Pronosticos de empleo formal urbano.

• Author: Lora, Eduardo
• Pages: NA

Forecasting Formal Employment in Cities

Previsoes para o emprego formal urbano

Introduction

United Nations Sustainable Development Goal 8 is to "Promote sustained, inclusive and sustainable economic growth, full and productive employment and decent work for all". More specifically, target 8.3 seeks to "[b]y 2030, achieve full and productive employment and decent work for all women and men, including for young people and persons with disabilities, and equal pay for work of equal value". This paper assesses how achievable this target is for Colombia, based on a novel theory of formal employment creation in cities and two complementary forecasting methods: standard regressions and machine learning.

Cities are necessary for economic growth to take place through a process of diversification and innovation that leads to productive employment and decent work for larger shares of the population. However, urbanization is not a sufficient condition for industrialization and productive employment: the expected relation between urbanization, industrialization, and employment quality is absent in many parts of the world (Gollin et al., 2016). Urbanization patterns, not just urbanization rates or macroeconomic factors (such as natural resource abundance), may shed light on the role of cities in economic growth and formal employment creation as suggested by two reliable facts (O'Clery et al., 2020): (i) formal occupation rates are more variable across cities within countries than across developing countries and (ii) larger cities create proportionally more formal employment.

Theoretical Framework

One of the central issues in economic development theory is the reason for the size and persistence of informal labor in developing economies. Since formal firms have access to capital and technology that make them more productive than small or family businesses, what explains the large quantity of labor force not occupied in the formal sector where labor conditions are better than in the informal sector? Economic theory has provided several explanations. In dualistic models of informality, the self-employed and their family businesses are fundamentally different from formal firms in the type of human capital they use--mainly uneducated and unproductive entrepreneurs and managers--, and in what they produce--mainly low-quality products for low-income customers--. The formal and informal sectors co-exist because they are different (Lewis, 1954; Harris & Todaro, 1970; Rauch, 1991). An alternative view is that of De Soto (1989, 2000), who considers that informal firms are an untapped reservoir of productive resources held back by government regulations. Relatedly, Levy (2008) sees informal businesses as entrenched firms that survive despite their low productivity by avoiding taxes and regulations. Lastly, in labor search models, which take into account the costs and benefits of labor regulations, informal employment is not the consequence of exclusion, but the result of labor market frictions between heterogeneous workers and firms (Albrecht et al., 2009; Bosch & Maloney, 2010; Ulyssea, 2010; Meghir et al., 2015).

While empirical evidence has been provided to support each of these explanations of informality, none of them recognizes the two facts mentioned in the introduction: that formal occupation rates across cities (within a given country) have a larger variance than across countries and that formal occupation rates are directly and significantly associated with city size. In other words, none of the mainstream theories can explain the role of cities in formal employment creation. Furthermore, some of the main variables put forward by those theories to explain the presence of informality--such as social security regimes and labor hiring and firing legislation--have little or no variance across cities within each country.

In view of these shortcomings, this paper adopts the theoretical framework developed by O'Clery et al. (2019, 2020), which differs from previous theories in a number of ways. First, it focuses on cities rather than countries because cities are the actual locations where workers and their employers interact. Second, it emphasizes skill diversity--which is central in urban economics--rather than skill levels, educational attainment, or managerial capabilities. Third, it assumes that firms evolve by tinkering with skills because many feasible technologies cannot be known in advance but need to be discovered. Formal employment creation in cities results from this evolutionary process. In larger cities, firms have better access to the diverse skills they need to produce more sophisticated goods.

Accordingly, in O'Clery et al. (2019), the creation of formal employment between period t and period t+n in city c, AempC depends on "complexity potential" cp in period t, which is a measure of the availability of skills of the local labor force needed in the more complex industries not yet present in the city:

[DELTA][emp.sub.C] = [emp.sub.C,t+1] - [emp.sub.C,t] = f (C[P.sub.c,t]) (1)

C[P.sub.c,t] = 1/[absolute value of ([M.sub.c,t])] [summation]i[member of][M.sub.c,t] [d.sub.c,i][C.sub.i] (2)

Notice that complexity potential is a weighted average of the industry complexity [C.sub.i] of the missing sectors at time t [M.sub.c,t], with weights given by the density [d.sub.c,i] of industries that use skills similar to those of the missing sectors.

In order to operationalize equation (2), data are needed on industry complexity, missing sectors, and skill similarity between all pairs of industries. Since skills are tacit knowledge and therefore unobservable, industry complexity and complexity potential must be computed indirectly. To that end, O'Clery et al. (2019) make use of the methodologies developed by Hidalgo and Hausmann (2009) and Neffke and Henning (2013). In essence, industry complexity is a measure of the range of skills needed in an industry, which is obtained from the number of industries present in the cities that have the industry (i.e., those industries that have revealed a comparative advantage greater than 1 in a city, based on formal employment shares) and the number of cities that have the industry (i.e., those cities where the industry has revealed a comparative advantage greater than 1). Skill similarity between a pair of industries is measured by the relative intensity of the labor flows between the two industries, and missing industries in a city are those with a revealed comparative advantage lower than 1.

Data and Empirical Definitions

Like in O'Clery et al. (2019, 2020), I use data for Colombian cities larger than 50.000 inhabitants. My definition of cities rests on the methodology proposed by Duranton (2015) to define metropolitan areas. It consists of adding iteratively a municipality to a metropolitan area if there is a share of workers above a given threshold that commute from the municipality to the metropolitan area. Assuming a 10 % threshold, the methodology generates 19 metropolitan areas that consist of two or more municipalities (comprising a total of 115 municipalities). Since another 43 individual municipalities have populations above 50.000 inhabitants, a total of 62 cities was obtained.

The main data source for the 62 cities was the social security administrative data collected by the Ministerio de Salud y Seguridad Social (Health and Social Security Ministry), known as pila (Planilla Integrada de Liquidacion de Aportes). PILA contains information of workers and firms on the days worked, the sector of activity, and the municipality. (1) To aggregate these data, I count the share of the year t that each worker effectively contributed to the social security system through firms per city c per industry j ([emp.sub.c,j]). This is the formal employment for a given sector (or for the aggregate of all sectors within a city). Sectors are defined at the 4-digit industry level of the International Standard Industrial Classification (isic, revision 3.0).

The formal employment rate in city c in year t ([f.sub.c,t]) is defined as formal employment divided by the city-wide population 15 years old or older ([pop.sub.c,t], estimated by DANE):

[f.sub.c,t] = [emp.sub.c,t]/[pop.sub.c,t] (3)

The (simple) average formal occupation rate in cities was only 20.3 % of the working age population in 2015 with a relatively large standard deviation (11.1 %). Important changes in urban formal occupation rates occurred between 2008 and 2015: the aggregate formal occupation rate for the 62 cities went up from 21.1 to 31.2 % with a (simple) average increase across cities of 8.1 % and a standard deviation of 5.4 %. Formal occupation was facilitated by a rate of gdp growth of 4.1 % and probably by the elimination in May of 2013 of payroll taxes and surcharges representing up to 13.5 % of the wage bill of some groups of workers (Kugler et al., 2017).

Since the formal employment rate is a variable bounded between 0 and 1, and the aim is to assess how fast it approaches 1, it is transformed to its logistic form, time-differentiated and expressed in annual terms:

[mathematical expression not reproducible] (4)

Where [y.sub.c,t-i] will be the dependent variable and the subscript i is the year-interval or number of years for the time-differentiation (which may take values between 1 and 7, given that the data cover an 8-year span). For intuition's sake, I will refer to the dependent variable as the annual speed towards full employment, or speed, for short.

The independent variables (at time t-i) were complexity potential, C[P.sub.c,t-i], as explained above, the (log of) working age population, [lpop.sub.c,t-i], the logistic of formal occupation rate, [mathematical expression not reproducible], a dummy for the oil-producing cities (those with more than one oil well per 10.000 inhabitants: Acacias, Arauca, Barrancabermeja, Neiva, and Yopal), and a synthetic measure of the exogenous sectoral shocks by city c (following...