What’s in a name? Dynasties, selection, and talent allocation among classical composers

Abstract

A central question in economics is how talent is allocated across elite professions and whether access is shaped by family ties or by open, standardized systems of selection. We study these questions in Western classical composition, a setting in which family background historically influenced entry through both inherited skill and inherited privilege, and where rich biographical records make it possible to trace selection over multiple centuries. Using data on more than 16,000 composers from 450 CE to the present, we identify dynastic ties from Grove Music Online and measure prominence by biographical entry length. Dynastic composers are 15–21 percent less prominent than non-dynasts, conditional on country and birth cohort, and this discount is driven by descendants: founders perform about as well as non-dynasts, while descendants perform significantly worse. Archival manuscript evidence from the Répertoire International des Sources Musicales yields similar results, suggesting that the pattern is not an artifact of editorial selection. In the twentieth century, however, the dynasty discount reverses, consistent with the transition from informal, family-based entry to more standardized selection through conservatory training. Consistent with this interpretation, dynasts are less likely to have formal training mentioned in their biographies, more likely to be described as advancing through family connections, and face a smaller dynasty discount in regions and periods with conservatories. Taken together, the findings suggest that expanded access to standardized training can weaken inherited advantage, reshape elite formation, and improve talent allocation, with broader implications for long-run innovation and economic performance.

1 Introduction

Family dynasties are common in high-status professions such as politics, medicine, and law (e.g., Dal Bó et al., 2009; Lentz and Laband, 1989; Raitano and Vona, 2021). Children enter these professions at rates far above population baselines, but what does this persistence signify? Dynasties may reflect the productive transmission of tacit knowledge and the advantages of early exposure (i.e., Becker and Tomes, 1979, 1986), or they may indicate selection failures in which access is shaped by family ties rather than ability (e.g., Aina and Nicoletti, 2018; Mocetti, 2016). Distinguishing these mechanisms is difficult because talent is unobserved, outcomes are noisy, and family ties affect entry thresholds, causing standard parent–child comparisons to conflate inherited ability with preferential access (see de la Croix and Goñi, 2024). The issue is further complicated by institutional change. In some eras, access to elite occupations depended heavily on patronage or hereditary privilege, while in others it was shaped increasingly by formal schooling and credentialing. These shifts influenced who entered the upper tiers of creative, political, and professional life, with potentially important consequences for talent allocation and aggregate innovation (e.g., Hsieh et al., 2019).

We study these questions in Western classical composition, a historical setting that offers unusual leverage for examining how family ties, training institutions, and selection interact over the long run. Composition offers a particularly useful lens for studying intergenerational persistence because, unlike more structured occupations, creative careers were historically less regulated and success was harder to verify. Whereas professions such as law or medicine are shaped by formal credentials, hierarchical progression, and more standardized performance metrics, musical careers depended more heavily on informal training, patronage, and subjective evaluation. As a result, both productive skill transmission and preferential access were especially salient: musical knowledge and technique could be passed informally within families, while family connections could ease access to commissions and employment opportunities. This makes composition a valuable setting for examining how inherited advantage shapes selection into elite occupations, a central concern in the literature on dynastic persistence and talent allocation (e.g., Galor and Tsiddon, 1997; Hassler and Mora, 2000; Hassler et al., 2007; de la Croix and Goñi, 2024).¹

Another advantage of this setting is that composers’ lives are unusually well documented. Biographical sources record extensive kin links—parents, siblings, cousins, and grandchildren—allowing us to identify dynasty founders and trace descendants across multiple generations.² These links extend far beyond the parent-child relationships typically observed in the literature, providing unusual leverage for studying how dynastic advantage operates within extended families. Combined with multiple, independent measures of success, they allow us to study dynasties at scale and over long historical horizons in a way that is rarely feasible in other creative or scientific settings.

Changes in musical training also offer a rare opportunity to observe how institutional reform reshapes the balance between inherited advantage and individual ability. For centuries, instruction was informal and family-based; beginning in the nineteenth century, conservatories introduced standardized curricula, broader access, and more competitive admissions. This transition parallels developments in other skilled occupations, such as law, medicine, and engineering, that moved from lineage-based apprenticeship toward formal training. It also aligns with theories of intergenerational mobility (e.g., Galor and Tsiddon, 1997; Hassler and Mora, 2000) which emphasize how new modes of knowledge transmission alter the relative importance of inherited background and individual ability. The rise of conservatories therefore allows us to examine how standardized training changed the role of family background in shaping entry and success.

To organize the empirical analysis, we develop a simple conceptual framework with two channels through which family background can influence creative success. On one hand, inherited human capital—the transmission of tacit knowledge and skills within families—raises productivity conditional on entry. On the other, inherited privilege—access to patronage, reputation, or professional networks—lowers entry costs, allowing less able individuals to enter the profession. These channels have opposite implications for average performance: the first strengthens selection on talent, while the second weakens it. The framework therefore yields two empirical predictions. If inherited human capital dominates, dynastic founders should outperform outsiders, with descendants gradually regressing toward the mean as inherited skills decay. If inherited privilege dominates, descendants should underperform, reflecting weaker selection on ability. The balance between these channels depends on the institutional environment. When training and access are informal, privilege is likely to matter more. As standardized training systems such as conservatories emerge, inherited advantages erode: formal instruction narrows gaps in music-specific human capital across families, while more open recruitment compresses lineage-based access advantages. This mechanism parallels models of dynastic persistence in changing institutional or technological environments, where the relative importance of inherited background and individual ability shifts as systems of knowledge transmission evolve (e.g., Galor and Tsiddon, 1997; Hassler and Mora, 2000; Hassler et al., 2007).

Using a newly constructed dataset of more than 16,000 Western classical composers from the fifth century to the present, we document systematic differences in prominence between dynastic and non-dynastic composers. Our main finding is that dynastic composers are, on average, less prominent than their non-dynastic peers, and that this pattern is robust when comparing composers born in the same countries and historical periods, exposed to similar local environments, and holding comparable occupational profiles. We measure prominence using the length of each composer’s biographical entry in Grove Music Online, a leading encyclopedia of music and musicians, where entry length reflects expert assessments of historical significance and correlates with output, influence, and recognition (Borowiecki et al., 2023, 2025; Borowiecki and Law, 2026). Using documented family relationships in Grove, we distinguish between dynasty founders, who have no known older composer-relative, and descendants, defined as later-born members of the lineage.³ This distinction is central to the analysis because it allows us to separate the performance of those who establish dynasties from that of later family members.

Looking more closely, we find that the dynastic discount is driven entirely by descendants. Founders, who establish new dynasties, perform about as well as outsiders, whereas descendants perform significantly worse. Within our framework, this pattern is difficult to reconcile with inherited human capital alone but aligns more closely with weaker selection among family insiders, consistent with inherited privilege. In line with this interpretation, dynasts are less likely to have formal training mentioned in their Grove biographies and more likely to be described as advancing through family connections, based on keyword and semantic similarity analysis. While lower rates of formal training could in principle reflect early exposure within families, that explanation sits uneasily with the underperformance of descendants relative to non-dynasts.

A potential concern is that our results reflect editorial bias. Because Grove is a curated reference work, dynasts may have been more likely to be included because of family ties, causing the observed discount to reflect asymmetric inclusion thresholds rather than genuine differences in prominence. To address this concern, we replicate the analysis using the Répertoire International des Sources Musicales (RISM), an independent archival catalog of surviving musical manuscripts.⁴ Because RISM records which composers’ works were historically copied, preserved, and circulated, it provides an independent measure of success not shaped by modern editorial discretion. The dynasty discount persists in the RISM data and is again driven by descendants rather than founders, making it unlikely that our results are driven by editorial selection.

We next examine how the dynastic discount evolved over time. The pattern is not stable: in the twentieth century, the dynasty discount reverses, with dynastic composers becoming more prominent, on average, than their non-dynastic peers. This reversal coincides with the rise of conservatories, which replaced informal, family-based apprenticeship with standardized instruction and selective admissions (Weber et al., 2001). Within our framework, conservatories are expected to narrow entry-cost differences between insiders and outsiders and to compress gaps in music-specific human capital by offering formal training to those without family-based access to expertise. Consistent with this interpretation, the dynasty discount is smaller in settings where formal training systems were established. Placebo tests, falsification exercises, and text-based proxies for formal training reinforce this result.⁵ Taken together, these findings suggest that the spread of conservatories reshaped selection in the composition profession, weakening inherited advantage and increasing the importance of merit and standardized credentials.

A growing body of work shows that insider ties—whether strictly familial or more broadly social—can distort the allocation of talent in elite settings. Evidence on family dynasties finds that dynastic scholars underperform in historical academia, an effect that fades in Protestant regions and in high-return scientific fields (de la Croix and Goñi, 2024). That study also offers a formal framework for distinguishing between nepotism and mean reversion, showing that dynastic underperformance is driven more by lower-ability sons than by exceptionally talented fathers. Studies of other inherited or socially mediated ties point to similar misallocations: firms run by family heirs earn lower returns (Pérez-González, 2006); connected candidates win election to elite science academies yet publish less (Fisman et al., 2018); and personal connections boost promotion chances in academia but reduce ex-post productivity (Zinovyeva and Bagues, 2015). Related work finds that credentialing reforms in Italy eliminated the advantage once enjoyed by dynastic lawyers, suggesting that changes in entry standards can mitigate inherited privilege (Raitano and Vona, 2021). Together, these findings motivate our analysis of musical dynasties and how credentialing reforms—via conservatories—altered the selection of talent over several centuries.

We extend this literature by focusing on a historical creative profession and by introducing several innovations in measurement and conceptualization. First, we distinguish between dynasty founders and descendants, enabling us to assess whether inherited access is associated with lower entry thresholds and weaker selection within families. Second, we define dynasties using extended kin networks—including siblings, cousins, and multi-generational links—rather than limiting attention to direct patrilineal succession, as is common in prior work. Third, we draw on a comprehensive biographical encyclopedia to map those kin networks, pair them with an independent archival catalogue (RISM), and develop text-based measures of formal training; this strategy broadens coverage beyond formal organizations, provides complementary outcome measures, and makes key human capital investments observable. Finally, we develop a simple model linking family background and training regimes to occupational sorting, which helps interpret observed changes in dynastic performance and generates testable predictions about selection dynamics.

Our results speak to research on human capital formation and credentialing. In Roy-style models, individuals sort into occupations or markets based on ability and returns, with patterns shaped by training costs, entry barriers, or migration incentives (e.g., Roy, 1951; Borjas, 1987; Heckman and Honoré, 1990). Conservatories acted as a credentialing reform: proximity to one narrows the dynasty discount, echoing evidence that modern credentials curb insider advantage in other professions by equalizing entry costs and lowering barriers for talented outsiders (e.g., Raitano and Vona, 2021). Similar patterns arise in academia, where changing entry barriers shifted the selection margin for women, producing a U-shaped pattern in relative performance (Iaria et al., 2024). We contribute to broader debates on how institutions shape the transmission of tacit knowledge. de la Croix et al. (2017) show how clan-, guild-, and market-based training systems influence the intergenerational transfer of skill. Our findings point to a similar change in music with the shift from family apprenticeship to conservatory training restructuring how musical knowledge was passed across generations. Building on recent text-as-data methods (see Gentzkow et al., 2019, for an overview), we convert historical biographies into structured measures—using keyword counts and sentence-embedding scores—to create novel proxies for formal training and broaden the toolkit for studying human capital investments in historical settings.

Finally, we contribute to cultural economics by analyzing how changing training structures affected access to creative careers. Prior work shows that copyright laws and agglomeration influenced creative output (Borowiecki, 2013; Giorcelli and Moser, 2020), while teacher-student linkages shaped stylistic transmission (Borowiecki, 2022). Our focus on dynastic access complements recent evidence that conservatories helped reduce gender gaps in musical prominence (Borowiecki et al., 2025). By combining family linkages with editorial and archival measures of success, we show how artistic training systems shaped the allocation of creative talent.

The remainder of the paper proceeds as follows. Section 2 presents a Roy-style model that links family background, training regimes, and performance, highlighting how inherited privilege and inherited skill jointly shape selection into composition. Section 3 describes the data, explains how dynasties are identified, and details our two prominence measures. Section 4 presents the core finding: a dynasty penalty, driven by descendants, and confirmed with both Grove and RISM data. Section 5 turns to training. We show, first, that dynastic composers are less likely to have formal instruction mentioned in their biographies and that their Grove entries also place greater narrative emphasis on family connections as a route to advancement; second, that the dynasty discount flips to a premium after conservatories spread; and third, that the penalty is attenuated in times and places where conservatories are present. Section 6 concludes and offers a back-of-the-envelope estimate of the welfare loss from dynastic misallocation.

2 Conceptual framework

We develop a simple Roy-style model to illustrate how family background shapes both selection into composition and subsequent performance. Family ties can influence two parameters: the cost of entering the profession and the amount of music-specific human capital inherited prior to entry. These two channels jointly determine who becomes a composer and how successful those entrants are, on average. The model is deliberately stylized, intended to organize empirical predictions rather than provide a fully structural representation of dynastic transmission.

Each individual has a general ability $a$, which is productive both in composition and in alternative occupations. This ability is drawn from a common distribution $F(a)$, identical across all lineage types. In composition, effective ability ($\tilde{a_j}$) depends on general ability and inherited, music-specific human capital ($h_j$):

$$\tilde{a}_j = a + h_j,$$

where $j \in \{n, f, d\}$ indexes non-dynasts ($n$), founders ($f$), and descendants ($d$). Non-dynasts have no inherited musical human capital ($h_n = 0$) and face an entry cost $c_n$.⁶ Founders may possess inherited musical capital ($h_f \ge 0$)—for instance, through early exposure or family involvement in music—but face similar entry costs as non-dynasts ($c_f \approx c_n$).⁷ Descendants, by contrast, benefit from lower entry costs through family reputation, patronage, or networks ($c_d < c_f$) while inheriting a fraction of the founder’s musical human capital ($h_d = \rho h_f$, with $0< \rho < 1$).

An individual chooses between composing and pursuing an outside occupation. Expected payoffs are

$$\pi _C = r(a + h_j) - c_j, \qquad \pi _O = w a,$$

where $r$ and $w$ denote the returns to ability in composition (C) and outside work (O). The individual enters composition if

$$a \ge \tau _j = \frac{c_j - r h_j}{r - w}.$$

Lower entry costs or higher inherited human capital both reduce the threshold and make entry more likely.

The model yields three testable implications. First, founders face essentially the same entry costs as non-dynasts ($c_f \approx c_n$) but may possess additional inherited musical human capital ($h_f> 0$). Because the entry threshold is $\tau _j = (c_j - r h_j)/(r-w)$, this inherited skill lowers the threshold for founders relative to outsiders even though their entry costs are the same. As a result, founders and non-dynasts undergo nearly identical selection on innate ability, and any performance advantage for founders must come from $h_f$ rather than from differences in entry cost. When inherited musical human capital is the dominant channel, this structure implies a founder premium: founders should outperform non-dynasts because their inherited skill augments effective ability without meaningfully altering selection on innate ability.⁸ Second, descendants inherit lower entry costs ($c_d < c_f$) but possess less music-specific human capital ($h_d = \rho h_f$); weaker selection then leads them to underperform founders on average. Third, the relative performance of descendants and non-dynasts depends on the balance of these effects: when inherited skill dominates ($rh_d \gg c_n - c_d$), descendants may outperform non-dynasts (a human-capital effect), whereas when access advantages dominate, ($c_d \ll c_n$ and $rh_d$ modest), descendants enter with lower innate ability and underperform (a privilege effect). Over time, inherited musical capital decays, generating mean reversion if family transmission is primarily skill-based; if privilege dominates, selection weakens and dynastic underperformance persists.

The institutional environment in which composers are trained shapes these dynamics. In particular, if conservatories reduce lineage-based access advantages by standardizing admissions and shrinking entry-cost differentials ($c_n - c_d$), and if they also narrow differences in music-specific human capital ($h_d - h_n$) by providing formal training to those without family-based instruction, then selection should become increasingly ability-based. Under these conditions, the performance gap between dynastic and non-dynastic composers is predicted to narrow following the spread of conservatories.⁹

3 Data

3.1 Sources and sample construction

Our primary dataset is constructed from Grove Music Online, an authoritative reference work on music and musicians. Grove includes entries on composers, performers, ensembles, genres, instruments, and musical terminology. As of 2024, the database contains over 60,000 entries, of which nearly 35,000—or approximately 57 percent—are biographical. Within this biographical subset, we identify more than 16,000 individuals whose occupational description includes “composer.” This subset forms the basis for our analysis.

Each Grove biography follows a standardized format, typically including a main biographical section as well as additional sections such as “Works,” “Writings,” and “Bibliography.” The main narrative provides a chronological account of the composer’s life, including details about education, professional appointments, compositional milestones, and their influence on music history. Because this section is available for all individuals in our sample, we use its word count as our primary measure of prominence. This measure reflects editorial judgments about the subject’s historical importance and is strongly correlated with other indicators of recognition, such as compositional productivity, scholarly attention, and inclusion in multiple reference sources.

We identify dynastic composers using Grove’s dedicated family entries, which provide structured accounts of prominent musical lineages.¹⁰ A composer is classified as a dynast if they are listed within one of these family entries. Using this approach, we are able to identify a total of 247 musical dynasties consisting of a total of 670 composers. We designate as the founder the earliest-born individual in each family group; all others are treated as descendants. This classification relies exclusively on inclusion in Grove’s family entries and does not incorporate inferred or unverified kinship ties.¹¹

To supplement our prominence measures, we incorporate data from the Répertoire International des Sources Musicales (RISM), which catalogs surviving musical manuscripts and printed sources. For each composer, we record the number of sources attributed to their name. This count provides an independent proxy for compositional output and historical circulation, capturing a dimension of prominence distinct from editorial assessments. Whereas Grove reflects curated evaluations of biographical significance, RISM offers a window into the documentary footprint of a composer’s work. Consistent with this distinction, the Pearson correlation between Grove word count and RISM source counts is approximately 0.39, indicating that while the two measures capture related aspects of prominence, they are not interchangeable. This moderate correlation is helpful for our purposes: it shows that RISM provides genuinely independent information rather than simply replicating Grove’s editorial assessment.

One advantage of using Grove and RISM is the breadth and consistency of coverage they provide across historical periods and geographic regions. Grove’s editorial oversight ensures structured entries and clearly defined family groupings. RISM, in turn, contributes a complementary perspective by documenting the physical survival of musical works. Because its source counts are based on archival cataloging rather than narrative curation, RISM is especially valuable for capturing the historical footprint of lesser-known composers who may receive limited biographical treatment in Grove. A key limitation, however, lies in the uneven availability of biographical information about family connections. Prominent composers have more detailed biographical entries and, are therefore more likely to be included in family groupings if applicable. In contrast, dynastic ties may go unrecorded for obscure individuals in two ways: their relatives—such as composer-parents—may not be included in Grove at all, or their entries may omit mention of familial relationships even when both individuals are present. This generates a form of non-classical measurement error, in which the likelihood of correctly identifying a dynast increases with prominence. As a result, true dynasts are more likely to be misclassified at the lower end of the prominence distribution. We return to this issue when interpreting our main results.

To examine broader changes in the training of composers, we supplement our dataset with information on the founding dates of music conservatories. These data are drawn from the International Directory of Music & Music Education Institutions (IDMMEI), an initiative affiliated with UNESCO that compiles information on post-secondary schools offering degrees in music or music education. From this source, we identify 2,174 conservatories worldwide, recording each school’s name, location (city, state, and country), and founding year. Founding dates are extracted manually from the descriptive text associated with each entry, and all locations are geocoded. These data allow us to track the historical expansion of formal music training and to examine how the growth of conservatories relates to the decline of family-based transmission within the profession

3.2 Descriptive statistics

Table 1, Panel A reports summary statistics for individual composers. The sample includes more than 16,000 composers born between 450 CE and the end of the twentieth century, with mean birth and death years of 1825 and 1848. Roughly four percent are identified as dynasts, meaning they appear in a Grove family entry, though this likely understates the true prevalence of dynastic ties among lesser-known figures. The average composer in the sample was therefore born during the Romantic era, when the number and visibility of professional composers expanded substantially.

Grove also reports composers’ additional occupations, many of which overlap. About ten percent are identified as conductors and another ten percent as pianists or organists, with smaller shares listed as teachers, violinists, or singers. Such overlap is especially common among earlier composers, who often combined composing with performing, church music, and teaching, whereas twentieth-century composers were more likely to specialize.

Prominence is measured using word counts from the main sections of Grove entries. The average main biographical entry contains 490 words, though the distribution is highly skewed: Beethoven’s exceeds 50,000. The “Works,” “Bibliography,” and “Writings” sections show similar dispersion. The average number of RISM sources is 151, but this too varies widely, from zero to several thousand.

Panel B reports summary statistics at the dynasty level. The average dynasty contains just over four composers, with the largest including 18. Dynastic composers are concentrated in earlier periods: their mean birth and death years are 1737 and 1771, placing the average dynast in the late Baroque or early Classical era, before the rise of nineteenth-century conservatory training. Grove entries for dynasty members are longer on average than those in the full sample, but the gap between mean and maximum word count within dynasties indicates substantial internal heterogeneity, with many families containing one especially prominent figure alongside lesser-known relatives.

Table 1 Summary statistics

3.3 Dynastic patterns over time

The presence of dynastic composers has varied substantially over the course of music history. As shown in Fig. 1, dynastic representation rose from an average of 4.5 percent in the sixteenth century to 8.5 percent in the seventeenth and just over 12 percent in the eighteenth-century, before falling sharply after 1800 to 4.2 percent in the nineteenth century and only 0.5 percent in the twentieth. This pattern suggests that family-based transmission played a much larger role in earlier periods of musical life than in the modern era.

Fig. 1

The decline of dynasties and the rise of conservatories

Figure 1 also shows that this decline coincided with the spread of formal music education. Conservatory openings remained limited until the nineteenth century and then increased rapidly, especially after 1850, as the dynastic share of composers fell. This divergence is consistent with the view that conservatory-based training reduced the importance of family-based transmission by broadening access to formal instruction.

Table 2 shows substantial regional variation in dynastic representation. Dynastic affiliation was most prevalent in France, the Low Countries, and the Germanic countries, where over 8 percent of composers were dynasts. Shares were lower in Italy and the British Isles and much lower in Russia, the United States, and the rest of the world. One possible explanation is that Russia and the United States entered the Western classical tradition later, when formal training had become more important and dynastic transmission had less scope to become entrenched.¹²

Table 2 Representation of dynastic composers by region

Figure 2 plots the lifespan of each dynasty with at least one composer recorded in Grove. Dynasty lifespans are measured from the birth year of the founder to the death year of the last descendant. Each horizontal line represents a single dynasty, sorted vertically by the birth year of its founder. The figure reveals a concentration of dynasties between 1600 and 1800, with particularly dense activity in the eighteenth century. The relatively steep slope of the curve during this period reflects a rapid rate of dynasty formation. In contrast, the curve flattens after 1800, indicating that new dynasties became increasingly rare. Most dynasties were relatively short-lived, typically spanning about 80 to 130 years, or roughly three to five generations.¹³ The overall pattern confirms that dynasties were historically concentrated in earlier periods, when musical training and employment were often mediated by family ties.

Fig. 2

Lifespan and historical clustering of composer dynasties

How large were dynasties in terms of membership? Fig. 3 reports the distribution of dynasty sizes. Most dynasties were small: nearly half consisted of exactly three composers, and the number declines steadily as family size increases. Large dynasties were rare, although a few—most famously the Bach family—contained more than ten composers. This pattern suggests that dynastic transmission in composition was typically limited to a few closely connected relatives rather than sustained through very large lineages.

Fig. 3

Distribution of dynasty size by number of members

4 Dynasty and prominence: empirical analysis

4.1 Baseline estimates of the dynasty effect

To estimate the relationship between dynastic background and a composer’s prominence, we begin with the following general specification:

$$\begin{aligned} \log (Y_i) = \beta \cdot Dynastic_i + \alpha _c + \delta _t + \theta _o +\eta _j + \varepsilon _i \end{aligned}$$

(1)

In this equation, $Y_i$ denotes a measure of prominence for composer i. Our primary specification uses the natural logarithm of the word count in the main biographical entry in Grove Music Online, but we also estimate parallel regressions using the number of sources attributed to each composer in the Répertoire International des Sources Musicales (RISM), applying the transformation $log(Y_i + 1)$ in that specification to accommodate zero values in the number of sources. The variable $Dynastic_i$ is a binary indicator equal to one if composer i is listed as part of a musical family based on Grove’s family entries. The specification includes fixed effects for country of birth ($\alpha _c$), to control for geographic variation in opportunities and documentation; for half-century-of-birth cohort ($\delta _t$), to absorb long-run historical changes in the size and structure of the composer population and to account for differences in average prominence across musical eras; for occupation ($\theta _o$), to capture systematic differences in prominence across occupational categories; and—when using Grove as the outcome—for biographical author ($\eta _j$), to control for stylistic variation across contributors. Standard errors are clustered at the country level.¹⁴

Columns (1) to (4) of Table 3 presents the results using Grove word count as the outcome variable. In the baseline specification (column 1), which includes country and cohort fixed effects, the coefficient on the dynastic indicator is negative and statistically significant, implying that dynastic composers are less prominent than their non-dynastic counterparts. Adding occupation fixed effects in column (2) does little to alter the estimate, suggesting that the discount is not driven by systematic differences in the types of roles dynasts held.¹⁵ Column (3) adds author fixed effects to account for variation in contributor style. The estimated discount increases slightly in magnitude to 0.156 log points, equivalent to a 14.4 percent reduction in word count. When all fixed effects are included (column 4), the magnitude increases slightly, to 0.162 log points (15 percent discount). The dynasty discount is also robust to alternative specifications that use word counts from other Grove sections—including “Works,” “Writings,” and “Bibliography”—as dependent variables (see Online Appendix Table A.5). These sections capture different facets of historical recognition: productivity (works), scholarly contributions (writings), and later citations or biographical treatments (bibliography). The findings from Grove are therefore consistent with the inherited privilege channel dominating: the negative coefficient suggests that lower entry costs weakened selection on innate ability, leading to reduced average prominence as reflected in expert assessments.

Table 3 Dynasty premium

Columns (5) and (6) of Table 3 present analogous regressions using RISM source counts as the dependent variable. As with the Grove results, dynastic composers appear less prominent than their non-dynastic counterparts. In the baseline specification (column 5), the dynasty coefficient is -0.261, corresponding to 23.0 percent fewer sources. When occupation fixed effects are included (column 6), the magnitude remains similar (-0.244 log points, or 21.6 percent fewer sources). That a similar dynasty discount emerges using a completely independent measure of prominence helps to alleviate concerns that measurement error in our primary success measure is driving the results.

These results help address two additional empirical concerns. First, our primary measure of prominence—Grove entry length—may be subject to editorial bias. Dynastic composers, especially descendants, might be included for the sake of genealogical completeness, while non-dynasts must clear a higher bar, producing a mechanical dynasty discount. Our inclusion of RISM mitigates this concern by providing an outcome based on archival evidence of source survival rather than editorial curation. That the discount persists—and indeed appears more pronounced—when using RISM suggests the pattern is not simply an artifact of asymmetric inclusion thresholds. Second, dynastic status is likely measured with error: some composers may be misclassified as non-dynasts either because relatives are omitted from Grove or because family ties are unreported. If this error is classical, it attenuates the coefficient toward zero; if non-classical—e.g., if prominent figures are more likely to have documented family ties—it biases our estimates upward. In either case, the true dynasty discount is likely larger than what we report, reinforcing our interpretation that dynastic composers, especially descendants, were on average less prominent than their peers.¹⁶

4.1.1 Robustness checks

To assess the stability of the dynasty discount, we estimate two additional sets of specifications that tighten the comparison groups. First, we include country-by-cohort fixed effects, which absorb broad cross-country differences in historical institutions—such as copyright regimes, patronage systems, and modes of musical training—as well as period-specific editorial practices in Grove. Second, because country-level categories can mask important local variation (and because national borders shifted over the long period we study), we re-estimate the models with sub-national fixed effects at the city, municipality, or district of birth.¹⁷ These finer controls compare composers from the same locality and absorb persistent regional differences in training or market conditions. As shown in Online Appendix Tables A.6 and A.7, the dynasty discount remains negative and statistically significant.¹⁸

A second set of robustness checks examines whether the dynasty discount depends on the functional form used to model prominence. Specifically, we supplement the baseline log specification with two common alternatives to the log transformation: percentile ranks and the inverse hyperbolic sine (IHS). Percentile ranks place composers on a uniform 0–1 scale that is unaffected by extreme values, while the IHS provides a smooth, log-like transformation that naturally accommodates zero observations. As shown in Online Appendix Table A.8, the estimated dynasty coefficients remain negative and of comparable magnitude using both approaches—the percentile-rank regressions imply a decline of roughly 5–10 percentile points, and the IHS estimates show similarly sized penalties—indicating that the dynasty discount is not driven by the choice of transformation.

A final robustness check uses unconditional quantile regressions (Firpo et al., 2009) to examine whether the dynasty discount varies across the prominence distribution (see Online Appendix Fig. A.9). For Grove, the estimated effect is negative at low quantiles, hovers around zero through most of the middle, and turns positive only at the very top, where a handful of exceptionally prominent dynasts drive the reversal. For RISM, the dynasty effect is negative across nearly the entire distribution, becoming imprecise only at the extreme upper tail. This pattern suggests that, on average and for most of the distribution, the privilege channel dominates the human-capital channel for dynasts: more marginal, lower-prominence composers are drawn into the profession under weaker selection thresholds. The fact that a few dynastic families (e.g., the Bachs) reach very high prominence shows that dynastic background can also coincide with unusually strong within-family advantages, but these cases are the exception rather than the rule.

4.2 Founders versus descendants

The dynasty discount estimated in Section 4.1 reflects the average prominence gap between composers from dynastic families and their non-dynastic peers, but this average may mask differences within dynasties themselves. Distinguishing between founders—the first family members to achieve recognition—and their descendants helps disentangle the two mechanisms outlined in Section 2. If dynastic persistence arises mainly through inherited human capital, founders should outperform non-dynasts, reflecting both individual talent and early exposure, with descendants gradually regressing toward the mean as inherited skills decay. If instead it stems from inherited privilege, founders—who faced similar entry barriers as outsiders—should perform about the same as non-dynasts, while descendants, benefiting from name recognition or family connections that lowered entry costs, should underperform.¹⁹

To assess which mechanism dominates, we next estimate regressions that separately identify the performance of founders and descendants relative to non-dynasts:

$$\begin{aligned} \log (Y_i) = \beta _1 \cdot Founder_i + \beta _2 \cdot Descendant_i + \alpha _c + \delta _t + \theta _o + \eta _j + \varepsilon _i \end{aligned}$$

(2)

As before, $Y_i$ is a measure of the prominence of composer i—either the word count of their Grove entry or the number of sources attributed to them in RISM. The variable $Founder_i$ is an indicator equal to one for the first born member of a dynasty, while $Descendant_i$ is an indicator equal to one for any subsequent member of a dynasty. The reference group is therefore the set of non-dynastic composers (i.e., those for whom both indicators are equal to zero). Under this coding, the coefficient $\beta _1$ captures the average difference in prominence between founders and non-dynasts, while $\beta _2$ captures the difference between descendants and non-dynasts. The difference between founders and descendants is therefore given by the $\beta _1 - \beta _2$. We also include country-of-birth fixed effects ($\alpha _c$), half-century of birth fixed effects ($\delta _t$), occupation fixed effects ($\theta _o$) and—when using Grove word count as the outcome—author fixed effects ($\eta _j$).

Table 4 Founders and descendants relative to non-dynasts (pooled descendants)

Columns (1) to (4) of Table 4 present the results using Grove word count as the dependent variable, while columns (5) and (6) report the corresponding estimates based on the number of sources in RISM. Regardless of specification or measure of prominence, founders are statistically indistinguishable from non-dynasts, while descendants are significantly less prominent than non-dynasts by between 16 and 28 percent. The difference between founders and descendants is statistically significant in the Grove regressions, indicating that descendants are less prominent than the founders of their lineages. This pattern suggests that dynastic underperformance reflects weaker selection among later-born entrants rather than family background per se. The fact that founders are roughly average while descendants are below average is also difficult to square with the human capital channel, which predicts exceptional founders followed by mean reversion toward average descendants. Instead, the pattern is more consistent with family ties lowering entry thresholds for later generations.

To examine how prominence varies across successive generations within dynastic families, we further disaggregate descendants according to their generational distance from the founder. Birth-year spacing provides a natural way to approximate lineage positions: composers born within roughly 20 years of the founder (generation 0) are most plausibly younger siblings or cousins; those born 20–40 years later (generation 1) are typically children, nephews, or nieces; those born 40–60 years later (generation 2) correspond to grandchildren or grand-nephews; and those born more than 60 years later (generation 3+) represent later collateral branches. This decomposition allows us to assess whether the prominence gap between founders and later family members narrows, widens, or remains stable across generations, thereby clarifying the generational pattern underlying dynastic persistence.

Table 5 Founders and descendants relative to non-dynasts by generational distance

Across all specifications and for both measures of prominence, the estimates in Table 5 display a clear pattern. Founders remain statistically indistinguishable from non-dynasts, while most descendant generations are substantially less prominent than non-dynasts. The magnitude of underperformance is largest for the groups closest to the founder, particularly generation 0 and generation 1; for later generations, the point estimates remain mostly negative but become increasingly imprecise, reflecting smaller sample sizes.²⁰ This pattern is difficult to reconcile with a pure inherited human capital story, which would predict exceptionally prominent founders and a gradual regression toward the mean. Instead, descendants underperform relative to outsiders from the earliest descendant cohort onward, consistent with weaker selection thresholds for entry into the profession. The generational decomposition therefore reinforces the main conclusion from the founder–descendant analysis: dynastic persistence reflects lowered barriers to entry rather than the intergenerational transmission of unusually high ability.²¹

To this point, our analysis has compared founders and descendants to non-dynasts in the full sample. While informative, these comparisons may still reflect systematic differences across families—some lineages are simply more visible, more extensively documented, or more historically esteemed than others. To verify that this difference is not driven by variation across families, we estimate a within-dynasty specification that includes dynasty fixed effects and restricts the sample to composers from dynastic lineages. This design compares founders and descendants from the same family while holding constant dynasty-level prominence and editorial inclusion thresholds. By absorbing family-specific visibility and reputation effects—for instance, the greater editorial attention or name recognition accorded to prominent lineages such as the Bachs—the within-dynasty approach yields a more credible estimate of the prominence gap attributable to founder status.

We implement this strategy by estimating the following within-dynasty regression model:

$$\begin{aligned} \log (Y_{id}) = \beta \cdot Founder_{id} + \gamma _d + \alpha _c + \delta _t + \varepsilon _{id} \end{aligned}$$

(3)

Here, $\log (Y_{id})$ denotes a measure of the prominence of composer i’s within dynasty d, proxied either by the word count of their Grove entry or by the number of sources attributed to them in RISM. The variable $\textit{Founder}_{id}$ is a binary indicator equal to one if composer i is the earliest-born member of their dynasty. The model includes dynasty fixed effects $\gamma _d$, which absorb all family-level characteristics that are shared by members of the same lineage, as well as country-of-birth fixed effects ($\alpha _c$) and half-century-of-birth cohort fixed effects ($\delta _t$).

Table 6 Within-dynasty premium

The results, displayed in Table 6, confirm that founders are substantially more prominent than other members of their family. In the Grove regressions, founders have biographical entries that are more than 50 percent longer than those of their dynastic descendants. The effect is even larger in the RISM regressions, where founders have at least 80 percent more surviving sources. These results are robust to the inclusion of additional controls for chronological position within the family, such as the number of years since the birth of the dynasty founder and an indicator for whether the composer is the last-known member of their dynasty. Because these estimates are identified from within-dynasty variation, they provide stronger evidence that the prominence gap between founders and descendants reflects real differences in selection, not unobserved variation across dynasties.

As a further check, we examine how a composer’s prominence varies with their position in the birth order within the dynasty. Rather than relying solely on a binary founder–descendant distinction, we estimate a model in which each dynasty member is categorized based on their chronological position within the family. Figure 4 plots the resulting coefficients, using the founder (first-born) as the reference group and including dynasty, country, and cohort fixed effects. All descendant categories lie well below zero, confirming that later family members are, on average, less prominent than the dynasty founder.

Fig. 4

Effect of order within dynasty on prominence

5 Training and access pathways, conservatories, and the transformation of dynastic advantage

The evidence presented so far shows that dynasty founders are, on average, no more prominent than non-dynasts, while their descendants perform markedly worse. This pattern sits uneasily with mean reversion and aligns more naturally with the view that dynastic ties lowered entry thresholds for later generations. However, that interpretation is based on outcomes, not on the investments or pathways that led to them. To probe the mechanism more directly, we turn to text evidence from Grove biographies.

Our text-analysis approach allows us to examine two complementary dimensions of early-career formation. First, we analyze the extent to which biographies emphasize formal musical training, using both keyword-based measures and semantic similarity metrics. If dynastic composers—especially descendants—entered under weaker selection thresholds, we would expect their biographies to devote less attention to structured instruction, either because they received less of it or because such training was less essential for securing early opportunities. Second, we construct a semantic measure of how strongly a biography emphasizes family connections or patronage in explaining a composer’s entry into the profession. Whereas the training measures speak to investments in skill acquisition, this connection-oriented index speaks to the channels of opportunity and advancement. Together, these text-based indicators characterize the pathways through which dynastic composers entered the profession and allow us to assess whether their routes differed systematically from those of non-dynasts.

While the textual analysis sheds light on whether dynasts relied on lower-cost or less formal entry routes, it does not, by itself, reveal how the prominence gap between dynasts and non-dynasts changed as the broader training environment evolved. To investigate how these developments shaped the dynasty effect, we proceed in two steps. First, we estimate cohort-specific dynasty coefficients to trace how the prominence differential shifted over time as conservatories expanded across Europe and North America. Second, we test whether the discount varied across space by linking composers’ birthplaces and dates to a proxy for local access to formal training. In the Roy framework developed earlier, the spread of conservatories should narrow the advantage of dynastic background by reducing differences in both entry costs and acquired human capital between dynasts and outsiders. Thus, the dynasty gap should shrink in periods and places where formal training became more accessible.

5.1 Training and access pathways into the profession

We use two complementary approaches to measure the emphasis on formal musical training in Grove biographies. The first relies on keyword analysis. We identify references to formal training in each composer’s entry using a curated list of training-related terms.²² Specifically, we code whether a biography contains any training reference, the number of times such keywords appear, and the fraction of words in the biography devoted to training-related content per 100 words. The keyword list is derived from word frequencies in the IDMMEI database’s descriptions of conservatories, academies, and other music schools. Because these profiles emphasize the core features of credential-based instruction, the most frequent terms provide a data-driven proxy for the language of formal musical education. To ensure relevance, we remove common stopwords and retain only terms that appear at least five times across all profiles. Keywords that could refer to informal or familial contexts (e.g., “training” or “student”) are excluded to ensure that the list reflects formal educational pathways rather than more diffuse learning environments. To validate this approach, we plot the share of Grove composer biographies with training-related keywords by birth-decade cohort against the number of conservatory openings by decade. The two series track each other closely ($r=0.92$), supporting the use of keywords as a proxy for exposure to formal training.²³

The second approach constructs a continuous index using pre-trained sentence embeddings from the Sentence Transformers library.²⁴ Each biography is represented as a vector that captures the contextual meaning of its sentences. We score these vectors based on their alignment with a semantic axis running from informal to formal training, defined by contrasting sets of anchor sentences such as “He earned a degree in composition from a prestigious institution” versus “He picked up piano by ear with no formal instruction.” This method improves on the keyword-based approach by capturing semantic meaning even when biographies use varied phrasing or implicitly refer to formal instruction. It also avoids misclassification in cases where keywords like “conservatory” appear in a negated context.

We focus initially on the first quarter of each biography, where early-life education and training are most likely to be discussed. This segment provides the clearest window into the conditions under which composers entered the profession, making it the most relevant for assessing whether founders, descendants, and non-dynasts followed different preparatory pathways. To evaluate the scope and persistence of any differences, we also replicate the analysis using the full text of each biography. This serves as a diagnostic check: if the dynastic gap is concentrated in the early part of the text and attenuates thereafter, it supports the interpretation that dynasts were less reliant on formal training at the point of entry, rather than reflecting systematic differences in the overall depth or structure of biographical coverage. This logic applies to both the keyword-based and embedding-based measures, each of which is computed separately for the first quarter and the full biography. All regressions control for log word count and include fixed effects for country and birth cohort to ensure comparability across biographies of different lengths, regions, and historical contexts.

Panel (a) of Table 7 reports the results for the first quarter of each biography, where early-life education is typically described. Relative to non-dynasts, dynastic descendants are about 15 percentage points less likely to have any mention of formal musical training, a sizable gap given the sample mean of 0.56. They also have significantly fewer training-related words, a lower fraction of training references, and a smaller score on the embedding-based formality index. Founders fall in between: they are more likely than descendants, but less likely than non-dynasts, to have formal training documented. This gradient suggests that formal instruction played a smaller role within dynastic families—most notably among descendants—consistent with the idea that inherited musical knowledge and access tended to substitute for credentialed education.

Table 7 Mentions of formal training in biography

Panel (b) extends the analysis to the full biographies. The estimated differences remain statistically significant in most specifications but are smaller than in the early-life results. Dynastic descendants are about 11 percentage points less likely than non-dynasts to have any mention of formal training, a decline of roughly 30 percent relative to the first-quarter estimate. The count and fraction of training-related keywords also shrink in magnitude, and the embedding-based index shows a similarly attenuated gap. Founders fall between descendants and non-dynasts, suggesting that differences in formal instruction diminished gradually across dynastic generations. The attenuation of the dynast–non-dynast gap when the full biography is used likely reflects the inclusion of mid- and late-career references—such as teaching appointments or honors—that dilute the signal from early training. Even so, the persistence of significant differences, particularly for descendants, supports the interpretation that family background increasingly substituted for credentialed instruction in shaping entry into the profession.²⁵

To complement the analysis of formal training, we examine whether dynastic composers are described in Grove as advancing through family connections. Using the same sentence-embedding approach introduced above, we construct a semantic index that captures the extent to which each biography uses language associated with family-based or patronage-driven access rather than merit-based advancement.²⁶ Table 8 shows that both founders and descendants receive substantially higher nepotism scores than non-dynasts. The coefficients—ranging from roughly 0.06 to 0.08 for founders and 0.07 to 0.09 for descendants—are large relative to the mean value of the index (about 0.05 in the first-quarter sample and 0.03–0.04 for the full biography). The larger coefficients for descendants suggest that biographers more frequently describe their careers in terms of inherited advantages and family networks. The positive coefficient on founders may partly reflect the construction of the measure, since a founder’s biography can register family-based advancement when it notes that he helped a later relative enter the profession. Taken together, these linguistic patterns indicate that dynastic composers are portrayed as advancing through family-based channels. This narrative emphasis aligns with the view that dynastic persistence operated primarily through inherited access rather than unusually high inherited ability.

Table 8 Mentions of family-based advancement in biography

5.2 Temporal patterns in dynastic advantage

To determine whether the dynasty effect shifted over time, we re-estimate our baseline Grove specification (i.e., Eq. 1) separately by century of birth. This cohort-based approach allows us to trace how the prominence gap between dynastic and non-dynastic composers evolved across periods marked by major changes in professional training that reduced entry costs for dynasts relative to non-dynasts. For each century, we estimate multiple versions of the model that vary in their inclusion of country, cohort, and author fixed effects, providing a robustness check on the resulting pattern.

Figure 5 plots the estimated coefficients on the dynastic indicator for each century, across the different fixed effects specifications. The results show a marked dynasty discount in the eighteenth and nineteenth centuries, where dynastic composers appear systematically less prominent than their peers. In the twentieth century, however, the dynasty effect reverses direction: the estimates turn positive, and in some cases statistically significant, despite smaller sample sizes. This shift suggests that the historical relationship between dynastic background and prominence was not constant but evolved over time, potentially in response to broader changes in how composers entered the profession.

Fig. 5

Temporal variation in the dynasty effect

This temporal pattern also offers a useful test of a concern raised earlier: that the dynasty discount found in Grove might reflect editorial bias, specifically, a tendency to include dynasts for genealogical reasons, even when their accomplishments were limited. If this were the case, we would expect the estimated dynasty effect to persist or even grow more negative over time, especially as genealogical information became more complete and editors were better able to identify family relationships. Later dynasts, particularly those born in the twentieth century, would then be even more prone to inclusion at low thresholds of prominence. Yet we find the opposite: the dynasty discount disappears and becomes a premium. This reversal is difficult to reconcile with the editorial bias hypothesis.

A more plausible explanation is that the selection mechanism into musical composition changed over time. In earlier centuries, dynastic background often granted access to positions through family ties, patronage, or informal training networks.²⁷ By the twentieth century, however, access increasingly depended on formal instruction and credentialed evaluation. The spread of conservatories may have narrowed entry-cost differentials by providing outsiders with structured pathways and reduced human-capital differentials by offering systematic training that partially substituted for music-specific knowledge historically transmitted within families.²⁸ When standardized training sufficiently diminishes dynasts’ relative access advantages—or even raises their effective thresholds—the model admits a reversal: only high-ability dynasts choose to enter, allowing inherited musical capital to dominate weakened privilege. Although we do not explicitly model these mechanisms, the disappearance of the dynasty penalty is consistent with a broader shift from inherited roles to competitive credentialing. At the same time, other institutional and technological changes also shaped musical careers in this period, making the precise sources of the reversal difficult to isolate. The resulting dynasty premium should therefore be interpreted cautiously.²⁹

5.3 Conservatories and dynastic prominence

To explore further whether the prominence of dynastic composers is shaped by access to formal training, we estimate regressions that interact dynastic status with exposure to conservatories. The goal is to test whether the dynasty discount varies systematically across different training environments. We proxy conservatory access using a binary indicator for whether composer i was born in a location and time period where a conservatory existed. The specification includes an interaction term to capture whether the prominence gap between dynasts and non-dynasts is attenuated—or even reversed—under formal training regimes:

$$\begin{aligned} \log (Y_i) =\ &\beta _1 \cdot Dynasty_i + \beta _2 \cdot Conservatory_i \nonumber \\&+\ \beta _3 \cdot (Dynasty_i \times Conservatory_i) + \alpha _c + \delta _t + \varepsilon _i \end{aligned}$$

(4)

As before, $Y_i$ denotes the prominence of composer i measured using either Grove word counts or the number of RISM sources. The coefficient $\beta _1$ captures the average dynasty discount in settings without conservatory access, while $\beta _3$ captures how that discount changes in the presence of formal training institutions. All regressions include fixed effects for country of birth and half-century of birth. In some specifications, we also include occupation fixed effects and fixed effects for Grove author. If conservatories narrow entry cost differentials and human capital gaps between dynasts and non-dynasts, we expect $\beta _1<0$ (a dynasty discount in the absence of conservatories) and $\beta _3>0$ (an attenuation of that discount when conservatories are present).

There are several reasons to treat this analysis cautiously. First, conservatories were not randomly placed; founding decisions reflected local politics and demand, and the resulting bias could go in either direction. Second, exposure is measured coarsely: any composer born within 100 km of an operating conservatory by age 15 is coded as treated, although actual enrollment is unobserved. Accordingly, non-differential misclassification will attenuate estimates toward zero.³⁰ Third, the binary indicator conflates elite academies with small teaching schools, and many conservatories were founded in regions with few dynasts (e.g., Russia and North America), both of which further bias the interaction toward zero. Finally, the twentieth century also saw major changes in musical labor markets, including the rise of public funding bodies, recording and broadcasting technologies, and copyright regimes that altered the returns to talent and reputation in ways we cannot separately identify from conservatory access. Some of these forces likely worked in the same direction as conservatories in weakening lineage-based privilege, while others may have offset it. In light of these limitations, we interpret the conservatory results as descriptive rather than causal. The measurement issues we highlight (coarse exposure and the heterogeneity of institutions) imply that, to the extent conservatories did narrow dynastic gaps in training and access, our estimated attenuation is probably conservative. But we remain cautious about attributing the full reversal of the dynasty discount solely to conservatory expansion, given the broader set of institutional and technological changes unfolding over the same period.

Table 9 Conservatories and dynastic prominence

Regression results are shown in Table 9. In columns (1)-(4) the dependent variable is Grove word count while in columns (5) and (6) it is RISM source counts. We report Conley standard errors, which account for spatial correlation in unobservables that decay with distance (Conley, 1999).³¹ This adjustment is important because treatment is geographic: composers born near the same conservatory may share unobserved regional advantages not fully absorbed by country or cohort fixed effects. In all specifications, the interaction between dynastic status and conservatory exposure is positive. This suggests that the dynastic discount narrows in settings where formal training was more accessible. The interaction remains consistently positive and of a similar order of magnitude when we vary the distance used to define exposure (from 20 to 200 km).³² These results align with our Roy model: credential-based entry appears to reduce outsiders’ cost disadvantage and narrow the selection gap, thereby diminishing the dynasty penalty.

To test whether the conservatory interaction reflects genuine selection rather than omitted advantages, we run three checks. First, we ask whether dynastic composers in conservatory regions live longer—using age at death as a health proxy—and find the interaction term is indistinguishable from zero. Second, we repeat the exercise with an elite social status index built from sentence embeddings of the opening lines of each Grove biography; this interaction term is small and statistically insignificant.³³ Both results indicate that healthier or better-connected dynasts are not driving our main finding. Third, we create an exposed too late variable that equals one if a conservatory first opened within 100 km of the composer’s birthplace only after the composer had turned 30—too late to affect training. If geography rather than access explained the pattern, the interaction between dynastic and exposed too late should still be positive; instead it turns negative (and is sometimes significant), reinforcing the view that timely conservatory access, not regional advantages, narrows the dynasty gap. See Online Appendix Tables A.9–A.11 for full results).

Finally, as a robustness check, we re-estimated our interaction regressions using a proxy for exposure to formal musical training based on the density of formal training-related keywords in the first quarter of each Grove biography.³⁴ This keyword-based measure complements our geographic exposure strategy and yields consistent results: the interaction between dynastic status and training exposure is positive and statistically significant across both Grove word count and RISM source count specifications, reinforcing the conclusion that access to formal training mitigates the dynasty penalty.³⁵

6 Conclusion

This paper examines how family background shaped entry into the profession of classical composition over a millennium of music history. Drawing on new data covering over 16,000 composers, we link dynastic status to long-run measures of recognition and archival presence. We find that dynastic composers are, on average, less prominent than their non-dynastic peers, but this discount is driven entirely by descendants. Founders—those who initiated a musical lineage—perform on par with non-dynasts, while descendants underperform both. The empirical pattern we uncover—average founders and underperforming descendants—points toward nepotism rather than inherited ability as the primary force sustaining dynastic persistence.

To shed further light on how dynasts entered the profession, we analyze the biographical narratives in Grove using text-based indicators of training and entry pathways. Dynastic composers, particularly descendants, receive less emphasis on formal instruction and more on family-based advancement, suggesting that connections often substituted for conventional training. To investigate how the prominence gap between dynasts and non-dynasts evolved as the broader training environment changed, we trace the dynasty discount across cohorts and across regions with differing access to conservatories. Consistent with the idea that credentialing reforms narrowed gaps in both entry costs and human capital between dynasts and outsiders, we find that the dynasty discount is smaller in regions and periods with greater access to conservatory training. While not a causal design, the association survives placebo and timing falsification tests, holds under text-based proxies for training exposure, and aligns with the model’s predictions, lending support to the interpretation that credentialing reforms helped level the playing field.

To assess the broader implications of dynastic underperformance, we conduct an illustrative counterfactual exercise that combines two components: the average shortfall in measures of composer prominence among dynastic descendants, and a stylized estimate of the loss from excluding exceptionally high-ability outsiders who might otherwise have entered the field. The first component is based on regression-adjusted gaps in prominence—24 percent in Grove word counts and 30 percent in RISM archival source counts (see Table 4)—under the assumption that dynastic descendants performed as well as average non-dynasts. Given that descendants comprise 2.8 percent of our sample, this implies baseline losses of 0.67 percent in Grove and 0.83 percent in RISM. To this, we add an estimate of tail losses, assuming that dynastic entry displaced ten individuals who would have been two standard deviations above the mean and one who would have been three standard deviations above, consistent with a normal distribution applied to a group of 456 displaced non-dynasts. Because creative impact is highly skewed, even a few such exclusions can generate disproportionately large losses. Including these tail losses and adjusting the baseline to avoid double-counting, we estimate combined welfare losses of 1.05 percent in Grove and 1.40 percent in RISM. If the share of dynastic descendants were as high as 15 percent—a conservative figure, well below the roughly one-fifth to one-quarter prevalence documented for European craft guilds and late-medieval Flemish trades (Prak et al., 2020; Stabel, 2007)—the implied welfare losses would rise to 5.67 percent for Grove and 7.5 percent for RISM. See Online Appendix G for a full derivation of these estimates.

These estimates are conservative in several respects. First, the descendant–non-dynast performance gap is estimated over the full sample, including the twentieth century, when the dynasty discount reverses. The true gap in earlier centuries, when family background and informal training dominated, was likely larger. Second, we assume a normal distribution of talent to approximate tail losses, but the actual distribution of creative impact is almost certainly more fat-tailed. This makes the assumed tail-loss scenario conservative, and suggests that the real consequences of lost genius may be substantially greater (Jones, 2014; Merton, 1968).³⁶ Third, our calculations assume one-for-one displacement of non-dynasts by dynastic composers, but dynastic dominance may also have discouraged broader participation, either by reducing the expected returns to training or by shaping social norms around who belonged in the profession.³⁷ These general equilibrium effects are not captured in our estimates but could magnify the long-run consequences of restricted entry (e.g., Heckman and Honoré, 1990; Hsieh et al., 2019). Fourth, our outcome measures reflect only the most visible forms of historical significance, and do not account for lost stylistic diversity, interrupted pedagogical lineages, or the broader cultural costs of excluding high-ability outsiders.

More generally, these findings contribute to a growing literature on the allocation of talent and the conditions under which inherited advantage either facilitates or distorts achievement. While our study focuses on classical music, the mechanisms we identify—linking family background, selection thresholds, and career outcomes—apply broadly across high-skill professions. Recent empirical work shows that family background and early exposure shape who enters innovative and high-impact careers, limiting the realization of latent talent among disadvantaged groups (Bell et al., 2019). Theoretical and quantitative research further demonstrates that such distortions in the allocation of talent can have substantial aggregate consequences for innovation and long-run growth (e.g. Murphy et al., 1991; Jones, 2014; Hsieh et al., 2019). Understanding when lineage enhances performance and when it undermines selection is essential for evaluating how talent is cultivated and deployed over time. Although historical misallocations cannot be reversed, they offer a powerful lens for understanding the costs of inefficient talent allocation and the potential gains from expanding opportunity.