The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
•••••••••••••••• • •••••••••• xko BonIer: Pia .H. On'enills B2B eMarketplace Announcements and Shareholder Wealth Alldrl:tU'1-1 Cl ell emd 77'01110' F 'iem Consolidation, Technology, and the Changing Structure of Banks' Small Business Lending David P. Ely and Kennelb]. Robinson This publication was digitized and made available by the Federal Reserve Bank of Dallas' Historical Library (FedHistory@dal.frb.org) [eonomie dna findneidl Review Federal Reserve Bank of Dallas Robert D. McTeer, Jr. P"esident and Chief Executive Officer Helen E. Holcomb First Vice President and Chief Operating Officer Robert D. Hankins Senior Vice President, Banking Supervision Harvey Rosenblum Senior Vice President and Director of Research W. Michael Cox Senior Vice President and Chief Economist Editors Stephen P. A. Brown Senior Economist and Assistant Vice President John V. Duca Senior Economist and Vice President Jeffery W. Gunther Research Officer Mark A. Wynne Research Officer Director of Publications Kay Champagne Associate Editors Jennifer Afflerbach Monica Reeves Art Director Gene Autry Design & Production Laura J. Bell Economic and Financial Review (ISSN 1526-3940), published quarterly by the Federal Reserve Bank of Dallas, presents in-depth information and analysis on monetary, financial, banking, and other economic policy topics. Articles are developed by economists in the Bank's Economic Research and Financial Industry Studies departments. The views expressed are those of the authors and do not necessarily reflect the positions of the Federal Reserve Bank of Dallas or the Federal Reserve System. Articles may be reprinted on the condition that the source is credited and the Public Affairs Department is provided with a copy of the publication containing the reprinted material Subscriptions are available free of charge Please direct requests for subscriptions, back issues, and address changes to the Public Affairs Department, Federal Reserve Bank of Dallas, P.O Box 655906, Dallas, TX 75265-5906; call 214-922-5254; or subscribe via the Internet at www.dallasfed.org. Economic and Financial Review and other Bank publications are available on the Bank's web site, www.dallasfed.org. Contents IIleg~llmmigr~tion ~nd [nforcement Along t~e U.S.-Mexico Border: An Overview Pia M. Orrenius Page 2 B2B emM~etplace Hnnouncement~ ~nd ~~Me~older We~lt~ Andrew H. Chen and Thomas F. Siems Page 12 (onsolid~tionr Tec~nologYr ~nd t~e (~~nging Structure of B~nhr Sm~1I Business lending David P. Ely and Kenneth J. Robinson Page 23 Illegal Mexico-U.S. migration has increased dramatically in recent decades. In this article, Pia Orrenius evaluates the causes of this migration and gives an overview of the enforcement and policy responses to date. Orrenius assesses the effectiveness of border enforcement by looking at developments in the smuggling industry, such as smuggler use rates and fees, as well as changes in bordercrossing sites. The findings suggest early attempts at enforcement fueled an increase in the demand for and supply of smugglers, with no rise in prices. Only the most recent enforcement initiatives, most significantly Operations Hold-the-Line and Gatekeeper, have been successful in reversing the thirty-year decline in smugglers' fees and moving migrants to remote crossing points. Risks have risen along with smugglers' fees, as reflected in an increasing number of crossing-related deaths since 1995. Orrenius concludes that Mexican and u.s. policymakers should consider a bilateral labor and migration agreement. In the business-to-business (B2B) sector, new supply-chain models within electronic marketplaces (eMarketplaces) offer firms significantly lower procurement costs, increased operating efficiencies, and expanded market opportunities. Using event-study methodology to look at the period July 1999-March 2000, Andrew Chen and Thomas Siems find that investors reacted favorably to B2B eMarketplace announcements, with slightly higher abnormal returns associated with vertical than with horizontal eMarketplaces. They also find significant positive abnormal returns for e-commerce technology providers that partnered with computer industry giants or with competitors in B2B e-commerce initiatives. The abnormal returns are more than three times greater than returns from creating a B2B eMarketplace alone or with Old Economy leaders. These results suggest that, at least for the period studied, shareholders valued alliances between B2B eMarketplace developers more than firms developing e-commerce strategies on their own or with an Old Economy partner. The U.S. banking industry continues to consolidate, with large, complex banking organizations becoming more important. Traditionally, these institutions have not emphasized small business lending. On the other hand, technological advances, particularly credit scoring models, make it easier for banks to extend small business credit. To see what effects these influences might have generated on small business lending, David Ely and Kenneth Robinson explore the small business lending patterns at U.S. banks from 1994 through 1999. They find that larger banks are increasing their market share, most noticeably in the smallest segment of the small business loan market. The authors also present evidence that the size of the average small business loan has declined, especially at larger organizations, and that the gap in lending focus on the smallest small business loans has narrowed between small and large banks. These trends are consistent with increasing use of credit scoring models. The U.S.–Mexico border is experiencing an era of unparalleled trade and exchange. But at a time when legal flows of goods and people are at historical highs, so are illegal cross-border flows of undocumented migrants. Illegal immigration from Mexico became more common in the late 1960s, following the end of the Bracero Program in 1964. The Bracero Program allowed Mexican guest workers to work legally in the United States. Over the past three decades, illegal immigration along the Southwest border has increased, and enforcement efforts have intensified as a result. Border apprehensions have grown from 200,000 in 1970 to more than 1.5 million in 1999. The cumulative impact of this immigrant flow is a sizable illegal immigrant population. The undocumented immigrant population from Mexico was estimated at 3.1 million in 1997.1 Mexicans make up about 60 percent of the total undocumented population of the United States, and Central Americans from El Salvador, Guatemala, Honduras, and Nicaragua make up another 13 percent (U.S. INS 1999). Whereas some undocumented immigrants arrive legally and simply overstay their tourist visas, the majority of illegal Mexican and Central American immigrants residing in the United States cross the border without documents. It is estimated that the net inflow of illegal immigrants from Mexico, excluding short-term cyclical migrants, averaged about 202,000 immigrants per year between 1987 and 1996.2 This article evaluates the determinants of illegal Mexico–U.S. migration and gives an overview of enforcement and policy responses. Many observers, noting the large number of illegal immigrants, have concluded that border enforcement provides little deterrent. Some research supports this view. Singer and Massey (1998) show declining apprehension rates along the border in the 1980s and early 1990s. Other research shows that apprehended migrants simply attempt additional border crossings until they succeed —also suggesting increased enforcement has little impact (Kossoudji 1992). On the other hand, there is evidence increased border enforcement is correlated with falling wages for young males in Mexican border cities (Hanson, Robertson, and Spilimbergo 1999). This could imply that tougher border enforcement has the effect of trapping would-be immigrants on the Mexican side of the border. Publicized reports about the increase in migrant deaths also imply that tougher border enforcement sends migrants on circuitous routes into the United States (Nevins 2000, Rosenblum Illegal Immigration and Enforcement Along the U.S.–Mexico Border: An Overview Pia M. Orrenius T his article evaluates the determinants of illegal Mexico–U.S. migration and gives an overview of the enforcement and policy responses to date. Pia M. Orrenius is an economist in the Research Department of the Federal Reserve Bank of Dallas. 2 FEDERAL RESERVE BANK OF DALLAS Jalisco, and Guanajuato—is a traditional source of U.S.-bound migrants. The MMP survey asks randomly sampled heads of households for family, job, and migration histories.3 The migration rate is depicted in Figure 1 and includes both legal and illegal trips. As shown, migration rates more than doubled between 1965 and 1995, rising from 3.7 percent to 7.5 percent by the end of the sample period. Sustained increases in migration are associated with the 1970s and the mid-1980s, with an all-time peak of almost 10 percent reached in 1988. The other data source on illegal immigration is Immigration and Naturalization Service (INS) data on the number of illegal aliens apprehended by the Border Patrol each year. Although apprehensions also reflect the intensity of enforcement, discussed in detail below, the time series shown in Figure 2 is largely consistent with the migration patterns observed in the household survey data in Figure 1. Apprehensions rose from about 21,000 in 1960 to more than 1.5 million in 1999, with steep increases in the 1970s, in the mid-1980s leading up to the Immigration Reform and Control Act (IRCA), and again in 1994–96. For comparability, I also plot the rate of illegal immigration in the MMP sample for the years available. Illegal immigration in the MMP sample is highly correlated with apprehensions up until the early 1990s. As households drop out of the MMP data (households are sampled only once) or are legalized through amnesty under IRCA, the sample becomes less representative. This problem becomes more severe after 1991. Figure 1 Mexico – U.S. Migration Rate, 1965–95 Percent 12 10 8 6 4 2 0 ’65 ’68 ’71 ’74 ’77 ’80 ’83 ’86 ’89 ’92 ’95 SOURCE: Mexican Migration Project. 2000). In this article, I explore the timing and pattern of substitution among border-crossing sites. I also look at developments in the smuggling (coyote) industry, such as changes in smuggler use rates and smugglers’ fees, to assess the effectiveness of border enforcement over the past thirty-five years. Migration is the outcome of both push factors within Mexico and pull factors in the United States. Migrant family networks and smugglers have facilitated illegal immigration, while wage and employment differentials have encouraged it. Although early border enforcement had little impact, more recent efforts are having an effect. Early enforcement attempts fueled an increase in the demand for smugglers, with no corresponding rise in coyote prices. Only the most recent enforcement initiatives, most significantly Operations Hold-the-Line and Gatekeeper, have been successful in reversing the thirty-year decline in smugglers’ fees and moving migrants to remote crossing points. Risks have risen along with smugglers’ fees, as reflected in an increasing number of crossing-related deaths since 1995. In light of these developments, I conclude that now is a good time for Mexican and U.S. policymakers to consider a bilateral labor and migration agreement. Figure 2 Border Patrol Apprehensions and Illegal Immigration, 1960 – 99 Apprehensions (in thousands) Migration rate (percent) 1,800 7 Rate of illegal immigration 1,600 6 1,400 5 1,200 4 1,000 Apprehensions ILLEGAL IMMIGRATION: DATA SOURCES AND TRENDS 800 2 To get an idea of the changes in illegal immigration on the Southwest border, I rely on two data sources. The first is the Mexican Migration Project (MMP 1999), collected in western Mexico between 1987 and 1997. Western Mexico—particularly the states of Michoacán, ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 3 600 400 1 200 0 0 ’60 ’63 ’66 ’69 ’72 ’75 ’78 ’81 ’84 ’87 ’90 ’93 ’96 SOURCES: Immigration and Naturalization Service; Mexican Migration Project. 3 ’99 ORIGINS AND DETERMINANTS OF MEXICO –U.S. MIGRATION Determinants of Illegal Immigration The resumption of illegal immigration following the end of the Bracero Program illustrates the power of economic and social factors. The links formed between employers, recruiters, and migrant workers during the Bracero period lowered the costs and risks of migrating to the United States for work (Massey et al. 1987). Key factors such as networks and the availability of people smugglers known as “coyotes” allow migration to rise. The underlying determinants of Mexico–U.S. migration, however, are the higher wages and job availability in the United States. The policy backdrop is also important. Laws that exist but are not enforced, such as IRCA employer sanctions, signal tolerance for illegal immigration. The Bracero Program Large-scale immigration from Mexico has its roots in the Bracero Program, a guest-worker program started in 1942 (Massey et al. 1987). It arranged for the temporary legal immigration of workers from Mexico to the United States, in part to cover U.S. labor shortages resulting from World War II. Following the war, however, U.S. farmers and fruit and vegetable growers successfully lobbied the government to extend the program through the 1950s. It was not until 1964 that organized labor’s call for an end to foreign contract labor was heeded and the Bracero Program was abandoned (Calavita 1992). From 1942 to 1964, the Bracero Program brought in an average of more than 200,000 workers annually (Calavita 1992). The majority of braceros were concentrated in Texas, California, Arkansas, Arizona, and New Mexico. When the bracero agreement was terminated in 1964, it was under a cloud of controversy, and consequently no worker exchange was put in its place. The program’s abrupt end eventually resulted in a new era of largely illegal immigration from Mexico. The new era had a slow start, in part because of strong economic growth in Mexico in the 1960s. Also, in 1965, Mexico instituted a border industrialization program that has become known as the maquiladora program, specifically intended to create jobs for laid-off migrants.4 Nevertheless, by the early 1970s, the movement of Mexican migrants to the United States was accelerating again. Networks Although economic factors such as wage differentials are considered paramount, they cannot be acted upon if migration costs are prohibitive. A first-time illegal migrant must overcome significant fixed costs to obtain information about the destination and how to make a successful trip.5 Immigration researchers have found networks —family members and friends with migration experience —as the most common way in which this crucial information is transmitted to the potential migrant. The Bracero Program laid the foundation for mass illegal immigration partly through the creation of networks and the dissemination of information pertinent to Mexico–U.S. migration and employment in U.S. labor markets. Since then, networks have continued to expand, as Figure 3 shows for sibling networks (defined as having a sibling with U.S. migration experience). Nineteen percent of MMP households had access to at least one sibling network in 1965, whereas 41 percent had access in 1991. These same households averaged 1.7 sibling networks in 1965 and 2.3 in 1991. Moreover, an increasing proportion of sibling networks settled permanently in the United States over this period. In other words, both the quantity and quality of migrant networks are changing. Figure 3 Access to Sibling Networks and Number of Sibling Networks, 1965 – 95 Percent Number of networks 45 3 40 2.5 35 Number of networks 30 2 Smugglers Along with migrant networks, the availability of people smugglers, or coyotes as they are commonly called, makes the cross-border trip possible for many undocumented immigrants. Coyotes can be hired in a migrant’s hometown or along the border and typically accompany the migrant to his ultimate destination.6 The smuggler’s fee, or “coyote price,” rep- 25 1.5 20 Percentage with networks 15 1 10 .5 5 0 0 ’65 ’68 ’71 ’74 ’77 ’80 ’83 ’86 ’89 ’92 ’95 SOURCE: Mexican Migration Project. 4 FEDERAL RESERVE BANK OF DALLAS Figure 4 ficiently to allow potential migrants to respond to changing factors such as relatively low Mexican wages and economic downturns. Massey and Espinosa (1997) and Orrenius (1999) provide a comprehensive look at determinants of Mexico–U.S. migration. Economic downturns cause unemployment in cities, depress agricultural prices in the countryside, and make loans difficult to repay. Figure 4 plots Mexican GDP per capita (total GDP and agricultural GDP) since 1965. The apparent surge in Mexican emigration in the mid-1980s is consistent with declines in real income at that time. Mexican manufacturing wages tell a similar story and, in 1999, were still below the peak levels reached in 1981. Agricultural sector output, although less volatile than national output, fell throughout the latter half of the sample period. Total and Agricultural Mexican GDP, 1965–97 Index, 1965 = 100 (GDP per capita) 200 180 160 Total GDP 140 120 Agricultural GDP 100 80 60 40 20 0 ’65 ’67 ’69 ’71 ’73 ’75 ’77 ’79 ’81 ’83 ’85 ’87 ’89 ’91 ’93 ’95 ’97 SOURCES: Instituto Nacional de Estadística, Geografía e Informática; Banco de México. resents a major cost of illegal immigration. Interestingly, despite increased enforcement, coyote prices were on a steep downward trend during most of the sample period. By 1994, coyote prices averaged about $300—one-third of 1965 prices. (Coyote prices are discussed in more detail below.) The most important reasons for the falling coyote prices were the development of infrastructure and free entry into the coyote industry. Construction of infrastructure such as roads and airports and the growth of twin cities along the border, such as Tijuana/San Diego and Ciudad Juárez/El Paso, made the border more accessible to travelers from Mexico’s interior. Before 1930, for example, only two railways connected central Mexico with the U.S. border, and no major roads connected the Mexican interior with any U.S. border city (Scott 1982). Most roads linking the interior to the border were built between 1940 and 1960. Similarly, the expansion of commercial air transport during these years was dramatic. As a result, travel times were significantly shortened, allowing coyotes to charge less for their services. Another factor in falling coyote prices has been free entry into the industry. In theory, any migrant who has undertaken an illegal border crossing can use the experience to work as a coyote. This implies that as illegal immigration became more commonplace, more and more migrants entered the smuggling trade.7 More competition among suppliers pushed prices lower. Insurance and Capital Markets Figure 4 suggests that low incomes and intermittent downturns or economic crises act as push factors and generate out-migration. However, one strand of literature argues that simply the risk of recession can also generate emigration in good times. This theory emphasizes the need to insure the household’s income against negative local shocks by coordinating the migration and remittances of particular household members in good and bad times (Stark and Bloom 1985). Underdeveloped capital markets also make borrowing difficult or impossible for many Mexicans. In surveys, migrants often cite the need for capital to start a business, build a house, repay a loan or fund a medical procedure as a major reason for migrating to the United States. Immigration Policy Policies in both home and host countries also affect the dynamics of migration. In Mexico, the government’s failure to generate consistent economic growth and stable financial institutions leads to higher emigration. In the United States (and many other countries), simply restricting immigration to below the global demand for visas creates an incentive for foreigners to illegally immigrate. Other U.S. policies, such as the generosity of public assistance programs and the availability of health care, also have an impact. The two most significant U.S. policies enacted in recent years, however, are the Immigration Reform and Control Act (IRCA) and the Illegal Immigration Reform and Immigrant Responsibility Act (IIRIRA). IRCA was passed in 1986 in response to increasing illegal immigration in the mid-1980s. Economic conditions worsened in Mexico in the Wages With networks and coyotes in place, the costs of illegal immigration have decreased suf- ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 5 false attestation of citizenship punishable by up to five years in prison. Meanwhile, welfare reform legislation passed in 1996 denied illegal immigrants and many legal immigrants access to most public benefits. IIRIRA has probably not had a strong effect on the volume of illegal immigration, but it has made adjustment to legal status more difficult for illegal aliens residing here. The welfare reform and denial of benefits such as food stamps have mostly affected legal immigrants. Figure 5 Illegal Crossings, 1965 –96 Percent of total crossings 90 80 70 60 50 40 30 20 EVALUATING RECENT TRENDS IN BORDER ENFORCEMENT 10 0 U.S. authorities responded to rising illegal immigration by increasing enforcement. As shown in Figure 6, enforcement hours (measured by the number of hours Border Patrol agents spend on linewatch duty) grew in three phases over this period (Dunn 1996). In this section, each enforcement phase is reviewed with regard to its likely effect on the probability of apprehension. Coyote use rates, coyote prices, and migrant crossing patterns are proxies for enforcement efficacy. ’65 ’67 ’69 ’71 ’73 ’75 ’77 ’79 ’81 ’83 ’85 ’87 ’89 ’91 ’93 ’95 SOURCE: Mexican Migration Project. early 1980s, and by 1983 both the migration rate and apprehensions began to rise. IRCA sought to deter illegal immigration by allocating more funds to the INS and border enforcement, imposing sanctions on employers who knowingly hired undocumented workers, and creating an agricultural guest worker program (H-2A). IRCA also offered amnesty to agricultural workers and long-term U.S. residents. The IRCA amnesty ultimately legalized nearly three million illegal immigrants —about two-thirds of them from Mexico. In the short term, IRCA’s passage is correlated with an increase in total immigration but a lowering of both the illegal immigration rate in the MMP data and the number of border apprehensions (Figure 2 ) (Bean, Edmonston, and Passel 1990; Orrenius 2000). IRCA may have failed to stem illegal immigration in the medium to long term in part because employer sanctions are rarely enforced and the guest worker program is too small and narrow to fill employers’ and migrant workers’ needs. The short-run effects may have been the result of legalizing the usual stream of illegal immigrants, many of whom cycle regularly between the United States and Mexico. Figure 5 shows the dramatic impact the IRCA amnesty had on migrants in the MMP sample, reducing the percentage of illegal crossings from 77 percent in 1986 to 29 percent in 1989. Passed ten years after IRCA’s implementation, IIRIRA followed up on some IRCA provisions by further increasing penalties on smugglers as well as illegal entrants. IIRIRA also mandated a doubling of the Border Patrol by 2001, increased penalties on document fraud, streamlined deportation proceedings, limited judicial review of deportation orders, and made Three Phases of Enforcement In early enforcement efforts, up until 1986, linewatch hours lagged the influx of migrants. Hours rose in the late 1970s when, in the face of rising illegal immigration, the Carter administration approved increased INS funding (Rosenblum 2000). Much of the additional money went to hardware and equipment. In the next phase, during the Reagan administration, border and immigration issues took on heightened urgency, and the passage of IRCA in 1986 marked a turn- Figure 6 Border Patrol Linewatch Hours, 1964 –99 Hours (in thousands) 9 8 Phase 1 Phase 2 7 Phase 3 6 5 4 3 2 1 0 ’64 ’66 ’68 ’70 ’72 ’74 ’76 ’78 ’80 ’82 ’84 ’86 ’88 ’90 ’92 ’94 ’96 ’98 SOURCE: Immigration and Naturalization Service. 6 FEDERAL RESERVE BANK OF DALLAS Figure 7 ing point for border enforcement and for immigration policy more generally. A large portion of the 33 percent increase in INS funding was earmarked for the Border Patrol, and the effect on linewatch hours is apparent in Figure 6. Congress also strengthened penalties against migrant smugglers. As illegal immigration began to grow again in the early 1990s, yet another enforcement initiative was undertaken. The third phase of enforcement, which started in 1993 and used site-specific crackdowns, marked by far the biggest increase in linewatch hours. The objective was to make illegal immigration costly by diverting illegal traffic out of border cities and away from roads and buildings (U.S. GAO 1999b). Agents took up fixed positions along commonly used paths within urban areas. Along with fencing and surveillance equipment, this forced illegal migrants away from densely populated areas. Once in remote areas, the illegal aliens could be more easily spotted and detained by the Border Patrol.8 The strategy was first implemented in El Paso (Operation Hold-the-Line), then in 1994 in San Diego (Operation Gatekeeper) and Nogales, Arizona (Operation Safeguard), and last in 1997 in South Texas (Operation Rio Grande). As a result, between 1993 and 1997, the budget for enforcement along the Southwest boundary more than doubled. The number of Border Patrol agents rose from 4,200 in 1994 to 7,700 in 1999 (U.S. GAO 1999a). Smuggler Use Rates and Fees, 1965– 97 Percent 1,000 90 80 900 Smuggler fee Use rate 800 70 700 60 600 50 500 40 400 30 300 20 200 10 100 0 0 ’65 ’67 ’69 ’71 ’73 ’75 ’77 ’79 ’81 ’83 ’85 ’87 ’89 ’91 ’93 ’95 ’97 SOURCE: Mexican Migration Project. likely to use a coyote when they perceive a higher chance of apprehension from attempting a crossing on their own. Another related variable is the price coyotes charge. Coyote prices should rise with apprehension probabilities, all other things the same, since the risk to the smuggler increases with the likelihood of getting caught. The bars in Figure 7 plot the percentage of illegal immigrants in the survey data that hired coyotes in each year. Coyote use increased steeply in 1970 and trended upward for the rest of the decade. By 1979, more than 70 percent of illegal immigrants in the sample were hiring coyotes. After leveling off in the early 1980s, coyote use rates trended slightly upward in the early years of IRCA (1986–1990). This pattern provides some evidence that, despite the increase in overall illegal immigration, costs to migrants rose during the first two enforcement phases. Figure 7 also illustrates, however, that increases in the supply of smugglers outpaced increases in the demand for smugglers, since prices fell despite higher use rates. In real terms (1994 dollars), the median reported coyote price fell from more than $900 in 1965 to about $300 in 1994. By 1996 –1997, however, the coyote price trended upward again. Higher post-1994 prices are consistent with an impact of heightened enforcement on smuggling fees. Coyote use rates peak in the survey data in 1996. Because the MMP sample becomes thinner at the end of the sample period, the last data point is slightly less reliable. In any case, anecdotal evidence supports the premise that by 1995, the border had become much harder to cross. In fact, the most recent border initiative, the series of site-specific crackdowns start- The Evidence on Smuggler Use Rates and Smuggling Fees To deter illegal immigration, heightened border vigilance must raise the costs migrants face. This is usually done by increasing the probability of apprehension (but can also be accomplished by raising other risks to the migrant such as the probability of injury or death). Has the probability of being apprehended, and hence the cost and risk to the migrant, increased during the enforcement periods under study? Apprehension probability cannot be directly measured because the number of illegal immigrants attempting crossings is unknown.9 Apprehensions could be rising because of increased numbers of immigrants and not because of increased probability of capture. An alternative measure of changes in the probability of apprehension is changes in related variables not directly affected by the volume of illegal immigration, for example, illegal immigrants’ propensity to hire coyotes. Migrants should be more ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 1994 dollars 100 7 Figure 8 movers headed toward California. In Mexico, cities along the way, such as Mexicali and Tijuana, began to grow. From the survey data in Figure 8, we can see that from one-half to three-fourths of all border crossings between 1965 and 1990 were into California. Following IRCA, there was a decline in the fraction of crossings into California and a corresponding increase in the propensity to cross into Texas. These trends intensified following the implementation of Operation Gatekeeper. Gatekeeper also led to increased crossings into Arizona, although this fact is not obvious from patterns in the MMP data (it is clear from the INS apprehensions data). The empirical trends are consistent with the interpretation that, with the enactment of IRCA and Gatekeeper, border enforcement in California became more effective than Texas enforcement. Border crossers responded by shifting to Texas and Arizona. Border Crossings by State, 1965 – 94 Percent of crossings 80 70 California 60 50 40 Texas 30 20 10 Arizona 0 ’65 ’67 ’69 ’71 ’73 ’75 ’77 ’79 ’81 ’83 ’85 ’87 ’89 ’91 ’93 SOURCE: Mexican Migration Project. ing with Operations Hold-the-Line and Gatekeeper, has probably been the most successful enforcement effort to date. For the first time, anecdotes of border crossers being apprehended with such frequency that they turn back, giving up their hopes of reaching the United States, are common. Enforcement hours and apprehension statistics are at all-time highs and coyote prices are increasing for the first time in more than thirty years. Intrastate Reallocation of Migrants Within Texas the changes are equally striking. Looking at the data in Figure 9, the increase in Texas crossings beginning in 1990 is almost entirely concentrated in the El Paso sector. The resumption of crossings in El Paso following IRCA was an important factor in the decision to implement Operation Hold-the-Line. The data reveal the impact of the 1993 crackdown with a 75 percent decrease in apprehensions within one year. The subsequent rise in apprehensions in the other Texas sectors, however, suggests migrants responded by crossing farther south. The change in preferred bordercrossing sites is particularly noticeable following The New Enforcement Strategy and the Change in Crossing Sites Another telling sign that recent crackdowns are a deterrent is the disruption of longstanding border-crossing patterns. Immigrants today shun formerly popular crossing points in California in favor of Texas and Arizona. Within states, the change is also noticeable. In California, migrants choose to cross the harsh deserts of El Centro rather than risk a crossing south of San Diego. In Texas, migrants are less likely to attempt an El Paso crossing, preferring to cross farther south in Texas through Laredo, McAllen, Brownsville, and most recently, Del Rio. Figure 9 Border Patrol Apprehensions by Texas Sectors, 1960 – 99 Apprehensions (in thousands) 450 400 Interstate Reallocation of Migrants As home to most of the U.S. border with Mexico, Texas historically has been the primary site of economic and cultural exchanges between the two countries. In 1900, the population of California was only half that of Texas, and the California–Mexico border was largely unpopulated (Lorey 1999). After the Depression, however, California overtook Texas in both economic and population growth. An enormous westward migration ensued, within both the United States and Mexico, with the majority of McAllen–Laredo 350 300 250 200 El Paso 150 100 50 Del Rio–Marfa 0 ’60 ’63 ’66 ’69 ’72 ’75 ’78 ’81 ’84 ’87 ’90 ’93 ’96 ’99 SOURCE: Immigration and Naturalization Service. 8 FEDERAL RESERVE BANK OF DALLAS On the Mexican side, policymakers face a different dilemma, although there are pros and cons to Mexico–U.S. migration for Mexico as well. The out-migration of Mexican citizens brings in $4 billion to $7 billion in remittances each year, with funds flowing to some of the country’s most poverty-stricken areas. Emigrants are the third-largest source of foreign reserves after trade and tourism. Emigration has reduced the pressure on politicians in handling economic crises at the local and national levels. However, Mexico has lost millions of workingage men and women to the United States. At the local level, the impact of mass emigration has been severe in places. Villages and towns have been depopulated. In years of economic growth, labor markets have had to adjust, partly through the rising labor force participation of women. For Mexican policymakers, the best-case emigration scenario might be a population of emigrants who leave in bad times, remit lots of cash, and come home in good times to work, invest, and run businesses. The policy implication is for Mexico to foster closer ties with the emigrant community and encourage the United States to allow more border-crossing mobility. This view is seemingly shared by Mexican President Vicente Fox, who proposes a renewal of bilateral migration agreements with the United States and suggests a border that would allow for the freer movement of both people and goods. One perverse outcome of the border crackdown and other U.S. immigration laws has been to discourage illegal immigrants who used to cycle in and out from returning home (Orrenius 2000). From both countries’ perspectives, a program incorporating temporary, workbased migration of Mexicans to the United States may prove the most beneficial arrangement (Orrenius and Viard 2000).11 This would limit the fiscal impact on U.S. taxpayers and allow Mexican migrant workers to keep their Mexican residences and cycle freely between the two countries. the December 1994 peso crash. Apprehensions in the McAllen–Laredo sector rose to unprecedented levels after 1995. Border-Crossing Deaths A specific intention of the new border enforcement strategy has been to eliminate illegal alien traffic from city centers. The consequence has been to divert migrants into more sparsely populated areas. Illegal immigrants today cross through inhospitable terrain and expose themselves to dangerous climactic extremes far more than they did ten or twenty years ago. Critics of the border offensives claim that injuries and deaths along the border are at an all-time high as a result. Rosenblum (2000) cites the number of crossing-related deaths at 324 in 1999, up from single digits before 1995. Deaths are believed to have numbered 388 in 2000. The Mexican estimate is 430. POLICY IMPLICATIONS OF ILLEGAL IMMIGRATION Policymakers face difficult choices on the issue of immigration. Just as the individual migrant faces costs and benefits from migrating, host (home) countries experience costs and benefits from immigration (emigration). Host countries often fight illegal immigration to minimize the fiscal burden of immigrants, to limit workplace competition for natives, and to heed native concerns about issues ranging from immigrant assimilation to cultural and linguistic erosion. At the same time, policymakers have come to understand migrant workers’ role in a growing economy. In the U.S. case, authorities seem unwilling to incur the economic consequences of ending illegal immigration. Consequently, IRCA-imposed sanctions against employers who hire undocumented workers are rarely enforced. Moreover, the INS has largely abandoned its former tactic of work-site raids, and its de facto policy since 1997 has been “once you are in, you are in.”10 The outcome of the two opposing forces has been a steady stream of illegal aliens. Foreign policy toward Mexico seems to imply, however, that large and sudden changes in that stream are undesirable. The loan bailout of 1995 was partly defended on the grounds that it would slow the Mexican out-migration resulting from the peso’s 1994 crash. IRCA was similarly intended to defend against the fallout of Mexico’s debt crisis in the 1980s. NAFTA proponents also argued that improved economic conditions, attained through trade, would generate less emigration from Mexico. ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 CONCLUSION Since World War II, boundaries between Mexico and the United States have diminished. A hundred years ago, wage differences were as large as they are today, yet there was virtually no migration between the two countries. Exchange of people and goods was limited by distance, the lack of roads and airplanes, a paucity of information, and linguistic and cultural differences. Today, after fifty years of large- 9 scale migration and settlement, the scenario is vastly different, and Mexico–U.S. trade and migration have grown significantly. The response to trade has been positive, but the response to illegal immigration has been a corresponding increase in the intensity of border enforcement. Early enforcement efforts in the 1970s and 1980s were largely ineffectual. They succeeded in raising coyote use rates among migrants, which created a flourishing smuggling industry offering a steadily decreasing fee schedule. The more recent enforcement initiatives have been successful, however, resulting in higher coyote prices and possibly higher rates of discouraged migrants who give up trying to cross the border. Additional evidence is the change in migrant crossing patterns. When one site has been chosen for a crackdown, the effect has been almost immediate, as migrants have responded by crossing elsewhere. Migrants currently shun formerly popular crossing points such as San Diego and El Paso in favor of more remote routes. The inter- and intrastate reallocation of migrants is apparent in both INS apprehensions data and Mexican survey data. Unfortunately, as border-crossing options have been reduced, migrants are risking more to make it to the United States. The result has been a record number of crossing-related deaths. The controversy over border enforcement’s impact on illegal immigration has led naturally to more debate on the larger question, namely the costs and benefits of illegal immigration. There are policy instruments that would allow both countries to garner the benefits of Mexico–U.S. migration while mitigating the costs. One of these is the development of bilateral migrant worker agreements that would provide for the legal and temporary entry of Mexican workers into the U.S. labor market. and Components of Change: 1987–1997,” http://wwwa.house.gov/lamarsmith. 2 3 4 5 6 7 8 9 10 11 The number of entries into the Mexican-born unauthorized-resident population is estimated at 330,000 per year. (See the report referenced in footnote 1). The difference is due to emigration, status adjustment, deportation, and death. The migration history includes number, timing, and destination of trips; legal status on trip; mode of crossing (alone, with family, or with smuggler); and amount of smuggler’s fee. The sample is selected on male household heads aged 15 – 65. Initially, maquiladoras largely employed female workers. As a result, the border industrialization program did little to help employ displaced braceros (Gruben 1990). In an expected utility framework, the risk-averse migrant suffers disutility from the uncertainty surrounding a trip to an unknown destination and involving an illegal border crossing. The perceived risks are therefore a migration deterrent. Spener (forthcoming) describes the smuggler’s role and methods in more detail. The increased drug trade in the 1980s may have lured more people into the smuggling trade. According to agents, this strategy has worked less well where vegetation hides the migrant — as in Laredo and the lower Rio Grande Valley. Despite these difficulties, Singer and Massey (1998) developed estimates of the probability of apprehension. They find that the average (1965– 92) probability of apprehension is 0.33 and that apprehension probabilities peaked in the late 1970s but have declined since. Internal enforcement since 1997 has focused on the apprehension and deportation of criminal immigrants. The guest worker program proposed by Sen. Phil Gramm incorporates some of these ideas, including issuing temporary visas to Mexican nationals working illegally in the United States. REFERENCES NOTES 1 Bean, Frank D., Barry Edmonston, and Jeffrey S. Passel (1990), Undocumented Migration to the United States: IRCA and the Experience of the 1980s (Washington, D.C.: The Urban Institute Press). The author would like to thank Gordon Hanson for providing the enforcement data, Jason Saving for his comments on an earlier draft, Jennifer Afflerbach for careful editing, and Border Patrol Supervisory Agent LeRoy Schleinkofer for sharing his experiences and insight. This number does not include the estimated 2.8 million illegal immigrants who were given legal residency under the 1986 amnesty provision of the Immigration Reform and Control Act. See the report entitled “Annual Estimates of the Unauthorized Immigrant Population Residing in the United States Calavita, Kitty (1992), Inside the State: The Bracero Program, Immigration, and the INS (New York: Routledge, Chapman & Hall). Dunn, Timothy (1996), The Militarization of the U.S.– Mexico Border, 1978 –1992 (Austin: CMAS Books, University of Texas Press). 10 FEDERAL RESERVE BANK OF DALLAS Gruben, William C. (1990), “Mexican Maquiladora Growth: Does it Cost U.S. Jobs?” Federal Reserve Bank of Dallas Economic Review, January, 15 – 29. Orrenius, Pia M., and Alan D. Viard (2000), “The Second Great Migration: Economic and Policy Implications,” Federal Reserve Bank of Dallas Southwest Economy, Issue 3 May/June, 1– 8. Hanson, Gordon H., Raymond Robertson, and Antonio Spilimbergo (1999), “Does Border Enforcement Protect U.S. Workers from Illegal Immigration?” NBER Working Paper Series no. 7054 (Cambridge, Mass.: National Bureau of Economic Research, March). Rosenblum, Marc R. (2000), “U.S. Immigration Policy: Unilateral and Cooperative Responses to Undocumented Immigration,” IGCC Policy Paper no. 55 (La Jolla, Calif.: Institute on Global Conflict and Cooperation, May). Kossoudji, Sherrie A. (1992), “Playing Cat and Mouse at the U.S.–Mexican Border,” Demography 29 (May): 159 – 80. Scott, Ian (1982), Urban and Spatial Development in Mexico (Baltimore, Md.: The Johns Hopkins University Press). Lorey, David E. (1999), The U.S.–Mexican Border in the Twentieth Century (Wilmington, Del.: Scholarly Resources Inc.). Singer, Audrey, and Douglas S. Massey (1998), “The Social Process of Undocumented Border Crossing Among Mexican Migrants,” International Migration Review 32 (Fall): 561– 92. Massey, Douglas S., Rafael Alarcón, Jorge Durand, and Humberto González (1987), Return to Aztlan: The Social Process of International Migration from Western Mexico (Berkeley: University of California Press). Spener, David (forthcoming), “Smuggling Mexican Migrants Through South Texas: Challenges Posed by Operation Rio Grande,” in Global Human Smuggling: Comparative Perspectives, ed. David J. Kyle and Rey Koslowski, (Baltimore, Md.: The Johns Hopkins University Press). Massey, Douglas S., and Kristin E. Espinosa (1997), “What’s Driving Mexico–U.S. Migration: A Theoretical, Empirical and Policy Analysis,” American Journal of Sociology 102 (January): 939 – 99. Stark, Oded, and David E. Bloom (1985), “The New Economics of Labor Migration,” American Economic Review 75 (May): 173 –78. Mexican Migration Project (1999), Population Studies Center, University of Pennsylvania, Philadelphia (producer and distributor), www.pop.upenn.edu/mexmig/ welcome.html. U.S. General Accounting Office (1999a), Border Patrol Hiring: Despite Recent Initiatives, Fiscal Year 1999 Hiring Goal Was Not Met, GAO/GGD-00-39 (Washington, D.C.: GAO). Nevins, Joseph (2000), “Immigration Control and the Remaking of the California–Mexico Boundary in the Age of NAFTA,” in The Wall Around the West: State Borders and Immigration Control in North America and Europe, ed. Peter Andreas and Timothy Snyder (Lanham, Md.: Rowman and Littlefield), 99 –114. ——— (1999b), U.S.–Mexico Border: Issues and Challenges Confronting the United States and Mexico, GAO/NSIAD-99-190 (Washington, D.C.: GAO). U.S. Immigration and Naturalization Service (1999), 1997 Statistical Yearbook of the Immigration and Naturalization Service (Washington, D.C.: Government Printing Office, October). Orrenius, Pia M. (2000), “Does Increased Border Enforcement Trap Illegal Immigrants Inside the United States?” Federal Reserve Bank of Dallas, unpublished paper. ——— (1999), “The Role of Family Networks, Coyote Prices and the Rural Economy in Migration from Western Mexico,” Federal Reserve Bank of Dallas Research Paper no. 99-10 (Dallas, November). ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 11 The advent of the Internet as an instrument for business commerce has fundamentally altered the economy by ushering in increased efficiencies and more transparent markets. Since businesses started conducting Internet transactions in 1995, the growth has been impressive. Forrester Research estimates that worldwide electronic commerce (e-commerce) revenues were about $650 billion in 2000 and projects they will grow to $6.8 trillion by 2004.1 The greatest impact is in the businessto-business (B2B) sector, where new supplychain models within electronic marketplaces (eMarketplaces) enable companies to significantly lower procurement costs and increase operating efficiencies. B2B eMarketplaces streamline the supply chain by making better use of more information. The time it takes to match buyers and sellers can be radically reduced, precautionary inventory levels can be lowered, and the range of potential suppliers and distribution outlets can be expanded as geographic boundaries disappear. Projections of online B2B revenues differ vastly, primarily because defining what counts and how to count it varies widely. But despite these differences, forecasters agree that online B2B trade will grow substantially. B2B e-commerce is generally believed to account for about 80 to 90 percent of total e-commerce today. Forecasts typically project U.S. online B2B revenues of about $2 trillion by 2003, up from roughly $336 billion in 2000. B2B e-commerce is expected to impact the U.S. economy significantly. Brookes and Wahhaj (2000) argue that the rapid growth of B2B e-commerce will have an economic impact over and above that of the normal process of innovation and productivity growth. They suggest that as a result of B2B e-commerce, annual GDP growth in the large industrialized countries should rise an average 0.25 percent for the next ten years—with the level of GDP eventually 5 percent higher than it would otherwise have been. Brookes and Wahhaj conclude that the dominant long-run effect of B2B e-commerce will be on output and equity markets, rather than on inflation and bond markets. Investors should respond favorably to announcements of new B2B e-commerce initiatives, as long as they believe these moves will ultimately result in higher profits and increased productivity without fueling inflation. We examine the potential impact of B2B e-commerce initiatives on the New Economy paradigm using the efficient markets hypothesis (Fama et al. 1969), which implies stock prices reflect all B2B eMarketplace Announcements and Shareholder Wealth Andrew H. Chen and Thomas F. Siems C lassifying B2B eMarketplace announcements by the type of eMarketplace and the type of partner reveals key differences in how the financial markets assess B2B e-commerce strategies. Andrew H. Chen is distinguished professor of finance at Cox School of Business, Southern Methodist University. Thomas F. Siems is a senior economist and policy advisor in the Research Department of the Federal Reserve Bank of Dallas. 12 FEDERAL RESERVE BANK OF DALLAS be assessed. The returns to firms that ally themselves with other technology providers and those that team with Old Economy leaders can also be compared. Classifying B2B eMarketplace announcements by the type of eMarketplace and the type of partner reveals key differences in how the financial markets assess B2B e-commerce strategies. available information about individual companies and about the economy as a whole. Information is the key input. So in efficient capital markets, prices will immediately adjust to reflect any new information. Thus, B2B e-commerce announcements should immediately raise stock prices if investors believe a firm’s value will be increased by higher net future cash flows resulting from higher productivity, lower costs, or higher revenues. This article empirically investigates B2B eMarketplace announcements from the financial market’s perspective. Overall, are online B2B exchanges creating shareholder wealth? Do the returns to shareholders of firms that announce vertical (intra-industry) exchanges and those that announce horizontal (cross-industry) exchanges differ? Are the returns higher when firms go it alone in developing an eMarketplace than when they do so with other B2B e-commerce companies or Old Economy leaders? And what are the returns to shareholders of firms that acquire other B2B e-commerce technology providers? This article addresses these questions using event-study methodology, a useful tool for examining the consensus estimates of future benefits attributable to organizational initiatives.2 Stock returns are analyzed relative to a portfolio of stocks representing the market. Differences in returns are analyzed on days leading up to and following the event date —in this case, the B2B eMarketplace announcement date—to determine whether shareholder returns differ significantly from the general market return for stocks. The strength of this methodology is that it captures a large number of investors’ overall assessment of a firm’s discounted present value. Subramani and Walden (1999) were the first to use event studies to explore why firms might pursue e-commerce initiatives. They examined 305 e-commerce announcements made between October and December 1998 and found that these announcements resulted in positive cumulative abnormal returns to firms’ shareholders. Contrary to their hypothesis, they found that business-to-consumer (B2C) e-commerce announcements resulted in higher abnormal returns than B2B announcements. For B2B initiatives, Subramani and Walden found average abnormal returns of 5.9 percent on the event date and 11.3 percent for a three-day window starting one day before the event. Using the event-study methodology, differences in returns to companies engaged in vertical and horizontal B2B eMarketplaces can ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 WHY EMARKETPLACES? Information networks create transparency in markets that substantially reduces transaction costs. Previously unavailable or hard-to-obtain intelligence increases transparency. In exchange markets, transparency along the supply chain regarding price, availability, competing suppliers, and alternative products can radically change the dynamics of the buyer–seller relationship. Both parties can benefit as shared information increases competition and reduces costs for searching, bargaining, decisionmaking, policing, and enforcement. Internet exchanges introduce unprecedented market and process transparency. B2B eMarketplaces can provide nearly perfect information at all points along the supply chain, increasing efficiency and lowering participants’ costs. Such exchanges also enable companies to develop, manage, and monitor internal and external processes — including work in process and finished-goods inventories — far more efficiently and effectively. The improved coordination that results gets the right goods and services to the right places at the right times with lower costs. B2B eMarketplaces can be divided into two types: vertical and horizontal. Vertical markets are industry-specific; they focus on an individual industry, such as steel, plastics, electronic components, or chemicals. Electronic exchanges in vertical markets serve participants primarily by bringing buyers and sellers together to transact business up and down the entire industry supply chain. They also provide industry-specific news and information and other value-added services, such as employment opportunities, discussion forums, and event calendars that create community within the industry. These benefits can substantially reduce operating costs. In contrast, horizontal markets cross industries. They focus on creating an exchange for goods and services at a specific link in the supply chain that is common to multiple industries, such as MRO (maintenance, repair, and operations) supplies, logistics, and benefits administration. 13 Typically, goods and services exchanged over horizontal eMarkets are standardized and can be outsourced to third-party providers that have well-defined, fixed-price products. As a result, the value added by horizontal eMarketplaces is in automating workflow and reducing process costs to the participants of the exchange. This enables businesses in various industries to operate more efficiently and effectively. We further divide B2B eMarketplace announcements by the type of partner, if any, the e-commerce technology provider said it would be working with, using the following five categories of partnership: Acquisition. The e-commerce technology provider announced plans to acquire another technology firm to aid in the development of a B2B eMarketplace. Alone. The e-commerce technology provider announced plans to develop a B2B eMarketplace on its own. Alliance: Computer. The e-commerce technology provider announced plans to develop a B2B eMarketplace with a large and well-recognized computer industry leader, such as IBM Corp., Microsoft Corp., or EDS Corp. Alliance: Competitor. The e-commerce technology provider announced plans to develop a B2B eMarketplace with a competitor. Alliance: Old Economy. The e-commerce technology provider announced plans to develop a B2B eMarketplace with an Old Economy leader (for example, General Motors or Ford in the automotive industry, Shell or Chevron in the energy industry). Using these partnership classifications, various hypotheses can be tested to determine the value of different B2B e-commerce strategies. view a B2B eMarketplace as an unprofitable strategy, perhaps because they suspect factors other than shareholder maximization motivated the initiative. Such factors might include management’s level of compensation, job security, and span of control. Among the benefits touted in announcements of new B2B eMarketplaces are the ability to: • Expose sellers in one marketplace to all potential buyers. • Create a hub for development projects, market feedback, and customer collaboration. • Reduce time to market. • Provide expansive catalogs of products and services. • Provide end customers with fast response, high cost efficiency, and superior service. • Increase operating efficiency through an integrated Internet supply chain. • Streamline purchasing operations. • Reduce supply-chain costs, increase manufacturing efficiency, and reduce inventories. • Reduce cycle times, improve transaction flows, and manage parts inventories. Hypothesis 2: Higher Returns for Vertical eMarketplaces Than for Horizontal eMarketplaces Because vertical eMarketplaces focus on the needs of an entire industry (up and down the supply chain) and horizontal eMarketplaces focus on specific business processes that span multiple vertical markets (individual links in the supply chain), we expect vertical eMarketplaces will have higher abnormal returns than horizontal eMarketplaces. If horizontal eMarketplaces have abnormal returns higher than vertical eMarketplaces, this might indicate that investors consider productivity improvements gained through providing goods and services at a specific link in the supply chain across industries of greater value than efficiency gains along the supply chain. HYPOTHESES Hypothesis 1: Positive Returns for B2B eMarketplace Announcements As Subramani and Walden (1999) discuss, e-commerce initiatives should position firms to exploit the growing importance of and expected growth in electronic commerce, leading to benefits in the future. Such initiatives signal that a firm plans to use information technology to better manage industry supply chains. Consequently, we expect that investors will react favorably to B2B eMarketplace announcements, resulting in positive abnormal stock market returns (that is, risk-adjusted returns in excess of average stock market returns) around the date of the announcement. Alternatively, negative abnormal returns might indicate that investors Hypothesis 3: Insignificant Returns to Firms Announcing the Acquisition of Another E-Commerce Technology Provider Generally, alliances and mergers are designed to create competitive advantages and should therefore enhance market valuations. However, we expect that announcements of e-commerce technology firms’ plans to acquire another such provider will not result in signifi- 14 FEDERAL RESERVE BANK OF DALLAS then be attributed to individual events. The strength of the method lies in its ability to identify such abnormal changes because it is based on the overall assessment of many investors who quickly process all available information in assessing a firm’s market value (McWilliams and Siegel 1997). To know what a firm’s stock price would have been in the absence of the event (in this case, the B2B eMarketplace announcement), the price is regressed against a market index to control for overall market effects. To calculate abnormal returns, the estimated coefficients from the market-model regression are used to compute the predicted value of the firm’s stock. For each security j, the following regression model is used to calculate abnormal returns at time t: cantly positive abnormal returns. This is because acquiring firms typically must pay a substantial premium for target firms, which is often viewed unfavorably by the financial market (Roll 1988). If acquiring firms produce significantly positive abnormal returns, this might indicate that investors expect these firms to generate synergies via economies of scale or scope by reducing costs and eliminating redundancies. This outcome could also indicate that investors see potential gains from providing a larger selection of products and services or the possibility of enhancing market power by reducing price competition. Hypothesis 4: Higher Returns to Firms Forming Alliances When Announcing New eMarketplaces We expect significant positive abnormal stock returns to e-commerce technology firms announcing B2B eMarketplaces, whether they develop the marketplaces by themselves or with another firm. However, we expect that alliances with other companies —whether they are computer industry giants, competitors, or Old Economy leaders—will result in higher abnormal returns than creating an eMarketplace alone would.3 This would occur if investors foresee potential synergies and competitive advantages from allying with firms that have similar objectives. Alliances also create more support and depth for the eMarketplace. If an e-commerce technology firm that announces plans to develop B2B eMarketplaces by itself generates higher abnormal returns than firms that plan to align with others, this might indicate investors foresee potential problems with the proposed alliance. (1) where ARjt is the abnormal return for stock j at time t ; Rjt is the actual return for stock j at time t ; α j is the ordinary least squares (OLS) estimate of the intercept of the market-model regression; Rmt is the return to the market at time t, as approximated by Standard & Poor’s 500 stock market index; and βj is the OLS estimate of the slope of the coefficient in the market-model regression.4 The parameters α j and βj are estimated from the market model as follows: (2) Rjt = α j + βj Rmt + εjt , where εjt is the residual. Daily returns for individual-firm stock prices and the market index are from the Center for Research in Securities Prices database. The date of the event (announcement) is t = 0, the market model is estimated over the period from t = –165 to t = –15 days relative to the event date, and the event window is from t = –1 to t = +1. Once the market model is estimated, the resulting estimated values for α j and βj are used in Equation 1 with data for Rjt and Rmt to calculate the abnormal returns (ARs) over the event window for each e-commerce technology firm. Because the event date is known, a short window is used (Armitage 1995). In addition to the abnormal returns computed for the day before the announcement (t = –1), the day of the announcement (t = 0), and the day following the announcement (t = +1), we also compute cumulative average abnormal returns (CARs) for the periods from t = –1 to t = 0 and from t = –1 to t = +1. Dyckman, Philbrick, and Stephan (1984) find that two- and three-day METHODOLOGY AND DATA Event-study methodology is a forwardlooking approach that focuses on identifying abnormal returns to firms from a specific event. If investors react favorably to an event, positive abnormal stock returns around the event date would be expected. Consequently, abnormal returns provide a means of assessing an initiative’s impact on a firm’s future profitability. Event-study methodology is based on the efficient markets hypothesis (Fama et al. 1969) — that is, as new information becomes available, it is fully taken into consideration by investors assessing its current and future impact. The new assessment results in stock price changes that reflect the discounted value of current and future firm performance. Significant positive or negative stock price changes can ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 ARjt = Rjt – (αj + βj Rmt ), 15 Table 1 B2B eMarketplace Announcements, July 1999 – March 2000 Announcement date Type of B2B eMarketplace E-commerce technology provider(s) 7/28/1999 Horizontal Oracle Oracle announces online B2B eMarketplace 10/11/1999 Horizontal i2 Technologies i2 announces new business eMarketplace 10/21/1999 Vertical VerticalNet VerticalNet and IBM to create eMarketplace 11/02/1999 Vertical Oracle Ford and Oracle to create B2B eMarketplace 11/02/1999 Vertical Commerce One General Motors joins Commerce One to create B2B eMarketplace 11/05/1999 Horizontal Commerce One Commerce One acquires CommerceBid.com 11/08/1999 Horizontal Grainger Grainger.com debuts B2B eMarketplace 11/15/1999 Horizontal Ariba Ariba acquires TradingDynamics 12/16/1999 Horizontal Ariba Ariba acquires TRADEX 12/21/1999 Announcement Vertical i2 Technologies i2 announces eMarketplace for high-tech companies 1/10/2000 Horizontal Ariba EDS and Ariba to create group of B2B eMarkets 1/13/2000 Vertical Commerce One Shell and Commerce One announce joint venture to build energy industry eMarketplace 1/17/2000 Horizontal i2 Technologies i2 announces B2B eMarketplace for consumer goods and retail companies 1/19/2000 Vertical i2 Technologies, Commerce One General Motors and Commerce One add i2’s B2B supply chain services to eMarketplace 1/19/2000 Vertical Ariba Chevron and Ariba unveil eMarketplace for energy industry 1/20/2000 Vertical VerticalNet VerticalNet and Microsoft join to accelerate B2B commerce on the Internet 2/09/2000 Vertical Ariba Azurix introduces Ariba-powered eMarketplace for water buyers and sellers 2/14/2000 Vertical i2 Technologies United Technologies, Honeywell, and i2 to create eMarketplace for aerospace industry 2/17/2000 Horizontal Commerce One Citigroup and Commerce One announce plan to build eMarketplace 2/23/2000 Vertical i2 Technologies Toyota and i2 form eMarketplace for automotive replacement parts market 2/23/2000 Vertical Commerce One BellSouth and Commerce One launch joint venture to build eMarketplaces for telecommunications industry 2/25/2000 Vertical Commerce One, Oracle Ford, General Motors, and DaimlerChrysler create eMarketplace 2/28/2000 Vertical Oracle Sears, Carrefour, and Oracle to form eMarketplace for retail industry 2/28/2000 Vertical i2 Technologies i2 creates eMarketplace for softgoods industry 2/29/2000 Horizontal i2 Technologies i2 creates eMarketplace for logistics industry 3/01/2000 Vertical Ariba Sabre and Ariba announce B2B eMarketplace for travel and transportation industry 3/08/2000 Horizontal Ariba, i2 Technologies IBM, Ariba, and i2 form alliance to accelerate global adoption and benefits of B2B e-commerce 3/08/2000 Vertical VerticalNet VerticalNet to acquire Tradeum to expand B2B e-commerce platform 3/08/2000 Vertical Oracle Chevron, McLane, and Oracle to form eMarketplace for convenience store industry 3/13/2000 Vertical Oracle Oracle and fibermarket.com announce B2B eMarketplace for global forest products industry 3/14/2000 Vertical Ariba Cargill and Ariba announce eMarketplace for food and beverage industries 3/15/2000 Horizontal FreeMarkets FreeMarkets announces agreement to acquire iMark.com 3/22/2000 Horizontal FreeMarkets FreeMarkets announces agreement to acquire Surplus Record and SR Auction 3/22/2000 Horizontal i2 Technologies i2 announces eMarketplace for aftermarket parts and service management 3/23/2000* Vertical Commerce One Boeing, Lockheed Martin, BAE Systems, and Raytheon to create B2B eMarketplace for aerospace and defense industry 3/26/2000 Vertical Oracle Hutchison and Oracle announce B2B eMarketplace for transportation service industry * The press release was dated March 28, 2000, but on March 23, several newspapers reported these firms’ plans to form a B2B exchange. 16 FEDERAL RESERVE BANK OF DALLAS Table 2 Average Abnormal Returns by Type of eMarketplace event windows are preferable to one-day windows because of rumors of the announcement and insider information. These calculations indicate whether the returns to the shareholders of the e-commerce technology providers are abnormal compared with those expected from general market movements. The market model in Equation 2 breaks down the total return on stock j into two components: one that reflects general market movements and one that reflects price variations caused by firm-specific events. Deducting (αj + βj Rmt ) from Rjt (as shown in Equation 1) neutralizes the effect of general market movements but does not neutralize firm-specific price variations caused by events other than the eMarketplace announcement. To neutralize these firm-specific variations, the cross-sectional average of the abnormal returns for the total sample of stocks for each period is computed. For a sample of n stocks, the mean abnormal return for each day t is (3 ) MARt = 1 n CAR (−1, t 1 ) = Overall Horizontal eMarketplace Vertical eMarketplace Day before announcement (t = –1) 1.25% (1.893) 1.33% (1.228) 1.19% (1.442) Day of announcement (t = 0) 4.05%*** (4.401) 3.55%** (2.168) 4.36%*** (3.896) Day after announcement (t = +1) 2.08% (2.151) 1.18% ( .862) 2.64% (2.060) Two-day event window (t = –1 to t = 0) 5.30%*** (4.450) 4.88%** (2.401) 5.56%*** (3.774) Three-day event window (t = –1 to t = +1) 7.38%*** (4.875) 6.06%** (2.459) 8.20%** (4.271) Number of firms 39 *** Significant at the 0.01 level. ** Significant at the 0.05 level. NOTE: t statistics in parentheses. stock price events and cover the period from July 28, 1999, through March 26, 2000. n ∑ AR jt , j =1 SHAREHOLDER RETURNS where t = –1, 0, +1. The cross-sectional average neutralizes firm-specific price variations unrelated to the B2B eMarketplace announcements. Hence, the expected value of MARt is zero in the absence of abnormal returns due to B2B eMarketplace announcements. The final calculation of abnormal returns is to compute cumulative average abnormal returns from day t = –1 to t = 0 and from t = –1 to t = +1, using the formula (4 ) Event period Table 2 presents a summary of the average abnormal returns for all the firms in our sample, as well as a breakdown by whether they are announcing horizontal or vertical B2B eMarketplaces.6 Average abnormal returns for five event periods are reported: the day before the announcement, the day of the announcement, the day following the announcement, cumulative returns from the day before the announcement to the day of the announcement, and cumulative returns from the day before the announcement to the day following the announcement. Also reported are the t statistics and significance levels that test whether the returns differ significantly from zero. All of the announcements taken together produced positive ARs to shareholders. Most noteworthy are the two- and three-day CARs. The two-day CAR is 5.3 percent, and the threeday CAR is 7.38 percent, with twenty-six of the thirty-nine firms receiving positive abnormal returns during both event windows. Abnormal returns to shareholders are significantly different from zero for both windows at the 0.01 level. This result strongly supports Hypothesis 1, which postulates that investors react favorably to firms announcing B2B eMarketplaces. When the announcements are segregated by the type of B2B eMarketplace, we find both horizontal and vertical eMarketplace announcements result in significantly positive CARs. t1 ∑ MARt , t = −1 where t 1 = (0, +1) and CAR (–1,t 1) is the cumulative average abnormal return for the sample of n stocks over the event period interval from t = –1 to t 1. The expected value of CAR is zero in the absence of abnormal performance. Tests of significance are discussed in the box on page 20. Table 1 lists the announcements in our sample. We define an event as the release of a firm’s B2B eMarketplace announcement through the media.5 Our events are derived from a list of defining events in B2B by Phillips and Meeker (2000). The events in this report include announcements of B2B e-commerce IPOs, eMarketplaces, acquisitions, joint ventures, and alliances. We include all announcements that involved an established, publicly traded e-commerce technology provider except those that announced an IPO. The resulting thirty-six announcements include thirty-nine individual ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 17 15 24 Horizontal B2B eMarketplace announcements result in a two-day CAR of 4.88 percent and a three-day CAR of 6.06 percent. Ten of the fifteen firms making a horizontal B2B eMarketplace announcement received a positive return in both windows. For vertical B2B eMarketplace announcements, the two-day CAR is 5.56 percent and the three-day CAR is 8.2 percent. Sixteen of the twenty-four firms making a vertical B2B eMarketplace announcement received a positive return in both windows. While the returns are higher for firms making vertical eMarketplace announcements, they do not statistically differ from the returns for firms making horizontal eMarketplace announcements. These results lend weak support to Hypothesis 2, in that investors prefer vertical over horizontal eMarketplaces, although the positive abnormal returns for both groups are not statistically different. In any case, investors seem to view both types of eMarketplaces favorably and anticipate increased efficiencies and reduced costs that will produce future benefits. Table 3 shows the results of dividing the B2B eMarketplace announcements by the type of partner, if any, the e-commerce technology provider announced it would be working with. For the six firms that announced they were acquiring another e-commerce company, the two- and three-day CARs are 1.15 percent and 4.1 percent, respectively, neither of which is significantly different from zero. Interestingly, the AR for the day before the announcement is –3.24 percent, with five of the six firms experiencing negative abnormal returns. This negative average abnormal return is quickly erased, however, as five of the six firms received positive ARs on the day of the announcement, for an average event-day return of 4.39 percent. In conformance with Hypothesis 3, these results suggest investors see acquisitions of other technology providers as neither a positive nor a negative. For the nine firms announcing plans to develop B2B eMarketplaces on their own, the two- and three-day CARs are 3.95 percent and 4.84 percent, respectively. Neither CAR is significantly different from zero. However, it is interesting that the first five announcements (those prior to February 2000) resulted in significantly positive CARs, whereas the last four announcements (those after January 2000) resulted in significantly negative CARs. The two- and threeday CARs for the first five announcements are 9.38 percent and 15.91 percent, respectively. For the last four announcements, the two- and three-day event-window returns are –2.83 percent and –9 percent, respectively. These results suggest the possibility of a first-mover advan- Table 3 Average Abnormal Returns by Type of Partner Event period Day before announcement (t = –1) Acquisition Alone Alliance computer Alliance competitor Alliance Old Economy – 3.24%* (–1.025) .70% (.723) 4.82%** (1.774) .47% (–.062) 2.39% (2.148) Day of announcement (t = 0) 4.39%* (1.488) 3.25% (1.831) 7.40%*** (2.551) 11.09%*** (4.331) 1.40% (1.027) Day after announcement (t = +1) 2.96% (1.016) .89% (.925) Two-day event window (t = –1 to t = 0) 1.15% (.328) 3.95% (1.806) 12.22%*** (3.058) 11.55%*** (3.019) 3.79% (2.245) Three-day event window (t = –1 to t = +1) 4.10% (.854) 4.84% (2.008) 18.46%** (3.743) 14.43%** (2.911) 4.64% (2.101) 6 9 5 Number of firms 6.25% (2.158) 2.88% (.773) 4 .84% (.464) 15 *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. NOTE: t statistics in parentheses. 18 FEDERAL RESERVE BANK OF DALLAS Table 4 Average Abnormal Returns by the Timing of the Announcement Event period 1999 January 2000 February 2000 March 2000 Day before announcement (t = –1) 3.31% (2.114) 2.33% (1.143) 2.23% (1.699) –1.92% (–.942) Day of announcement (t = 0) 5.95%*** (3.218) 4.87%** (2.121) 3.94% (2.128) 2.08% (1.433) Day after announcement (t = +1) 6.78%*** (3.526) 7.55% (3.480) –2.40% (–1.147) –1.29% (–.952) Two-day event window (t = –1 to t = 0) 9.26%*** (3.770) 7.20%* (2.309) 6.16%** (2.706) Three-day event window (t = –1 to t = +1) 16.04%*** (5.114) 14.75%* (3.894) 3.76% (1.548) –1.13% (– .266) 7 10 12 Number of firms 10 .16% (.348) *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level. NOTE: t statistics in parentheses. from alliances with competitors and computer industry leaders. In fact, returns for alliances with competitors and computer industry leaders are significantly higher than those made through partnerships with Old Economy leaders. Investors may react more favorably to competitor and computer industry partnerships because of the perceived synergies, name recognition, and increased operating efficiencies created by such alliances. Also, these partnerships mean neither party has to expend additional resources to compete for B2B eMarketplace business. Table 4 shows the results from examining the timing of the announcements. The earlier announcements received ARs and CARs much higher than announcements made later during our sample period. The 1999 announcements have two- and three-day CARs of 9.26 percent and 16.04 percent, respectively. These returns are significantly different from zero at the 0.01 level, and all ten firms making these announcements experienced positive abnormal returns over the three-day event window. For the seven firms making announcements in January 2000, six experienced twoand three-day positive CARs. As a whole, these firms have two- and three-day CARs of 7.2 percent and 14.75 percent, respectively, which are both significant at the 0.10 level. But the returns get lower later in the sample period. For February 2000, six of the ten firms received positive tage to firms that position themselves as B2B eMarketplace leaders. (This is discussed with respect to the entire sample below.) For firms announcing alliances to develop B2B eMarketplaces with other firms, the returns are positive and mostly significant, as Hypothesis 4 suggests. The two- and three-day CARs for the five firms announcing an alliance to develop a B2B eMarketplace with a large and established computer industry business are 12.22 percent and 18.46 percent, respectively. For the four firms announcing an alliance with another e-commerce technology provider, the two- and three-day event-window returns are 11.55 percent and 14.43 percent, respectively. Taken together, such alliances result in two- and threeday CARs of 11.92 percent and 16.67 percent, respectively, both of which are significant at the 0.01 level.7 The firms announcing the development of B2B eMarketplaces with Old Economy leaders also received positive average abnormal returns, but they are not significantly different from zero. The two- and three-day CARs for the fifteen firms announcing B2B eMarketplaces with these leaders are 3.79 percent and 4.64 percent, respectively. Nine of the fifteen firms received positive returns over the two event windows. It appears investors view alliances with industry leaders favorably. However, the returns are fairly low (and not significantly different from zero), especially when compared with the returns ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 19 Tests of Significance To test the significance of MARt , the average standardized abnormal return is estimated using the following statistic, as described in Dodd and Warner (1983): Only four of the twelve firms making B2B e-commerce announcements in March have positive CARs over the event windows. Overall, for the two- and three-day event periods, CARs are 0.16 percent and –1.13 percent, respectively, neither of which is significantly different from zero. Thus, before February 2000, firms making B2B eMarketplace announcements generally received significantly positive abnormal returns. For the firms in our sample, the two- and threeday CARs are 8.41 percent and 15.51 percent, respectively, both of which are significantly different from zero at the 0.01 level. After January 2000, however, investors reacted less favorably to B2B eMarketplace announcements. While the CARs are still positive, they are not significantly different from zero. The two-day CAR for firms making announcements in February and March 2000 is 2.89 percent. The three-day CAR is 1.09 percent. These results suggest the possibility of significant first-mover advantages.8 However, this conclusion must be viewed with caution because of the extreme volatility of technology stocks at the end of our sample period. The Internet sector in general has experienced substantial volatility, making those firms particularly sensitive to announcements of B2B e-commerce initiatives. 1 n AR jt SARt = ∑ , n j =1 s jt (B.1) where sjt is the estimated standard deviation of the abnormal returns for stock j in event period t and is computed by 1 s jt = s 2j 1+ + T (B.2) (R − R ) , ∑ (R − R ) 2 mt m T k =1 2 mk m 2 where s j is security j ’s residual variance from the market-model regression, T is the number of days in the estimation period (150), Rmt is the rate of return on the market – index for day t of the event period, R m is the mean rate of return on the market index during the estimation period, and Rmk is the rate of return on the market index for day k of the estimation period. As shown in Equation B.2, the standard error of the forecast for the event period, sjt , involves a slight adjustment from the standard error of the estimate, sj . This adjustment reflects the deviations of the independent variables in the estimation period from the values employed in the original regression and are typically close to 1 (Peterson 1989). Assuming cross-sectional independence, SARt approaches a normal distribution and the test statistic is unit normal: (B.3) t statistic = nSARt . This test statistic is used to test the hypothesis that the average abnormal returns for a given sample of stocks (MARt ) are significantly different from zero at various levels for each of the event periods t = –1, 0, +1. A similar test statistic is employed to test the hypothesis that the cumulative average abnormal returns (CAR ) are significantly different from zero. In this case, the relevant test statistic must be modified to fit the particular interval over which the returns are calculated, as follows: (B.4) t statistic = t1 n ∑ SARt , (t1 + 2) t = −1 CONCLUSIONS where t 1 = (0, +1) to compute cumulative average abnormal returns over the two-day event period from t = –1 to t = 0 and the three-day event period from t = –1 to t = +1. To test whether abnormal returns from two groups of stocks statistically differ, we use (B.5) t statistic = CAR1 − CAR2 2 1 T ∑ (Z − Z ) (T − 2) t =1 t Table 5 summarizes our hypotheses and findings. Overall, we find that shareholders view B2B eMarketplace announcements favorably. These initiatives promised increased efficiencies and reduced costs from streamlining operations up, down, and across industry supply chains. Firms making B2B eMarketplace announcements received significantly positive average abnormal returns around the date of the announcement, suggesting that B2B e-commerce strategies create significant future benefits. We also find significant positive average abnormal returns associated with both vertical and horizontal eMarketplace announcements. The returns to firms making vertical eMarketplace announcements are slightly higher than those to firms making horizontal announcements, but they do not differ significantly. It appears investors foresee gains from both types of eMarketplaces, whether they create efficiencies up and down the supply chain or at a single point across the supply chains of different industries. , where CAR1 is the cumulative average abnormal return for one group of stocks, CAR2 is the cumulative average abnormal return for another group of stocks, T is the number of days in the estimation period (150), Z t is the difference in returns between CAR1 and CAR2 at time t, and Z is the average difference in returns between CAR1 and CAR2 over the estimation period. abnormal returns over the event windows. Overall, the ten firms have two- and three-day CARs of 6.16 percent and 3.76 percent, respectively. Only the two-day event-window return is significant at the 0.05 level. In March 2000, investors began to dump technology stocks and firms making B2B eMarketplace announcements no longer experienced significantly positive abnormal returns. 20 FEDERAL RESERVE BANK OF DALLAS Table 5 Summary of Hypotheses and Findings Hypothesis Findings 1. For e-commerce firms making B2B eMarketplace announcements, the abnormal returns should be positive. Strong statistical support. (Abnormal returns are positive and significantly different from zero.) 2. Abnormal returns to firms announcing vertical eMarketplaces should be greater than those to firms announcing horizontal eMarketplaces. Weak support. (The two groups do not statistically differ, although both groups have positive abnormal returns significantly different from zero.) 3. Firms announcing plans for the acquisition of e-commerce technology providers should experience insignificant abnormal returns. Support. (Abnormal returns are positive but not significantly different from zero.) 4. Firms announcing alliances to develop B2B eMarketplaces should receive higher abnormal returns than those creating B2B eMarketplaces on their own. Strong statistical support. (Abnormal returns are highest for alliances with competitors and computer industry giants. These returns are significantly different from the abnormal returns for alliances with Old Economy leaders and going it alone.) When subdividing the data by the type of partner the e-commerce provider aligns with, we find that investors reward firms the most when they partner with a competitor, especially a large computer-industry giant like IBM, Microsoft, or EDS. Abnormal returns from these announcements are more than three times higher than those from announcements of plans to develop a B2B eMarketplace alone or with an Old Economy leader. This is noteworthy, as it suggests that shareholders value alliances between e-commerce technology providers more than solo B2B e-commerce initiatives or those undertaken with an Old Economy leader. One explanation for this is that when e-commerce technology firms combine resources, there is one less competitor. When e-commerce technology providers develop B2B e-commerce strategies on their own or with an industry leader, competition is not lessened and investors view the news less favorably. Finally, we find that announcements made earlier in our sample period had much higher average abnormal returns than announcements made closer to the end of the period. One possible explanation is a first-mover advantage: investors may tend to reward firms that position themselves as leaders and pioneers in B2B e-commerce. Another explanation may be the sample period used. During 1999 and early 2000, technology stocks were the darlings of Wall Street. ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 But in late February and early March 2000, investors began to dump them because of increasing fear that these stocks were overvalued. Because our sample period covers this unsettled time, returns may be somewhat distorted. NOTES 1 2 3 4 5 21 The authors would like to thank John Duca, Kenneth Robinson, and Mark Wynne for their extremely insightful comments and suggestions, which improved the quality of this article. See www.forrester.com/ER/Press/ForrFind/ 0,1768,0,00.html. McWilliams and Siegel (1997) outline procedures for using the event-study framework. While we have no information about whether alliances with competitors, computer industry giants, or Old Economy leaders will result in higher abnormal returns, we can test for significant statistical differences between groups of stocks. Most published event studies use the S&P 500 index to estimate the parameters for calculating abnormal returns. Because information about B2B eMarketplace announcements may have leaked prior to the issuance of press releases, a search of major news and business publications using the Dow Jones Interactive News Service was conducted to see if any information was anticipated. In one case — Commerce One’s involvement with the creation of a B2B exchange for 6 7 8 the aerospace and defense industry — there were several news reports five days prior to the March 28, 2000, press release. As a result, the event window used for this announcement is based on a March 23 announcement date. All the abnormal returns this article reports are based on using the S&P 500 stock market index in the market-model regressions. These results are, however, qualitatively robust when using either the Wilshire 5000 stock market index or the Nasdaq composite stock index. These other indexes were used to test whether the technology-sector stock correction of early 2000 and potential investor sentiment swings affected relative returns. Because of the low number of observations, these results must be viewed with caution. Nevertheless, taken together, alliances with computer industry giants and competitors in B2B e-commerce initiatives do generate statistically significant positive abnormal returns. Further, the statistical significance for these results is not driven by any particularly large return for just one firm. Milbourn, Boot, and Thakor (1999) discuss why shareholders should benefit from scope-expanding early entry. Dyckman, Thomas, Donna Philbrick, and Jens Stephan (1984), “A Comparison of Event Study Methodologies Using Daily Stock Returns: A Simulation Approach,” Journal of Accounting Research 22 (Supplement): 1– 30. Fama, Eugene F., Lawrence Fisher, Michael C. Jensen, and Richard Roll (1969), “The Adjustment of Stock Prices to New Information,” International Economic Review 10 (February): 1– 21. McWilliams, Abagail, and Donald Siegel (1997), “Event Studies in Management Research: Theoretical and Empirical Issues,” Academy of Management Journal 40 (June): 626 – 57. Milbourn, Todd T., Arnoud W. A. Boot, and Anjan V. Thakor (1999), “Megamergers and Expanded Scope: Theories of Bank Size and Activity Diversity,” Journal of Banking & Finance 23 (February): 195 – 214. Peterson, Pamela P. (1989), “Event Studies: A Review of Issues and Methodology,” Quarterly Journal of Business and Economics 28 (Summer): 36 – 66. Phillips, Charles, and Mary Meeker (2000), “The B2B Internet Report: Collaborative Commerce” (Morgan Stanley Dean Witter Equity Research , April). REFERENCES Armitage, Seth (1995), “Event Study Methods and Evidence on Their Performance,” Journal of Economic Surveys 9 (March): 25 – 52. Roll, Richard (1988), “Empirical Evidence on Takeover Activity and Shareholder Wealth,” in Knights, Raiders, and Targets, ed. John C. Coffee, Jr., Louis Lowenstein, and Susan Rose-Ackerman (New York: Oxford University Press), 241– 52. Brookes, Martin, and Zaki Wahhaj (2000), “The ‘New’ Global Economy — Part II: B2B and the Internet,” Global Economics Weekly (Goldman Sachs, February 9, 3 –13). Subramani, Mani R., and Eric Walden (1999), “The Dot Com Effect: The Impact of E-Commerce Announcements on the Market Value of Firms,” Working Paper no. 99-02, Management Information Systems Research Center, University of Minnesota (September). Dodd, Peter, and Jerold B. Warner (1983), “On Corporate Governance: A Study of Proxy Contests,” Journal of Financial Economics 11 (April): 401– 38. 22 FEDERAL RESERVE BANK OF DALLAS Long-standing restrictions on where banks could locate their operations began to erode more than twenty years ago and were mostly eliminated with the passage of interstate branching and banking legislation in 1994. As a result, the U.S. banking industry experienced substantial consolidation.1 While this has likely contributed to the industry’s robust performance of late, it could have important consequences for banks’ small business lending. Large, complex banking organizations are traditionally not seen as significant sources of financing for small businesses. On the other hand, the banking industry, like other segments of the economy, is an active participant in the information and communications revolution. Credit scoring models lower the costs of extending credit and improve access to small business financing, especially for larger banks. So, while consolidation could reduce small business lending, technological advances may increase the flow of small business credit. In this article we summarize some of the ways consolidation and advances in technology may affect small business lending. We then examine the available data on small business loans over the period 1994 through 1999 to detect any changes in small business lending patterns and their possible consequences. Although small business lending has increased since 1994, we find that the share of total lending devoted to small business loans has declined. However, the aggregate numbers conceal some important trends across organizations of different sizes. We find evidence that large banks are increasing their presence in the smallest segment of the small business loan market and that the average loan size has declined, especially at larger institutions. Larger banks also appear to have narrowed the gap relative to small banks in their focus on the smallest loans. These trends are consistent with the view that new information technology, most notably credit scoring, is changing the structure of small business lending. Consolidation, Technology, and the Changing Structure of Banks’ Small Business Lending David P. Ely and Kenneth J. Robinson L arge banks are increasing their presence in small business lending, especially in the smallest-sized loan category, reflecting both consolidation trends and greater use of new information technology. WHY LOOK AT BANKS’ SMALL BUSINESS LENDING? Small business lending by banks has been the subject of extensive theoretical and empirical investigation. This reflects the value of small businesses to the U.S. economy and the potentially unique role of banks in small business lending. Small businesses (those with fewer than 500 workers) employ 53 percent of the private nonfarm workforce and are responsible for David Ely is a professor of finance at San Diego State University. Kenneth Robinson is a senior economist and policy advisor in the Financial Industry Studies Department at the Federal Reserve Bank of Dallas. ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 23 51 percent of private gross domestic product. Small businesses are also responsible for a major portion of job creation. From 1990 through 1995, small businesses created more than three-fourths of all new jobs (U.S. Small Business Administration 1999). preexisting relationship, regardless of the length of that relationship. One of the premises of relationship lending is that larger, more complex banks might find the gathering and monitoring of information for nonstandard small business loans too expensive. To the extent that larger banks find it more costly than smaller ones to evaluate small business borrowers, larger banks would be expected to extend less small business credit than smaller banks. On the other hand, larger organizations may enjoy greater diversification and lower costs, which could serve to increase small business lending. Relationships Banks fill an important niche in financing small businesses. Small firms are more likely to obtain financing from a commercial bank than from other sources, including depository and nondepository institutions (Cole and Wolken 1995). Small business lending is often viewed as idiosyncratic and relationship-based. It depends on collecting and analyzing detailed, proprietary information because public information on small firms is often lacking. Many small business loans are treated in the same manner as consumer loans because the creditworthiness of the firm’s owner—rather than the firm —is frequently a key factor in the lending decision. In contrast, ample public information is usually available about larger borrowers. The unique information requirements for small business loans may give smaller, more locally based banks an advantage in extending these types of loans (Berger and Udell 1996). Lending relationships between banks and firms can reduce the monitoring and oversight costs associated with small business loans. Theoretical models of relationship lending can be found in Greenbaum, Kanatas, and Venezia (1989), Petersen and Rajan (1994, 1995), and Boot and Thakor (2000). These articles stress the presence of information asymmetries between borrowers and lenders and how banking relationships can overcome the problems associated with providing small business credit. For an overview of issues involved in studies of relationship banking, see Boot (2000). An extensive number of empirical investigations also support the view that banking relationships generate information about bank customers and yield benefits, such as lower funding costs and increased availability of credit to customers. Petersen and Rajan (1994) find that the length of the borrower relationship affects the availability of small business lending but not the price. Berger and Udell (1995) use lines of credit to isolate relationship loans and find that firms having longer relationships with banks pay lower rates. They also report that a longer relationship decreases the likelihood that the lender will require collateral to secure a loan. Cole (1998) finds that banks are more likely to lend to firms with which they have a Consolidation and Small Business Lending Individual states began allowing out-ofstate-institutions to establish operations across state lines more than twenty years ago. This process culminated with the Riegle–Neal Interstate Banking and Branching Efficiency Act of 1994, authorizing interstate banking and branching nationwide. Proponents of the legislation pointed to the efficiency and cost-saving potential of freeing banks from geographic restrictions on their operations. However, as this legislation was likely to accelerate the consolidation trends already evident in the industry, it also raised concerns about the effects on banks’ small business lending. The number of U.S. banks peaked at over 14,400 in 1984 and currently stands at roughly 8,500. In 1990, the top ten banking organizations accounted for 25.5 percent of U.S. banking assets; by 1999, they held 46.2 percent. Several studies have looked at the effects of bank mergers and acquisitions (M&As) on small business lending. Berger et al. (1998) examine both the static and dynamic impacts of mergers and acquisitions, using more than six thousand M&As from the late 1970s to the early 1990s. The static effect is the predicted change in lending from simply combining the balance sheets of the participating banks. Using results from a model of lending activity, Berger et al. find that the small business lending predicted for the combined bank is less than that of the two (or more) pre-M&A banks. However, when dynamic effects are considered, such as changes in the consolidated institution’s lending focus or the response by other banks in the same market, they find that the static declines in small business lending are mostly offset. Peek and Rosengren (1998a,b) present evidence that acquiring banks recast their targets in their own image. But, because most mergers are between two or more small banks and because 24 FEDERAL RESERVE BANK OF DALLAS acquirers are likely to have larger small business loan portfolios than their targets, any concerns about the effects on small business lending from mergers and acquisitions may be unwarranted. Strahan and Weston (1998) point out that smaller banks may not realize lower costs relative to larger banks if size-related diversification advantages offset organizational diseconomies in business lending. Their finding that consolidation among small banks increases lending to small businesses while other types of mergers have little effect is consistent with important diversification effects that come with size. Jayaratne and Wolken (1999) provide evidence that small banks have no cost advantage in making small business loans. These authors also find that young firms and firms with poor credit histories are as likely to have a line of credit from a large bank as a small bank. Because loans to these firms usually require closer scrutiny, this result is also consistent with no cost advantage for smaller banks to engage in small business lending. As these studies show, fears that the ongoing consolidation of the U.S. banking industry may diminish small business lending are generally not supported by the evidence. Also, the possibility that larger banks find themselves at a cost disadvantage in extending small business loans is open to question. One factor that may reduce the costs of small business lending is banks’ increasing use of advances in information technology. Loan Officer Opinion Survey on Bank Lending Practices contained several questions on the use of credit scoring in small business lending. Thirty-eight of the fifty-four banks responding indicated they used credit scoring models in extending small business loans. Larger banks (those with assets greater than $15 billion) were more likely to use credit scoring models than smaller banks (Federal Reserve Board 1997). To the extent that credit scoring reduces large banks’ costs of extending small business loans, it would be expected to narrow the gap between large and small banks’ emphasis on small business lending. Levonian (1997) reports that lenders view business loans below a certain size as analogous to consumer loans, making these smaller loans attractive candidates for credit scoring models. An article in the Wall Street Journal noted that over the past five years banks have turned to scoring models in their small business lending and that about 90 percent of big banks use a credit scoring model known as the Fair Isaac system (Prager 1999). A recent survey of credit scoring found that the median loan size scored in 1998 was $150,000, up from $100,000 in 1997 (American Banker 1998). When looking at mergers, Peek and Rosengren (1998b) find that the largest acquiring banks (those with assets greater than $1 billion) show an increase in their portfolio shares of small business loans with original amounts of $100,000 or less. On the other hand, smallersized acquirers record decreases in their portfolio shares of loans of $100,000 or less. Peek and Rosengren argue that larger banks’ investment in information technology enables them to use credit scoring models to service small business loans at lower costs. To obtain some insights into what role consolidation and technological advances play in the small business loan market, we examine the available data on small business lending. Technology and Small Business Lending Banks’ growing presence online may be the industry’s most obvious embrace of the new economy (Couch and Parker 2000). But banks are also adopting recent advances in information technology and computing —particularly credit scoring models —to their small business lending decisions. In the past, banks relied on personal credit histories and their own judgment in deciding whether to extend credit. Credit scoring uses sophisticated statistical models to evaluate potential borrowers, isolating characteristics that best predict riskiness. These models then produce scores that banks can use to rank their borrowers in terms of risk. Originally used for credit card and other consumer loans, credit scoring is now making significant inroads into mortgage origination (Mester 1997). Even more important for our purposes, an increasing number of banks are adopting credit scoring models for use in small business lending. The Federal Reserve’s January 1997 Senior ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 SMALL BUSINESS LENDING: SOME PATTERNS AND IMPLICATIONS Banks’ Reports of Condition and Income, or call reports, contain data on small business loans based on the size of the loan, which serves as a proxy for the size of the borrower. Three categories are identified: loans with original amounts of $100,000 or less; those with original amounts of $100,001–$250,000; and those with original amounts of $250,001– $1 million. These data are collected yearly and appear in the second quarter call reports.2 25 Table 1 Banks’ Market Share by Bank Size, 1994– 99 (Percent of total loans) Asset size Less than $300 million $300 million to less than $1 billion $1 billion to less than $5 billion $5 billion and greater 1994 1995 1996 1997 1998 1999 12 7 9 72 11 6 8 74 11 6 8 75 11 7 8 75 10 7 8 76 9 6 7 77 NOTE: Assets in 1999 dollars. SOURCE: Report of Condition and Income. In the following sections, we investigate trends in aggregate small business lending at U.S. banks. However, because aggregate data could conceal substantial variation in lending activity and focus, we also examine lending at banks of four different asset sizes: those with assets less than $300 million; those with assets of $300 million to less than $1 billion; those with assets of $1 billion to less than $5 billion; and those with assets of $5 billion and greater. All total asset and loan values are expressed in 1999 dollars, using the consumer price index to remove the effects of inflation. We use data at the organization level because intracompany transactions among banks that are part of a multibank holding company could make banklevel comparisons misleading (Strahan and Weston 1998). For example, Demsetz (1999) presents evidence that such banks are more likely to buy and sell loans than are independent banks or banks that are part of a one-bank holding company.3 1999. These data are consistent with the available evidence that consolidation of the U.S. banking industry has not led to large declines in small business lending. The portfolio shares of the three lending categories from 1994 through 1999 for all U.S. banks are displayed in Figure 2. We measure the shares using both total assets and total loans as the base. The overall asset share of small business lending has held steady at U.S. banks at close to 8 percent. However, when using total loans as the base, portfolio shares drop from 13 percent in 1994 to 11.9 percent in 1999, reflecting differences between small business loan growth and total loan growth at U.S. banks. While total small business lending has increased 20 percent since 1994 (in inflation-adjusted terms), it has not kept pace with overall loan growth, which increased 31 percent between 1994 and 1999. Consolidation Trends Table 1 indicates the consolidation trends in the industry over our analysis period by revealing banks’ market shares in terms of total loans. Only the largest bank size classification recorded an increase in market share. Banks with assets of $5 billion and above accounted for 72 percent of all loans in 1994; in 1999, this share increased to 77 percent. The other size classifications recorded declines in their loan shares, especially banks with assets of less than $300 million. Figure 1 shows how total small business lending and the three categories of loans have increased from 1994 through 1999.4 All categories have grown, with the exception of loans of $100,000 or less, which declined slightly in Figure 1 Small Business Lending, 1994 – 99 Index, 1994 = 100 130 $100,000 or less $100,001–$250,000 $250,001–$1 million Total 125 120 115 110 105 100 ’94 ’95 ’96 ’97 ’98 ’99 SOURCE: Report of Condition and Income. 26 FEDERAL RESERVE BANK OF DALLAS Figure 2 Small Business Lending Portfolio Shares in Percent of Assets and Loans, 1994 –99 Assets Loans Percent Percent 14 14 $250,001–$1 milllion 12 12 $100,001–$250,000 $100,000 or less 10 10 8 8 6 6 4 4 2 2 0 ’94 ’95 ’96 ’97 ’98 0 ’99 ’94 ’95 ’96 ’97 ’98 ’99 SOURCE: Report of Condition and Income. Figure 3 highlights two general trends in small business lending over 1994–99 for the four bank size groups. The first trend is the growing presence of the largest banks in the small business loan market. While the value of loans controlled by the smallest-sized banks fell from $95 billion to $92.3 billion, the holdings of banks with $5 billion or more in assets increased from $158.2 billion to $204.3 billion. The middle-sized banks also show modest gains in holdings of small business loans. The second trend of note in Figure 3 is the shifting focus of small business loans. For the smallest banks, the value of loans of $100,000 or less decreased from $48.7 billion in 1994 to $39.8 billion in 1999, but loans greater than $100,000 increased. For the middle-sized banks, business loans of less than $100,000 increased between 1994 and 1999 but less so than the loans in the ranges of $100,001– $250,000 and $250,001–$1 million. The largest banks’ holdings of loans of $100,000 or less expanded grad- Figure 3 Small Business Loans Outstanding by Bank Size, 1994 – 99 Billions of 1999 dollars 225 $250,001–$1 milllion 200 $100,001–$250,000 $100,000 or less 175 150 125 100 75 50 25 0 ’94 ’95 ’96 ’97 ’98 ’99 ’94 < $300 million ’95 ’96 ’97 ’98 ’99 $300 million–<$1 billion ’95 ’96 ’97 ’98 $1 billion–<$5 billion SOURCE: Report of Condition and Income. ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 ’94 27 ’99 ’94 ’95 ’96 ’97 $5 billion + ’98 ’99 Table 2 Banks’ Ratio of Small Business Loans to Total Loans, 1994 – 99 Loan size $100,000 or less 1994 1995 1996 1997 1998 1999 Change 1994 – 99 All banks Loan share (percent) Bank asset size $300 million $1 billion Less than to less than to less than $300 million $1 billion $5 billion $5 billion and greater 4.23 3.91 3.81 3.69 3.54 3.36 – 20.57 15.17 14.57 14.13 13.71 13.42 12.72 –16.15 7.59 7.55 7.66 7.40 7.19 6.78 –10.67 5.23 4.69 4.63 4.75 4.59 4.68 –10.52 1.90 1.85 1.88 1.82 1.82 1.81 – 4.73 $100,001 to $250,000 1994 1995 1996 1997 1998 1999 Change 1994 – 99 2.59 2.54 2.50 2.49 2.48 2.44 – 5.80 5.23 5.36 5.51 5.47 5.55 5.64 7.84 5.12 5.35 5.31 5.36 5.46 5.56 8.59 3.59 3.58 3.70 3.81 4.10 4.31 20.05 1.78 1.75 1.69 1.66 1.65 1.60 – 10.11 $250,001 to $1 million 1994 1995 1996 1997 1998 1999 Change 1994 – 99 6.21 6.06 5.97 5.96 5.94 6.10 – 1.77 9.13 9.71 9.94 10.07 10.40 11.15 22.12 11.97 12.06 12.19 12.48 12.53 13.22 10.44 9.03 8.96 9.19 9.06 9.85 10.80 19.60 4.82 4.65 4.52 4.46 4.37 4.43 – 8.09 All small business loans 1994 1995 1996 1997 1998 1999 Change 1994 – 99 13.03 12.51 12.28 12.14 11.96 11.89 – 8.75 29.52 29.64 29.58 29.24 29.37 29.51 –.03 24.68 24.96 25.17 25.24 25.18 25.55 3.52 17.85 17.23 17.52 17.62 18.54 19.79 10.87 8.50 8.26 8.09 7.94 7.84 7.85 – 7.65 NOTE: Assets in 1999 dollars. SOURCE: Report of Condition and Income. the business cycle, lending would tend to increase. This cyclical effect could cause various loan-to-asset ratios to rise independent of any change in banks’ lending focus. Our emphasis is on the smallest loans— those with outstanding amounts of $100,000 or less — because they should show the most noticeable effects of credit scoring, given their similarity to consumer loans. As shown in the table, for all banks combined, the aggregate ratio of loans of $100,000 or less to total loans has declined more than 20 percent since 1994. A downward trend is also evident for the various bank size categories, especially the smallest. Banks with less than $300 million in assets recorded an aggregate loan share of nearly 15.2 percent in 1994 but ually from $35.4 billion in 1994 to $47.2 billion in 1999. While these trends largely reflect U.S. banking industry consolidation, a more detailed examination of changes in lending shares also provides evidence that technology is changing the structure of small business lending. TECHNOLOGY’S ROLE Lending Focus Table 2 presents details on banks’ share of total loans devoted to small businesses for 1994–99. We calculate the aggregate shares for each bank size classification as a group and for all banks combined. We focus on loan shares rather than asset shares to account for cyclical effects on lending. That is, during an upswing in 28 FEDERAL RESERVE BANK OF DALLAS Table 3 Banks’ Average Small Business Loan Size by Bank Asset Size, 1994 –99 (Thousands of dollars) Loan size 1994 1995 1996 1997 1998 1999 p -value 31.1 53.1 30.1 53.0 29.5 53.2 29.6 53.8 30.0 55.6 29.8 57.9 .000 .000 33.1 108.9 31.4 98.8 31.0 95.9 30.8 96.8 30.0 97.4 29.2 97.7 .000 .007 34.9 126.9 32.7 127.9 31.3 123.6 29.7 112.9 29.9 115.5 28.3 110.3 .000 .192 30.5 115.0 29.3 112.0 28.3 104.1 26.2 98.7 26.0 99.2 25.4 104.5 .033 .593 Banks with assets less than $300 million $100,000 or less Total small business loans Banks with assets $300 million to less than $1 billion $100,000 or less Total small business loans Banks with assets $1 billion to less than $5 billion $100,000 or less Total small business loans Banks with assets $5 billion and greater $100,000 or less Total small business loans NOTE: Assets in 1999 dollars. SOURCE: Report of Condition and Income. 12.7 percent by 1999. The declines in loan shares at banks with assets greater than $300 million were substantially less. These results show the industry has reduced its lending focus on the smallest small business loans. This pattern could reflect less lending demand or less supply or a combination of both. As is clear from Table 2, though, the largest banks have reduced their emphasis on this loan category by a smaller amount than other organizations, which is consistent with the large banks’ adoption of credit scoring for small business loans. The cost reductions made possible by credit scoring may have partially offset the forces otherwise causing the industry to cut back on the smallest loans. The aggregate shares of loans of $100,001 through $1 million declined slightly for the entire set of banks. However, these loan shares tended to increase for all but the largest banks, indicating that the largest organizations’ emphasis on the larger loan categories has not kept pace with other banks. Finally, Table 2 shows that total small business lending, as a percent of total loans, declined for all banks as a group. However, the lending shares varied little at the smallest banks, increased a bit at the intermediate-sized institutions, and declined steadily at the largest banks. These trends in lending share suggest larger banks are changing their small business lending focus relative to the smaller banks. While the smallest loans declined in importance ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 for all organizations, this decline was substantially less at large banks. At the same time, the smaller banks increased their emphasis on the larger loans, while large banks decreased their focus on these loans. Consequently, the gap in lending emphasis between large and small banks in the smallest loan category has narrowed, consistent with credit scoring models becoming more important. An examination of trends in the average size of small business loans can also help ascertain the growing relevance of credit scoring. Average Loan Size Because inflation would likely push loans above the $100,000 cutoff, we would expect the share of the $100,000 or less category to fall over time, as it has, especially at the smaller banks. One explanation for a more moderate decline in this category at larger banks is that the adoption of credit scoring models has offset this inflation effect. To explore further the possible impact of such technology, we examine patterns of average loan size over time. Because the costs of offering smaller loans would tend to fall when credit scoring is employed, the average small business loan size should also fall. We calculate the average loan size for small business loans of $100,000 or less and for total small business loans. These results appear in Table 3. The dollar value of small business loans was divided by the number of loans outstanding to calcu- 29 Table 4 Banks’ Market Shares of Small Business Loans Outstanding by Asset Size, 1994 – 99 (Percent) Percent of total small business loan category Loan size 1994 1995 1996 1997 1998 1999 45 25 18 28 43 24 18 27 41 24 18 26 40 23 18 26 38 22 17 24 35 21 17 23 12 13 13 13 12 13 13 13 13 13 13 13 13 14 14 14 12 14 14 14 12 15 14 13 11 12 13 12 10 12 12 11 10 12 12 11 10 12 12 12 10 13 13 12 11 13 13 12 32 49 56 47 35 51 57 49 37 51 57 49 37 50 56 49 39 50 56 50 41 51 56 51 Banks with assets less than $300 million $100,000 or less $100,001– $250,000 $250,001– $1 million Total small business loans Banks with assets $300 million to less than $1 billion $100,000 or less $100,001– $250,000 $250,001– $1 million Total small business loans Banks with assets $1 billion to less than $5 billion $100,000 or less $100,001– $250,000 $250,001– $1 million Total small business loans Banks with assets $5 billion and greater $100,000 or less $100,001– $250,000 $250,001– $1 million Total small business loans NOTE: Assets in 1999 dollars. Shares might not total 100 due to rounding. SOURCE: Report of Condition and Income. tent with anecdotal evidence on the growing use of such technology. late average loan amounts for each bank. These amounts were then averaged across organizations. The p -values are significance levels for F -statistics that test for differences in means over time.5 The average loan size decreases for loans of $100,000 or less for all size groups. For banks with assets less than $300 million, the average loan of $100,000 or less fell from $31,132 in 1994 to $29,840 in 1999, while the average loan amount of all small business loans increased from $53,092 to $57,917. The decreases in the average loan amounts in the $100,000-and-under category are more substantial for organizations with over $1 billion in assets. For organizations with assets of $1 billion to less than $5 billion, the average loan amount dropped from $34,883 in 1994 to $28,331 in 1999. For organizations with over $5 billion in assets, the average loan amount dropped from $30,461 in 1994 to $25,421 in 1999. While we have not offered a direct test on the use of credit scoring models to lending activity, the declines in the average loan amounts, especially at larger banks, are consis- CONSOLIDATION AND TECHNOLOGY: TRENDS IN MARKET SHARE An increase in the overall presence of larger banks in small business lending would be expected from the combined effects of consolidation and the greater use of credit scoring models. And calculations of the market shares of small business loans indicate substantial shifts in shares, especially between large and small banks and particularly in the smallest lending category. Table 4 contains the market shares of the three categories of small business lending, classified by bank size, for 1994– 99. The market shares of small business loans for the two intermediate-sized groups — banks with assets of $300 million to less than $1 billion and those with assets of $1 billion to less than $5 billion — have remained constant. The biggest shifts in market share have occurred at the largest and smallest banks and in the smallest loan category. In 1994, 45 percent of small business loans 30 FEDERAL RESERVE BANK OF DALLAS of $100,000 or less were held by banks with assets less than $300 million, and 32 percent were held by banks with assets of $5 billion and above. In 1999, the proportion of these loans held by the smallest banks had fallen to 35 percent while the proportion held by the largest banks had increased to 41 percent.6 These data on market shares are consistent with both a more consolidated banking industry and a greater role for technology in lending. The industry is becoming more concentrated, with the largest banks controlling a greater portion of loans, including small business loans. Consequently, small businesses have become more dependent on large banking organizations. And the sharp increase in the largest banks’ market share of loans with outstanding amounts of $100,000 or less could indicate a greater role for credit scoring technology in small business lending. perfect indicator of small business lending by banks. Also, the data are only for loans secured by nonfarm, nonresidential property and commercial and industrial loans, while bank credit to small businesses can take other forms, such as personal lines of credit and home equity loans. Throughout our analysis, we exclude any bank that did not report total loans for the year in question. Also, we exclude banks under five years of age because younger banks may exhibit unusual patterns in small business lending. See Goldberg and White (1998) and DeYoung, Goldberg, and White (1999). 3 4 5 CONCLUSIONS Consolidation and technological change have characterized the U.S. banking industry recently. Each of these has important implications for small business lending, which has grown steadily since 1994. Large banks’ market share of small business lending has increased at the expense of smaller banks. The largest banks are making the greatest inroads with the smallest loans, and the gap in lending focus between large and small banks has narrowed in this area. Moreover, the average small business loan has declined in size. These results are consistent with advances in technology playing a larger role in small business lending. However, given the variety of regulatory and market forces that could also affect small business lending, we cannot be sure that advances in technology are the driving force behind the changes we have observed. Beyond consolidation, though, the role of technology is well worth considering when trying to understand changes in banks’ lending to small businesses. 6 REFERENCES American Banker (1998), “Small Business Lending: Securitization Likely Next Step for Banks in the Market,” February 6, 37. Berger, Allen N., and Gregory F. Udell (1995), “Relationship Lending and Lines of Credit in Small Firm Finance,” The Journal of Business 68 (July): 351– 81. NOTES 1 2 ——— (1996), “Universal Banking and the Future of Small Business Lending,” in Universal Banking: Financial System Design Reconsidered, ed. Anthony Saunders and Ingo Walter (Chicago: Irwin Professional Publishing), 558 – 627. The authors would like to thank Jeff Gunther and Bob Moore for helpful comments and suggestions. See Moore (1995) for evidence that the industry was consolidating before passage of interstate branching legislation. All data used in this paper are expressed at the organization level, but for simplicity, we use the terms bank and banking organization interchangeably. Because the data are based on the size of the loan rather than the size of the business, they are not a ECONOMIC AND FINANCIAL REVIEW FIRST QUARTER 2001 Our conclusions are unchanged, however, if we use bank-level data. Data on small business lending are available beginning in 1993. However, the 1993 data were found to contain errors. See Berger and Udell (1996, 576 – 77, footnote 6), and Peek and Rosengren (1998a, 802, footnote 3). Although revisions were made to these data, we begin our analysis in 1994 to avoid any possible data inconsistencies. These tests are based on ANOVA methods to detect any statistically significant differences in the means in at least one year. An alternative way to measure lending to small businesses is to examine market shares based on the number of loans outstanding rather than aggregate dollar amounts. This approach will provide additional insights if the average loan size is changing over time. When we measure market share based on the number of loans, the most notable changes in market share occur, once again, in the smallest lending category at both the largest and smallest institutions. In 1994, the smallest and largest banks both held 40 percent of the number of loans with original amounts of $100,000 or less. In 1999, the smallest banks held only 19 percent of these loans, while the largest held 55 percent. Berger, Allen N., Anthony Saunders, Joseph M. Saclise, and Gregory F. Udell (1998), “The Effects of Bank Mergers and Acquisitions on Small Business Lending,” Journal of Financial Economics 50 (November), 187– 229. 31 Boot, Arnoud W. A. (2000), “Relationship Banking: What Do We Know?” Journal of Financial Intermediation 9 (January): 7– 25. Jayaratne, Jith, and John Wolken (1999), “How Important Are Small Banks to Small Business Lending? New Evidence from a Survey of Small Firms,” Journal of Banking and Finance 23 (February): 427– 58. Boot, Arnoud W. A., and Anjan V. Thakor (2000), “Can Relationship Banking Survive Competition?” The Journal of Finance 55 (April): 679 – 713. Levonian, Mark E. (1997), “Changes in Small Business Lending in the West,” Federal Reserve Bank of San Francisco Economic Letter, January 24. Cole, Rebel A. (1998), “The Importance of Relationships to the Availability of Credit,” Journal of Banking and Finance 22 (August): 959 – 77. Mester, Loretta J. (1997), “What’s the Point of Credit Scoring?” Federal Reserve Bank of Philadelphia Business Review, September/October, 3 –16. Cole, Rebel A., and John D. Wolken (1995), “Financial Services Used by Small Businesses: Evidence from the 1993 National Survey of Small Business Finances,” Federal Reserve Bulletin 81, no. 7 (July): 629 – 67. Moore, Robert R. (1995), “Does Geographic Liberalization Really Hurt Small Banks?” Federal Reserve Bank of Dallas Financial Industry Studies, December, 1–12. Couch, Karen, and Donna L. Parker (2000), “ ‘Net Interest’ Grows as Banks Rush Online,” Federal Reserve Bank of Dallas Southwest Economy, Issue 2, March/April, 1– 5. Peek, Joe, and Eric S. Rosengren (1998a), “Bank Consolidation and Small Business Lending: It’s Not Just Bank Size That Matters,” Journal of Banking and Finance 22 (August), 799 – 819. Demsetz, Rebecca S. (1999), “Bank Loan Sales: A New Look at the Motivations for Secondary Market Activity,” Federal Reserve Bank of New York, Staff Reports, no. 69 (March). ——— (1998b), “The Evolution of Bank Lending to Small Business,” Federal Reserve Bank of Boston New England Economic Review, March/April, 27– 36. DeYoung, Robert, Lawrence G. Goldberg, and Lawrence J. White (1999), “Youth, Adolescence, and Maturity of Banks: Credit Availability to Small Business in an Era of Banking Consolidation,” Journal of Banking and Finance 23 (February): 463 – 92. Petersen, Mitchell A., and Raghuram G. Rajan (1994), “The Benefits of Lending Relationships: Evidence from Small Business Data,” The Journal of Finance 49 (March): 3 – 37. ——— (1995), “The Effect of Credit Market Competition on Lending Relationships,” The Quarterly Journal of Economics 110 (May): 407– 43. Federal Reserve Board (1997), Senior Loan Officer Opinion Survey on Bank Lending Practices (January), http://www.federalreserve.gov/boarddocs/SnLoanSurvey/ 199701/default.htm. Prager, Joshua Harris (1999), “More Small Firms Get a Break from Banks,” Wall Street Journal, June 28, A2. Goldberg, Lawrence G., and Lawrence J. White (1998), “De Novo Banks and Lending to Small Businesses: An Empirical Analysis,” Journal of Banking and Finance 22 (August): 851– 67. Strahan, Philip E., and James P. Weston (1998), “Small Business Lending and the Changing Structure of the Banking Industry,” Journal of Banking and Finance 22 (August): 821– 45. Greenbaum, Stuart I., George Kanatas, and Itzhak Venezia (1989), “Equilibrium Loan Pricing Under the Bank– Client Relationship,” Journal of Banking and Finance 13 (May): 221– 35. U.S. Small Business Administration (1999), “The Facts About Small Business 1999,” U.S. Small Business Administration Office of Advocacy. 32 FEDERAL RESERVE BANK OF DALLAS