and make it difficult for banks to find new employees and bring on new management talent. Rural areas have thus become less attractive places to open a new bank. Reinforcing this negative effect on de novo bank formation is the competitive nature of urban banking markets, which can make it difficult to establish a successful de novo bank. Finally, owners of small rural banks may sell their bank and bank charter to investor groups; rather than open a new bank in an attractive location, investor groups can purchase a rural bank and open a branch in the attractive location.

Two other demographic changes occurred during the 1960-2023 period: The proportion of women in the workforce increased, and baby boomers (the largest age cohort) reached their prime working years. We measure these through the labor force participation rate (LFPR) because, unlike an age or gender variable, it also serves as a behavioral indicator of the balance between liquidity providers (passive depositors) and credit consumers (active earners) by measuring the segment of the population actively participating in production.

As baby boomers aged through their prime working years, the LFPR slowly declined starting around 2000. Life cycle theory suggests that people build wealth during their working years and use it to maintain their lifestyle later in life. They are more likely to be borrowers early in their careers. In later years, they have a higher demand for deposit products to hold accumulated wealth (Imam and Schmeider 2024). This tendency potentially changes the composition of banks' retail assets and liabilities. The necessity for de novo banks to build loans and deposits in tandem means that demographic shifts disrupting the balance between liquidity providers and credit consumers can create headwinds for de novo banks. However, because the optimal mix is nuanced, the effects of these demographic shifts on de novo bank formation are unclear a priori.

II. A Model of De Novo Bank Formation

To examine the factors that affect de novo bank formation, we analyze annual data from 1960 to 2023. We begin with 1960 because at this time, the banking industry appeared to be in a long-term equilibrium: Banking activities and industry structure were essentially the same as they had been since World War II.^_ Boundaries between financial services were largely intact, bank failures were rare, the regulatory environment was relatively stable, and technologies that would later gain prominence were in their infancy.^_ Industries-namely agriculture-were adopting new methods that would later determine the fate of many rural communities.^_

The dependent variable in our study is the annual number of de novo banks. This information is available from the FDIC for the years 1984 to 2023.^_ For earlier years, we estimate the number of de novo banks from FDIC data on new charters using a process similar to DeYoung (1999). We filter the charter data to remove new charters due to a change from a state to national charter (and vice versa) or the product of mergers. We do this by excluding newly chartered banks with total assets greater than $250 million (2023 dollars) at the end of their first year.^_

Our model can be summarized as follows:

Annual National De Novo Bank Formation = f(Economic Factors, Regulatory Factors, Technological Change, Demographic Change)

The dependent variable, the number of de novo banks, is count data consisting of integer values with a lower bound of zero. With this type of data, regression coefficients for ordinary least squares (OLS) regressions may be biased, and standard errors may not be reliable. For this reason, we use a Poisson regression, a generalized linear model that uses a logarithmic link function to model the relationship between predictors and the frequency of discrete events, ensuring that all predicted counts are nonnegative.

The explanatory variables include those that Adams and Gramlich (2014) find as important drivers of de novo bank activity-economic conditions, interest rates, and regulatory burden-as well as two additional factors, technological and demographic trends. We outline these factors and their relationship to our dependent variable below.

Economic conditions. We measure economic conditions using variables for economic growth (change in real gross domestic product) and short-term interest rates (federal funds rate). A strong, growing economy promotes loan growth, better asset quality, stronger deposit demand, and higher earnings, so investors are more likely to be interested in de novo bank formation when economic growth is strong.

We measure interest rates using the federal funds rate. The assets and liabilities of de novo banks, by definition, have been recently booked and will be more closely tied to recent short-term rates. New banks will therefore be asset sensitive-that is, earnings will be positively correlated with interest rates. Higher short-term interest rates will generally improve net interest margins and earnings, thereby making de novo banks a more attractive investment.

Regulatory environment. Rather than trying to measure, date, and incorporate specific banking regulatory changes, such as individual banking laws, we try to capture changes in the regulatory environment using three variables.^_ The first variable is the number of Federal Register pages, which is a commonly used proxy for cumulative regulatory burden. While the metric is easily understandable and has been published for decades, it includes regulations over the entire economy rather than those strictly pertaining to banking and does not provide insight into the content of published regulations (Gelfond 2020). Despite these limitations, an increase in Federal Register pages should generally be aligned with increasing regulation, which should in turn increase costs, decrease investor returns, and lessen interest in starting a new bank. In addition, in a tighter regulatory environment, chartering agencies may require more detailed application plans and take more time to review them. These effects delay opening while organizational costs continue to mount, making new bank chartering less attractive.

A second regulatory influence is the initial capital level of de novo banks. Capital levels are a critical factor for maintaining bank safety and soundness. Business plans for de novo bank applicants must be approved by regulatory agencies. As the regulatory environment becomes more stringent, de novo banks need to provide more initial capital to have their application approved. Higher initial capital levels may discourage de novo banks by requiring founders to raise more capital, which can be challenging for bank organizers. Also, higher capital levels reduce the return on equity for a given dollar level of earnings.

A third regulatory influence is a tax law change that is especially important for new and small banks. In 1997, the Small Business Job Protection Act expanded eligibility for Subchapter S tax treatment to small banks with private, concentrated (closely held) ownership (Federal Reserve Bank of Kansas City 2025).^_ For eligible banks, responsibility for the taxes on the bank's earnings can be passed on to shareholders. This allowance is attractive to investors in new banks as they avoid "double taxation" on earnings, and investors can manage their taxes more efficiently than the bank. We include a binary variable that has the value of 1 in the years 1997 to 2023. The advent of Subchapter S for small, closely held banks increases investor returns and makes new bank formations more attractive.

Trends in technology. We measure changes in technology using the level of private IT investment. The data measure investment in information-processing equipment and software in billions of inflation-adjusted dollars and are obtained from the U.S. Bureau of Economic Analysis.^_ The effect of technological advances on de novo bank formation is unclear: Although technology can improve operating efficiencies and may provide new business opportunities, it also requires escalating IT infrastructure costs and may increase digital competition.

Trends in demographics. Our analysis accounts for two demographic changes: the population shift from rural to urban areas and the population's changing age profile. We use the non-rural population percentage to account for geographic changes in population. We use the LFPR to capture the effects of changes in age distribution over time. The consequences of population migration and the aging baby boomer cohort may reduce returns to investors and reduce demand for new banks.

III. Explaining the Decline in De Novo Bank Formation

Table 1 shows that almost all variable coefficients are statistically significant with the expected sign and that our regression has strong explanatory power. The federal funds rate, GDP growth rate, labor force participation rate, and Subchapter S tax variables have a positive effect on de novo bank formation. Conversely, Federal Register pages, average opening capital, non-rural to total population percentage, and private IT investment have a negative effect. Alternative specifications are presented in the appendix.

As expected, our results show that economic growth and eligibility for Subchapter S tax treatment contribute to de novo formation. The coefficients for the federal funds rate and GDP growth variables are positive and statistically significant, supporting our expectation that de novo bank formation increases when the economy is growing and interest rates are not depressed. The Subchapter S binary variable is also positive and statistically significant: When tax liabilities can be passed through to investors and investors can avoid double taxation on earnings, more de novo banks are likely to be chartered.

Our results also show that greater regulation, demographic changes, and evolving technology have a negative effect on de novo formation. The Federal Register pages variable and the opening capital variable have coefficients that are negative and statistically significant, in line with expectations that fewer de novo banks are formed when the regulatory environment is more severe and when supervisory agencies require higher initial capital levels. The coefficient for the non-rural population variable is negative and significant in the regression, supporting our expectation that de novo bank formation weakened as the population shifted from rural to urban areas of the country. The labor force participation rate variable was positive and significant. De novo banks were more likely to be formed as more women entered the labor force and as the baby boomer generation aged into its prime working age years. Finally, the coefficient of the private IT investment variable is negative and significant. Technological change brings important advantages to banking, but it can slow de novo bank formation if incumbent banks can take advantage of the improvements more effectively than new banks. Further, technology can make it easier for incumbent banks to expand, increasing competitive pressure and reducing the attractiveness of de novo bank entry. Our results suggest that this is the case.

As noted earlier, the regression has strong explanatory power, indicated by a McFadden's R2 of 70.5 percent. As shown in Chart 4, the number of de novo banks predicted by the model tracks the actual number of de novo banks reasonably well. In general, the differences in the two series are smaller in the second half of the time series.

Sources: FDIC, FFIEC Call Reports (Board of Governors of the Federal Reserve System), and authors' calculations.

Relative importance of each factor

While Table 1 shows that our chosen factors have direct relevance to de novo bank formation, it does not show the relative importance of each factor, which depends on each variable's size and volatility.^_ To assess the relative importance of each variable to de novo bank formation, we compute the expected annual number of de novo banks using the mean value of each independent variable, except for Subchapter S status, which is set to one. Using these mean values, the model projects average annual de novo formation of 182 banks per year during the 1960-2023 period, higher than the actual average of about 138 de novo formations per year. While our model generally performs well, it failed to predict a couple of peaks in the 1960s and 1970s and predicted the recent decline in de novo formation to have occurred a bit earlier than what actualized bringing down the long-term predicted average.

Table 2 shows the change in the number of new banks when each variable is increased or decreased by one standard deviation. Shifts in economic, regulatory, technology, and demographic factors have collectively created a new normal in terms of de novo charter formation, with technological innovation appearing to account for the most significant reduction.

Note: Figures in the table are rounded to the nearest whole number.

Technology factors. Technological change has the largest effect on de novo bank formation. A one standard deviation increase in private IT investment reduces the number of de novo banks by 88. The exponential function is not symmetrical, in that the increases and decreases will have differing magnitudes. For example, a one standard deviation decrease in this variable has a much larger effect, raising the number of new banks by 172.

Demographic factors. Labor force participation has the second largest effect on de novo bank formation. A one standard deviation increase in the labor force participation rate is associated with an increase of 102 banks. In contrast, a one standard deviation decline in the non-rural population percentage (rising rural population) increases de novo banks by 31.

Regulatory factors. Reducing regulatory pressures increases the number of de novo banks. When Federal Register pages decline by one standard deviation, the number of banks increases by 42. Similarly, a one standard deviation drop in opening capital is associated with an increase of 82 banks. An increase of one standard deviation in Federal Register pages and opening capital reduces new de novo banks by 34 banks and 57 banks, respectively.

Economic factors. Improving economic conditions increases de novo bank formation. A one standard deviation increase in GDP growth would lead to 20 more banks. If the federal funds rate increases by one standard deviation, the model indicates that de novo bank formation would increase by 75 banks. In contrast, a one standard deviation decline in GDP growth and the federal funds rate would decrease de novo bank formation by 18 banks and 53 banks, respectively.

Current conditions

For policymakers interested in affecting de novo bank formation, recent values of our independent variables may be more useful, as values for many variables have diverged significantly from their long-term averages. Moreover, some variables (for example, private IT investment) are unlikely to ever revert to their long-term averages. To focus on the more recent behavior of the factors driving new bank formation, we replace the mean values of independent variables in Table 2 with their 2023 values. Using these recent values, the model predicts 16 new de novo banks in 2023, which is only slightly higher than the nine de novo banks actually chartered in 2023.

Table 3 shows the expected change in de novo bank formation for a one standard deviation change in the independent variables. Because the current level of de novo bank formation is lower, the absolute effects of changing policy variables are smaller. Technological change continues to have the greatest effects on de novo formation followed by opening capital needs and labor force participation.

Note: Figures in the table are rounded to the nearest whole number.

Economic factors. Based on 2023 values, a one standard deviation increase in the federal funds rate is associated with a seven-bank increase in the number of new banks. A one standard deviation increase in the GDP growth rate should increase de novo banks by two.

Regulatory factors. A one standard deviation decline in Federal Register pages, our general measure of regulatory pressure, would increase the expected number of de novo banks by four. A decrease of one standard deviation in initial opening capital would result in seven new banks. Reducing regulatory pressure (as measured by Federal Register pages) and initial opening capital is expected to lead to a 13 bank increase in de novo banks.^_ Furthermore, although not suitable for what-if analysis, the Subchapter S tax relief variable included within our model was quite influential, suggesting tax relief could be another avenue to stimulate de novo formation.

Despite model results indicating that economic conditions and regulatory relief can encourage de novo activity, results also indicate that they are unable to overcome the effects of long-term technological and demographic changes on de novo formation. However, because our model focuses on traditional banking charters, it is unclear what effect novel activities could have on future charter interest.

Conclusions

De novo bank formation is important in improving consumer banking access, promoting healthy competition in the industry, and offsetting the industry's consolidation. Given limited de novo activity in recent years, we develop a model to identify possible determinants of de novo formation over a long-term horizon. Our model shows that a combination of economic, regulatory, demographic, and technological factors can help explain the decline in de novo formation over time.

First, consistent with prior research, favorable economic factors are positively associated with increased de novo activity. Favorable economic conditions, including positive GDP growth, are associated with more loan demand and lower delinquencies, which would be favorable conditions for new banks. Further, a higher federal funds rate is likely to attract de novo bank investors since they increase the likelihood for higher net interest margins, higher profits, and greater investor returns.

Second, uncontrollable factors including technological change and demographic shifts play an important role in de novo activity. Technological change was the most influential factor within our model, although our results suggest it may have contributed both positively and negatively to de novo formation. In particular, technological changes appear to have increased both competitive pressures and bank business opportunities, although the negative effect on formation from competitive pressures outweighs any positive effects arising from added efficiencies. Demographic factors such as the declining labor force participation rate and population migration from rural to urban areas also appear to have weighed on de novo formation, potentially by reducing demand for new banking entities or changing the composition of products and services demanded by customers in way that dissuades investors from chartering new banks.

Third, although policy changes may increase de novo formation, even with large reductions in regulatory burden and initial capital requirements, de novo activity is unlikely to return to pre-recession levels. Based on our model, a one standard deviation reduction in both opening capital and Federal Register pages results in only about five to six more de novo banks opening each year. Even more discouraging, our model may have overestimated the effects of regulatory change on de novo activity since the banking agencies have attempted to tailor regulatory requirements more recently based on a bank's size and complexity. That being the case, regulatory burden relief would need to be targeted in ways that specifically benefit de novo banks rather than banking more broadly. Moreover, while relaxing policies, such as capital requirements, may help in encouraging de novo formation, the benefits should be weighed against costs of the increased failure risk early in the de novo lifecycle (Jones, Myers, and Wilkinson 2022). One shortcoming of our model is that we only consider factors that influence the formation of traditional banks. It is unclear if, and to what extent, unconventional business models may be a catalyst for new charter formation.

Appendix

Data Used in Analysis

Our study uses annual (year-end) data at the national level for the years 1960-2023. The dependent variable, the number of de novo banks, consists of the total number of newly formed commercial banks according to the FDIC's BankFind Suite's Annual Historical Bank Data. Because the FDIC's de novo bank count only goes back to 1984, we estimate prior years using a similar method to De Young (1999). We count the number of new charters in each year and remove any with over $250 million in total assets (2023 dollars) at the end of their first year to reduce the likelihood of including new charters that resulted from mergers or from existing bank charter changes. Chart A-1 shows the FDIC reported counts along with our estimates.

Sources: FDIC and FFIEC Call Reports (Board of Governors of the Federal Reserve System).

The independent variables consist of economic variables, demographic variables, regulatory variables, and a technology variable, as well as a binary indicator variable for Subchapter S tax eligibility given that preferential tax treatment improves shareholder returns for the small institutions. The economic variables consist of the federal funds rate and the annual rate of growth in U.S. real GDP. Demographic variables consist of the labor force participation rate and the non-rural to total population percentage. Regulatory variables consist of mean opening capital (inflation adjusted to 2023 dollars) and the total number of Federal Register Pages, a proxy for regulatory burden. Finally, private IT investment serves as proxy for technological innovation over the time horizon. Table A-1 provides full definitions for each variable in our model, Table A-2 provides full descriptive statistics for included variables, and Table A-3 presents the correlation matrix for the variables.

Alternative Measures of Technological Change

The paper's regression uses a broad measure of technological change. However, two technological changes, the development of the internet and digital banking, have been especially important to the banking industry. The internet promotes widespread information sharing. De novo banks can use the internet to provide information to potential customers about the new bank. De novo bank founders can use the internet to identify new market opportunities and learn how to implement new regulatory changes. Importantly, economies of scale may be less important for internet use, which may put small banks at less of a disadvantage relative to larger banks.

However, digital banking services-online and mobile-are likely to have economies of scale that advantage larger banks. Larger banks may be able to offer more digital services through platforms that are easier to use. In addition, digital banking services allow larger banks to cover a broader geographic area with fewer brick and mortar offices, which could reduce the availability of attractive areas for setting up a new bank. New and small banks may have access to some digital banking services through their core service providers. However, these services are unlikely to be as sophisticated or as user-friendly as digital banking services offered by larger banks.^_

We ran several regressions that included variables measuring internet access and digital banking. Variable descriptions and sources for these variables are in Table A-1. Data on internet adoption rates are from the World Bank via FRED. Values for the years before 1990 are imputed as zero. Digital banking adoption data are from Federal Reserve surveys and show the percentage of banking customers using online and/or mobile banking services. Values for the years before 1995 are zero. Due to multicollinearity issues, we did not run regressions that included all three technology variables. The results are shown in Table A-4.

Regression (1) is the standard regression from the main body of the paper. In Regression (2) the digital banking variable shows a significant negative effect on de novo bank formation. The negative effects of regulatory variables are smaller compared with the baseline regression. The effects of the Subchapter S variable are also smaller, and this is the case for all regressions with the alternative technology variables. The results are consistent with the expectation that digital banking disadvantages new and smaller banks relative to larger banks.

Interestingly, in Regression (3) the internet adoption variable has a significant positive effect on de novo bank formation. The internet allows new banks to advertise their services to new customers, and economies of scale may be less important with internet services. In this regression, regulatory variables have a larger negative effect on de novo banks.

Regressions (4) and (5) include the private IT investment variable with the digital banking adoption variable (4) and the internet adoption variable (5). In Regression (4), the digital banking variable has a significant negative coefficient, but the private IT investment variable is no longer significant. In addition, the labor force participation rate variable is no longer significant.

In Regression (5), the private IT investment variable is negative and significant, and the internet adoption variable is positive and significant. Other coefficients are similar to those in Regression (3).