Financial Development: Structure and Dynamics

What has shaped the process of financial development (FD), and how regular has it been? Has this process followed a single or multiple paths? What were the sequences and shapes of the paths followed by different financial services and activities (as measured by FD indicators) as economies developed? How did policy—whether aimed at strengthening the enabling environment for financial contracting or ensuring the sustainability of FD—affect these paths?

Remarkably, the literature (particularly the empirical literature) that attempts to explain how economic development and financial sector policies jointly affect FD is still nascent.² The proposition that financial structure is shaped by the efforts of market participants to circumvent and reduce the frictions that hinder financial contracting is of course a familiar one.³ However, few papers have attempted to analyze these frictions in terms of the way they interact and what this interaction may imply for the dynamics of FD.⁴ At the same time, FD has typically been understood as a relatively smooth and predictable march from “relationship-based finance” to “arm's-length finance” involving a systematic process of market completion driven by a gradual reduction of agency frictions.⁵ However, the global financial crisis showed that FD has a “dark side” associated with volatile boom-bust cycles. Thus, what may appear as progress in market completion can, in fact, exacerbate market failures, thereby undermining the sustainability of FD.

This paper begins by exploring and explaining the patterns of FD based on a simple typology of the frictions that hinder financial contracting. Following work by de la Torre and Ize (2010, 2011), the paper separates the frictions into agency frictions, which restrict the scope of bilateral contracting, and collective frictions, which restrict the scope of multilateral interaction and participation.⁶ The structure of the financial system and its evolution over time (hence FD) reflect efforts to find the path of least resistance around these frictions. This hypothesis leads to broad predictions regarding the sequencing of financial activities, the shape of their development paths, the importance of scale effects, and the role of public policy. The hypothesis also suggests that the same easing of frictions that underlies the “bright side” of FD (innovation, market completion, improved allocation of funds and risks) can breed the tensions and fault lines of the “dark side” (systemic fragility, excessive fluctuations, crises).

To explore such issues, we use a battery of 16 financial indicators that relate to different dimensions (depth-size or liquidity-efficiency) and channels (markets or institutions) of FD. To create a uniform set of dynamic benchmarks that facilitates comparisons across countries, we use a comprehensive set of country-level controls consisting of variables that reflect the stage of economic development (initial and current per capita income) as well as structural variables that are arguably exogenous to financial policy, including population size and density. We find that FD, as measured by the benchmarks, followed regular dynamics. However, development paths depended on the initial level of per capita income. Thus, the lower-income countries did not generally retrace the steps followed in the past by the current higher-income countries. This finding suggests that across-the-board innovations (that lift all boats at the same time) and path dependencies (that reflect dynamic interactions between financial and economic development) are both relevant for a country's FD trajectory. The fact that the data we use cover only a small slice of the world's FD history (35 years at most) compounds the importance of such initial income effects.

Reflecting the differential impact of frictions on different financial activities, paths differed widely across FD indicators. Indeed, the sequencing of financial services broadly conformed to what one would expect based on the gradual smoothing of the frictions along the paths of least resistance. For example, reflecting differences in the difficulty of coping with bilateral agency frictions (information or contract enforcement costs) across financial services, credit to governments developed before credit to private participants, bank deposits preceded bank credit, and the development of capital markets and associated institutions, such as mutual funds and other forms of asset managers, followed the development of bank credit. At the same time, reflecting collective frictions and network effects, external funding of the government preceded domestic funding, retail funding preceded wholesale funding, casualty insurance preceded life insurance, the development of capital markets lagged behind and was highly convex (a manifestation of large scale and network effects), and interconnectedness increased significantly as financial systems matured.

We further explore the dynamics of FD based on an analysis of how and why countries deviated from their benchmark paths. For this exploration, we use as additional controls a number of policy-related variables proxied by various enabling-environment indicators. We also include a financial crash variable to capture the dark side of FD. We find that enabling-environment factors, such as enforcement costs, creditor or property rights, and credit information, played an important role in promoting FD. However, we find that contractual frictions (rather than informational frictions) explained the bulk of policy-induced developmental differences across countries. Although informational frictions could be mostly overcome through technological innovations that could be more easily imported, contractual frictions mostly reflected weaknesses in local (nonimportable) institutions that were more difficult to resolve. Regarding booms and busts, we find that financial crashes accounted for large and lasting lags in FD that cut across a surprisingly large number of indicators and dimensions of FD. The lasting impact of financial crashes clearly puts a premium on central bankers' and supervisors' capacity to anticipate and control unsustainable booms through monetary or macroprudential policies.

The rest of the paper is organized as follows. Section 2 briefly presents the conceptual framework for financial frictions that underpins FD (the bright side of finance) and links it with financial instability (the dark side of finance). Section 3 describes the data and procedures that are used to estimate the FD paths. Section 4 presents the dynamics of FD as given by the benchmark paths. Section 5 explores the role of policy in explaining deviations from the benchmarks. Section 6 concludes.

I. Financial Frictions, Financial Development, and Financial Stability

The financial services industry emerges to help market participants find ways to reduce or circumvent the two classes of frictions that hinder financial contracting: bilateral agency frictions and multilateral collective frictions. Each class can be subdivided into two categories: informational frictions (which relate to agents' limited and often asymmetric capacity to understand information and to the costs of obtaining information) and relational frictions (which hinder agents' capacity to agree, act upon, and enforce bilaterally or collectively beneficial financial contracts). This simple dichotomy underpins four paradigms, of which two (asymmetric information and costly enforcement) are associated with agency frictions and two (collective action and collective cognition) are associated with collective frictions.⁷

Asymmetric information frictions hinder FD because they lead to a misalignment of incentives between the principal and the agent. This misalignment, in turn, can trigger the commonly known market failures of adverse selection, risk shifting, shirking, and false reporting. Thus, information asymmetry frictions limit financial contracts to those contracts in which the agent has sufficient resources of its own at risk (“skin in the game”) and/or where the principal can adequately screen and monitor the agent. Enforcement frictions also lead to a misalignment of incentives between the principal and the agent and, in this case, because of imperfect pledgeability, to a situation in which the agent is unable to credibly commit to honor the contract. Imperfect pledgeability thus restricts financial contracts to those contracts that can be effectively collateralized.

Collective frictions hinder FD by constraining collective (rather than bilateral) participation. Much of the gains from financial activity relate to the reduction in transaction costs and the increase in liquidity and risk diversification benefits that result from multilateral arrangements in which many players participate. Such arrangements can either take place in markets, where transactions can be conducted around a trading platform, or through financial institutions that offer services whose benefits are pooled across a large number of customers. The higher the number of participants is, the higher the benefits of participation will be. However, although participation creates positive externalities for society at large, it may be hindered by coordination failures. For example, because of first-mover disadvantages, an investor may abstain from buying a long-term security for which there is no secondary market and, hence, no liquidity. In this case, there is a multilateral (individual versus social) misalignment of incentives. Participation may also be hindered by collective (even symmetric) cognition frictions; one does not participate in an activity that is not well understood.

Therefore, market participants who wish to engage in financial contracting must find the path of least resistance around these frictions and the associated market failures. Once a decision is made to participate, private responses to coping with frictions can be divided into two subsets: responses aimed at lessening the frictions themselves (acquiring information, using collateral, delegating) and responses aimed at lessening the exposure to these frictions (diversifying and pooling risk, buying insurance and hedges, staying liquid). In turn, the state facilitates these private responses through a set of progressively more intrusive public interventions: (i) the provision of a basic contractual and informational infrastructure that facilitates contracting; (ii) coordination arrangements that facilitate participation (catalytic involvement to promote market development, systemic lending of last resort, government guarantees); (iii) the regulation and taxation needed to internalize externalities or protect consumers; and (iv) the direct provision by the state of financial services.

Financial structure is a snapshot, at a given point in time (hence, for a given technological and state of market development), of the composition of financial services aimed at coping with financial frictions. FD reflects the evolution of financial structure over time. At lower stages of FD, financial institutions resolve agency frictions by relying on nontradable and immovable collateral and relationship-based transactions. However, as the informational and contractual environment improves, private information becomes public, other types of collateral become available and tradable, and monitoring can increasingly rely on third parties, statistical methods, and accounting and disclosure standards.⁸ The gradual easing of agency frictions thus helps boost participation, which unleashes positive network and scale externalities (e.g., liquidity, learning spillovers, efficiency) and sets in motion a virtuous development cycle. In turn, rising participation gradually increases the degree to which financial institutions and capital markets complement each other. The entire process is accelerated by financial innovation.⁹

The gradual easing of finance frictions provides a few predictable regularities that can be empirically verified. The first regularity concerns the sequence in which various financial activities are likely to develop, the second regularity concerns the shape of the paths they are likely to follow once they begin, and the third concerns the volatility of FD, reflecting the seeds of financial instability that germinate while FD takes place and that may eventually develop into a financial crisis.

The order of development of financial activities should reflect the intensity of the frictions to which they are exposed. The activities that are the least constrained should develop first. However, those financial services that are strongly inhibited by collective frictions should develop only after a critical mass of participation has been reached that is sufficient to trigger the positive network externalities needed to sustain their development. We would thus expect participation-intensive financial activities to be those that have the most rapid development paths (the most “buoyant” or convex) once they have passed this initial threshold. Because activities that exhibit the highest returns to scale are expected to be those exposed to the highest collective frictions, scale effects should correlate with the order in which financial activities develop and, subsequently, the buoyancy of their development.¹⁰

However, the same frictions that underlie FD (the bright side of finance) are also likely to be at the root of its dark side (financial instability). We define the “dark side” as the type of financial malady in which the actual success in the process of smoothing frictions can endogenously lead to systemic instability.¹¹ This type of adverse dynamic may ensue if the easing of frictions boosts participation in such a way that it unleashes new and possibly more severe forms of frictions. FD may thus advance along a fragile or self-destructive trajectory that may lead to large and lasting reversals in developmental indices.

The dark side may arise at the interface between agency and collective action frictions. The smoothing of agency frictions facilitates a switch from private to public information, which, in turn, promotes participation but leads to a socially insufficient supply of screening and monitoring. Investors free ride on the increased availability of public information and choose to remain short and rely on market liquidity and public information to exit at the first sign of possible trouble.¹² The boost in participation can also promote risk origination without sufficient “skin in the game.” As observed in the global crisis, this situation can give rise to a complex and opaque chain of transactions, which is ultimately unstable.¹³

This situation may also arise from dynamics associated with the intensification of collective action problems. The positive externalities of increased market participation in good times may turn into crippling negative externalities in bad times. Thus, market withdrawal in times of stress may be individually optimal but socially harmful because it can trigger self-fulfilling liquidity collapses.¹⁴ Financial institutions may become too large (from a social point of view) because participants may not internalize the negative externalities (e.g., domino effects and contagion) associated with the failure of systemically important financial institutions and the too-big-to-fail or too-interconnected-to-fail syndromes that are associated with these failures.

The dark side may also be associated with the swelling of collective cognition problems. The successful easing of frictions (including through innovation) may lead to problems of collective cognition that, in a world of collective uncertainty, may result in wide mood swings. The bonanza associated with enhanced participation feeds a collective mood of optimism that unleashes bouts of exuberance. The euphoric mood initially accentuates the upswing; however, once unexpected realities emerge, euphoria can easily turn into despair, worsening the collapse.¹⁵

II. Data and Estimation Procedures

In this section, we conduct a simple empirical analysis of FD indicators. Our aim is to illustrate the above conceptual framework by showing that the sequence of development of various financial activities and the shape of the path that they follow as they develop is consistent with (and thus validates, at least broadly) the broad footprints described above.

We measure domestic FD based on 14 depth-size indicators that include the following:¹⁶ We complement these indicators with two indicators of efficiency-liquidity for which there is sufficient cross-country data: To make the data as comparable as possible across countries, we create FD benchmark paths for each FD indicator by controlling for structural factors that can be considered policy exogenous (at least in the short term), including population (size and density) and three other country-specific characteristics that, for a given level of economic development, have a sizable impact (positive or negative) on FD (fuel exporter, offshore financial center, transition country).¹⁷ We also control for economic development, as proxied by per capita GDP. Although the latter is clearly not exogenous to FD, this fact is not a significant concern for what we want to achieve here.¹⁸

• Key components of commercial banks' operations (retail and wholesale funding, credit to the private sector, credit to government, and claims on other domestic financial institutions)

• Insurance company premia (life and casualty)

• Mutual fund and pension fund assets

• Public and private debt securities (domestic and international capitalization)

• Equity (total stock market capitalization)

• Banks' net interest margin

• Equity market turnover

Reflecting that FD can be expected to be generally path dependent, we also include as controls the initial level of GDP per capita (as measured for the earliest date for which financial statistics are available) and its interaction with current GDP per capita. By introducing a level effect, including initial GDP per capita allows countries at different initial levels of economic development to follow their own dynamic FD path. The interaction term, for its part, allows the slope of this dynamic FD path to vary with the level of initial economic development.

As shown in the supplemental appendix S.II, this specification is key to properly representing the dynamics of FD, whether they reflect the impact of financial innovations or the interactions between economic and FD. Financial innovations are typically transferable across countries and thus tend to be introduced across the board (in low- as well as high-income countries)—that is, financial innovations can cause a boost in financial activity that “lifts all boats at the same time,” with dynamic paths suddenly surging above the cross-section (benchmark) line at all per capita income levels in a parallel fashion.¹⁹ In addition to allowing for such parallel surges, including the initial income level in the estimating regressions also helps to capture path dependence. Initial conditions matter because today's FD depends on today's output, which in turn depends on yesterday's FD. Path dependence thus implies that better initial institutions can become self-reinforcing.²⁰

The α₁ coefficient measures the elasticity of FD with respect to the country's initial per capita income. The higher this elasticity is, the more dependent the development of the financial indicator (or financial activity) is on the country's initial level of economic development (per capita income). Thus, this coefficient can be viewed as a proxy for sequencing: The activities that are more income elastic develop “later” than those activities that are less income elastic. Similarly, the α₂+α₃y₀ coefficient measures the elasticity of FD with respect to the country's per capita income growth.²² The higher this elasticity is, the faster the activity develops with the country's economic development. Thus, growth elasticity characterizes the “buoyancy” of the financial-activity path. Finally, the α₄ coefficient measures returns to scale (elasticity relative to population size). Activities with higher returns to scale prosper in larger countries.

We run this regression in two stages. First, to fully capture the longer-run dynamics of FD and thus obtain a more meaningful benchmark, we estimate the regression over the whole panel but without policy and institutional variables. The results of this first stage are discussed in section 3. In the second stage, we reestimate the equation with policy-related institutional variables to capture their specific impact on FD. This second stage, whose results are examined in section 4, is conducted over a narrower sample because of limitations in the data on enabling-environment indicators.

III. FD Growth Patterns

Table 1 reports the results of the first-stage regressions over the full sample. Note first the impact of the structural controls. Reflecting returns to scale and network effects, FD tended to lag in countries with smaller or more dispersed populations. In contrast, reflecting uneven growth across sectors, FD in oil-exporting economies lagged behind FD in other economies at similar levels of income. Similarly, FD lagged behind in transition economies, in this case reflecting a pretransition economic system that did not favor FD. Inversely, and for obvious reasons, FD in offshore centers generally led FD in other countries.

Table 2 shows the variance decomposition of these estimates, expressed by groups of variables, in terms of simple averages (of absolute values) over all indicators as well as ranges (maxima and minima). The initial level of economic development (per capita income) accounts for the bulk (nearly half) of the explained cross-country variations in FD, which, in turn, account for approximately half of the total variance. Thus, much of the cross-country FD pattern is simply explained by initial conditions, which, in an important sense, summarize the entire history of the economic and FD of a given country prior to the first observation in the sample period. However, countries' economic growth thereafter and over the sample period explains a substantial portion (approximately 30 percent) of the explained portion of FD paths, with population size and other structural controls accounting for the rest.

Figure 1 orders financial indicators in accordance with their elasticities with respect to initial per capita income and contrasts these elasticities with the growth and returns-to-scale elasticities. Panel C of figure 1 shows the matrix of cross-correlations across indicators. As a whole, figure 1 broadly validates one of the three main predictions of section 2: Sequencing, buoyancy, and returns to scale are all substantially correlated. The correlation between sequencing and buoyancy is particularly high; the later an activity develops, the faster it develops once it takes off. This result is expected once strong collective (participation) frictions are overcome. The positive correlation between sequencing and returns to scale also confirms that returns to scale relate to collective frictions. Larger returns to scale require higher critical mass for an activity to develop, and thus, it is more difficult for this activity to develop in the presence of substantial collective frictions.

To further test the hypothesis that sequencing, buoyancy, and returns to scale are interrelated, we group the indicators according to their initial income, income growth, and size elasticities (table 3). Regarding income, the high correlation between initial income and income growth elasticities simplifies this task because all FD indicators can be placed in three groups (column headings of table 3): early developers with low buoyancy ⁠, middle developers with average buoyancy ⁠, and late developers with high buoyancy ⁠. Financial indicators in the first group developed early but declined in importance as income grew. In contrast, the third group appeared late but grew very quickly. The middle group was somewhere in between.

Regarding size, we divide activities into those with negative returns to scale (α₄ < 0), low returns to scale ⁠, and high returns to scale (α₄ > .1) (row headings of table 3). Most financial activities (12 out of 16) cluster in the diagonal cells of the table, fully matching the predictions of section 3. That is, the later the initial level of economic development at which a financial activity developed is, the faster it developed as income grew and as it was increasingly subject to increasing returns to scale. The financial services above the diagonal or below the diagonal do not fully match the predictions of section 3. Above the diagonal are financial services that, given their low returns to scale, developed later than expected (casualty insurance) or more buoyantly than expected (life insurance).²³ Below the diagonal is one financial indicator (stock market capitalization) that, considering its high returns to scale, developed earlier than expected. However, these three FD indicators that deviate from predictions are only mildly deviant (there are no activities in the top-right cell or in the bottom-left cell of table 3), and the reasons underlying their deviancy can be easily identified, as discussed below.

To better comprehend the process of FD, we examine each financial activity by grouping them into four categories (government borrowing, banking services, capital markets, and institutional investors) that are broadly ordered according to the sequencing of their development. To illustrate the dynamics, we chart the predicted growth trajectories followed by some of the indicators for individual countries as their income grew over their initial income level (figures 2 and 3). We plot the initial and current income on the horizontal axis and the value of the indicator on the vertical axis. The continuous lines correspond to the initial cross-country benchmarks (i.e., the projected values of the indicators when initial income varies across countries while current income remains equal to initial income). The dotted lines correspond to the expected country-specific FD trajectories, given its initial income (i.e., the values of an FD indicator projected for a country as its per capita income grows above its initial income).²⁴

Government Borrowing

Somewhat surprisingly, public-sector borrowing was the financial activity that developed the earliest. Furthermore, public-sector borrowing developed in international markets before it did at home (figure 1). These features are easy to explain based on frictions. Public-sector borrowing developed early because sovereigns are well known; hence, agency frictions were comparatively easier to overcome.²⁵ However, public-sector borrowing developed abroad first because well-established international markets, by definition, have overcome the relevant collective frictions. Hence, it is not surprising that government borrowing initially took place abroad and in foreign currency.²⁶ As frictions eroded and countries' FD deepened, governments were able to substitute external borrowing with domestic borrowing. However, reflecting collective-participation frictions, governments typically started by borrowing from local banks instead of markets (figure 2). Thus, domestic public debt developed later than domestic bank credit to government or foreign public debt.

Banking Services

At the initial stages of FD, the lack of public information, a sound contractual framework, and tradable collateral restricts lending to mostly relationship lending based on private information rather than at arm's length (based on public information). Hence, banks would be expected to develop relatively early, ahead of capital markets or asset managers such as mutual funds.²⁷ The fact that banks can also solve collective frictions efficiently through liquidity pooling and the provision of basic payment services further enhances their early attractiveness. Indeed, this early attractiveness is exactly what one observes. Banking services were the next financial activity to develop (figure 1). However, because attracting deposits is easier than overcoming agency frictions that limit lending, bank deposit taking should precede bank lending. Moreover, banks initially lent to governments rather than to private agents. Indeed, both features are corroborated by figure 1. At the same time, bank retail funding (deposits) took off before wholesale funding. As frictions eased, retail investors were increasingly able to shift into higher yielding, market-based instruments or to have their funds managed by asset managers or institutional investors rather than banks.²⁸ Reflecting participation frictions, bank claims on other financial institutions developed later but grew much faster once they developed (figure 3).²⁹ The gradual overcoming of collective frictions allowed banks to increasingly participate in an interinstitutional market that became denser as the number of players rose and the latter became more interconnected. Finally, note in figure 1 that improvements in bank efficiency (a reduction in net interest margins) began occurring very early, presumably as soon as banks started to operate.

Capital Markets

In view of the stronger agency frictions associated with arm's-length transactions, capital markets should only develop once public information has improved sufficiently. This hypothesis is largely consistent with the evidence. In view of high returns to scale, stock markets developed after banking (figure 1). Nonetheless, they developed ahead of prediction, which may be explained by the fact that equity issuance is an essential component of corporate finance (and governance) for larger firms. However, reflecting size thresholds and collective frictions that limit participation, the development of primary markets (market capitalization) should precede the development of secondary markets (market turnover), where trading liquidity is essential. As corroborated by figure 1, the secondary market developed after the primary market. At the same time, equity markets are likely to develop ahead of corporate bond markets because their unlimited upsides can better compensate for the downsides associated with agency frictions. Again, figure 1 shows that this sequencing held: Private debt securities followed equity and developed abroad before developing at home.

Institutional Investors

The development of asset managers is likely to be constrained by the development of capital markets. In particular, the growth of mutual and pension funds should reflect the fact that marketable, liquid assets—which the funds need to invest in—appear relatively late. Furthermore, pension funds developed earlier than mutual funds (figure 1) because the collective frictions that needed to be overcome for their development were largely solved by an act of government (e.g., the creation of privately administered, fully funded, individual retirement savings accounts, often of a mandatory nature). In contrast, the frictions hampering the development of mutual funds (large returns to scale) had to be resolved by market forces. In both cases, however, their growth was extremely buoyant, reflecting the fact that marketable, liquid assets in which these funds invested grew very rapidly once they began to develop. Regarding insurance, life insurance companies are likely to develop relatively late because, in addition to being exposed to collective frictions (they also need to invest in market-based assets), they are exposed to larger agency frictions that reflect their longer investment horizons. This fact may explain why life insurance developed later than predicted. In contrast, casualty insurance is likely to develop earlier, both because it is partly influenced by policy (as in the case of mandatory insurance for motor vehicles) and because the risks it faces are normally distributed and can be more easily addressed through risk pooling. Furthermore, their negative returns to scale, which make this activity somewhat of an outlier, can be explained by the predominance of foreign trade insurance in smaller open economies.³⁰

IV. The Role of Policy

In this section, we briefly explore the role of policy and institutions in explaining the deviations from benchmark paths. We use three variables (creditor rights, investor protection, and enforcement costs) that reflect the quality of the contractual environment and one variable (creditor information) that reflects the quality of the informational environment.³¹ To measure the quality of macroprudential management, we construct a credit crash dummy that measures severe annual drops in the ratio of private credit to GDP.³² Table 4 presents the results of the regressions similar to table 1 but with the policy variables included, estimated over the more limited period for which the data are available. The variance decomposition for these new estimates, which is similar to the decomposition shown for the regressions without policy variables, appears in table 2.

Together, the enabling-environment indicators account for a significant (albeit limited) fraction of the total explained variance of the financial indicators, which rises from 48 to 60 percent when the policy variables are incorporated.³³ As expected, better creditor rights promoted banking activity (not only private credit but also, and apparently even more, bank claims on other financial institutions) as well as capital markets (stock market capitalization and, even more strongly, private bond market capitalization) and life insurance. Similarly, lower enforcement costs facilitated bank funding (both retail and wholesale) and bank private lending while contributing to lowering intermediation margins. Unsurprisingly, the quality of investor protection had a particularly large impact on stock market activity (both in the primary and secondary markets). Furthermore, as expected, the quality of the informational environment was important for bank private lending and, even more so, for private (corporate) debt capitalization (particularly at home).³⁴

It is noteworthy, however, that the contractual variables (particularly creditor rights) accounted for the lion's share of the cross-country differences, with information accounting for only a relatively small fraction.³⁵ This result likely reflects the fact that improvements in the informational environment, which are more technology dependent, could be more simply introduced and imported from abroad.³⁶ In contrast, improvements in the contractual environment required a strengthening of local institutions that was harder to deliver. This interpretation is bolstered by the results of a regression of creditor rights and credit information against country income (incorporated through both a linear term and a quadratic term). The creditor rights index is convex, whereas the credit information index is concave.³⁷ This result suggests that improvements in the informational environment are achieved early on, whereas improvements in creditor rights take much longer to materialize. Thus, one should observe fewer cross-country differences regarding the quality of information compared to the quality of contracts.

Remarkably, the credit crash dummy accounts for a sizable share of total explained variance, particularly in the case of private credit (where it accounts for nearly 30 percent) but also for other banking indicators, such as bank funding and bank margins, and even for financial activity outside banking. For example, the credit crash dummy had a sizable negative impact on stock trading. It is also remarkable that banking systems took such a long time to recover from such impacts. This finding can be inferred from the fact that the credit cash dummy, which is estimated over a 30-year time span, remains highly significant when regressed against the pure cross-section of last period indicators rather than the full panel.³⁸ This result is a noteworthy reminder that FD (the bright side) and financial instability (the dark side) strongly interact in lasting, complex ways.

The interaction between bright and dark sides is illustrated further in figure 4, which plots the predicted and actual dynamic paths of private bank credit for all countries in the database, separated into three groups according to their initial income levels. Even at such an aggregate level of analysis, a cursory inspection suggests that, in most cases, busts occurred following booms in which the actual development paths went over their predicted values. Thus, financial unsustainability (the dark side of FD) appears to be related to abnormal deviations from the regular development paths inferred from a cross-country benchmarking analysis (the bright side of FD). Of course, this hypothesis must be validated by a detailed statistical analysis at the country level, a topic left for further research.

V. Final Thoughts

The broad cross-country benchmarking approach we have developed here can help to measure the quality of policies at any point in time and to identify the forces and frictions shaping FD. This approach began from the premise that both the evolution of financial structure (FD) and its sustainability (financial stability) are shaped by four fundamental types of frictions and their associated market failures. Two such frictions restrict agents' capacity to establish and enforce bilateral contracts (agency frictions), and the other two frictions restrict agents' capacity to participate and coordinate their financial activities in ways that are collectively desirable (collective frictions). The predictions derived from this approach regarding the order of development, returns to scale, and the shape of the developmental paths of various financial activities were broadly satisfied. In particular, where strong participation frictions were responsible for hindering the growth of financial activities, development only began after some threshold. However, once this threshold was passed, it was followed by buoyant dynamics because rising participation and interconnectedness generated positive externalities that promoted further participation and interconnectedness.

This paper has also argued that the same frictions that feed the development forces of the bright side feed the forces of instability from the dark side, making them interact in complex and unexpected ways and leading to booms followed by busts.³⁹ Indeed, the paper has shown that FD was substantially volatile and that countries paid a heavy and lasting price for financial collapses. Avoiding such collapses requires the early identification of the buildup of systemic stress. By inferring a predicted development path from a broad universe of observations, our benchmarking approach may help to discriminate between sustainable and unsustainable trajectories. A key topic for further research is to investigate whether deviations from benchmarks can help to predict crashes on a country-by-country basis.

An important caveat applies, however. Although our analysis suggests the presence of developmental regularities, it does not ensure predictability, particularly as one goes farther into the future. On the one hand, financial innovations may introduce inflexion points that are undetectable from the existing historical data. If so, apparent short-term upward deviations from the benchmark paths could, in fact, become sustainable because the benchmark paths themselves bend upward. On the other hand, it is possible for the benchmark paths to bend downward, at least in the longer term. For example, some of the indicators that have exhibited rapidly growing paths may eventually slow down, in accordance with logistic (S-shaped) trajectories.

References

Notes