Economies grow as a result of new ideas enabling innovations that render existing technologies obsolete. Yet, explaining economic growth by using the growth of ideas just pushes the question a step further. If growth comes from new ideas, where do ideas come from? With the availability of new data sources on the inputs and outputs of innovation, we have started to build a fairly accurate description of our idea-producing machine. A picture emerges where ideas are cumulative, innovation relies on the ability of ecosystems to produce complex combinations of new ideas, and geography still poses a barrier for knowledge exchange. This work contributes to our understanding of innovation by documenting three stylized facts about knowledge creation: the type of knowledge matters for new companies, complex knowledge is better produced in large cities, and urban vibrancy can help enhance knowledge spillover. First, we document that when starting new ventures, knowledge about the industry is more important than knowledge about the occupations involved. Second, we find that complex economic activities tend to be disproportionately concentrated in large cities and that this concentration has been growing for the past one hundred and fifty years. Third, we use the staggered roll-out of state-level R&D tax credits in the US together with department-level publication data to measure the benefit to university researchers working in close physical proximity to private researchers. We find that urban vibrancy plays a role in increasing the spillover to academia. Our understanding of innovation used to be based on speculation built on anecdotes and stories of success. With the availability of new data sources and platforms that track different pieces of our idea-making machine, we are no longer restricted to study innovation by focusing only on the big winners. The first chapter focuses on how worker mobility can bring different types of knowledge to pioneer companies in Brazil. Using methods from network science to build indicators of knowledge relatedness, we explore the question: how does the success of entrepreneurial activities depend on the experience of a team? We measure the industry-, occupation-, and location-specific knowledge carried by workers from one establishment to the next, using a dataset summarizing the individual work history for an entire country. Our results show that hiring workers with industry-specific knowledge produces the largest and most significant boost in the survival and growth of new firms. This is particularly important for pioneer firms, which are firms operating in an industry that was not present in their region. Pioneers are of particular importance because the success of pioneers is the basic unit of regional economic diversification. The second chapter studies how the spatial concentration of economic activities depends on its knowledge complexity. Are economic activities that rely heavily on complex knowledge more concentrated? How has their concentration changed in the last decades? We find that complex economic activities, such as biotechnology, neurobiology, and semiconductors, concentrate disproportionately in a few large cities compared to less complex activities, such as apparel or paper manufacturing. We use multiple proxies to measure the complexity of activities, finding that complexity explains from 40% to 80% of the variance in urban concentration of occupations, industries, scientific fields, and technologies. Using historical patent data, we show that the spatial concentration of cutting-edge technologies has increased since 1850, suggesting a reinforcing cycle between the increase in the complexity of activities and urbanization. These findings suggest that the growth of spatial inequality may be connected to the increasing complexity of the economy. The third chapter explores the role of urban vibrancy in mediating knowledge spillover between two types of knowledge workers: private researchers and university researchers. Do university researchers benefit from private R&D? Does this benefit depend on the urban environment around them? Using the staggered roll-out of state-level R&D tax credits in the US together with department-level publication data, we measure the benefit to university researchers working in locations dense with related industry R&D. We use data on patents to calculate how exposed university researchers are to related private R&D, and data on the density of cafes and restaurants to build an index or urban vibrancy. We find that university researchers benefit from R&D tax credits only when located in areas dense with related industry R&D activity. More importantly, urban vibrancy increases the benefits from R&D tax credits when the academic department is located in areas dense with related industry R&D. These results highlight that although the urban environment can increase the positive externalities of investment in R&D, it cannot create innovation by itself. Understanding how ideas are created used to be based on speculation built on anecdotes and stories of success. With the availability of new data sources and platforms that track different pieces of the ideas-making machine, we are no longer restricted to study innovation by focusing only on the winners.Ph.D.