Policy Research Working Paper 10151 How Regulation and Enforcement of Competition Affects ICT Productivity Evidence from Matched Regulatory-Production Surveys in Peru’s ICT Sector Tanida Arayavechkit Charl Jooste Ana Urrutia Arrieta Latin America and the Caribbean Region & Macroeconomics, Trade and Investment Global Practice August 2022 Policy Research Working Paper 10151 Abstract How the enforcement of competition regulation of infor- much depends on the regulatory structure, which affects mation and communications technology affects growth productive firms differently depending on how long they depends on how well firms adapt to competitive pressure. have been in business. Highly productive older firms trans- This paper tests this empirically using Peruvian firm-level late regulations that make processes more complex (such data matched to a compilation of information and com- as raising quality standards) into more productivity; pro- munications technology regulations and competition ductive younger firms benefit more from simplifying rules enforcement cases over 10 years. Based on the theoretical that facilitate competition through lower entry barriers and dispersion in markups, the paper shows that by increasing improved operating conditions. This feature is consistent productivity, leaders in a market can avoid the effects of across different segments of the information and commu- competition while maintaining market share. However, nications technology sector. This paper is a product of the Office of the Chief Economist, Latin America and the Caribbean Region and the Macroeconomics, Trade and Investment Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at cjooste@worldbank.org, tarayavechkit@worldbank.org, aurrutiaarrieta@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team How Regulation and Enforcement of Competition Affects ICT Productivity: Evidence from Matched Regulatory-Production Surveys in Peru’s ICT Sector* Tanida Arayavechkit1 , Charl Jooste1 , and Ana Urrutia Arrieta1 1 World Bank JEL classification: C31, D24 Keywords: ICT, competition policy, productivity, markups. * We are grateful to James Sampi, Ekaterina Vostroknutova, Chad Syverson, Doerte Doemeland, Jorge Araujo, William Maloney, Ana Paula Cusolito, Martin Rama, Elena Ianchovichina and Tania Begazo, for helpful and constructive comments. Jose David Prieto Campo provided excellent research assistance . Corresponding authors can be reached at: tarayavechkit@worldbank.org, aurrutiaarrieta@worldbank.org, cjooste@worldbank.org 1 Introduction This paper studies the role of competition regulation and enforcement in the ICT ecosystem and the effect on economic growth. We analyze how regulations that make processes more complex or simpler interact with enforcement to produce productivity outcomes. Our economic framework characterizes the dispersion in markups as the channel of distortions. Markup dispersion is used to identify large market power and weak competi- tion, which together reduce incentives to improve performance and raise productivity. For ICT firms, reducing the markup dispersion when misallocation is serious generates both productivity gains for ICT and productivity spillovers to other sectors. Matching firm-level data to a compiled data set of regulations and enforcement sheds light on the underlying theory. Enforcing ICT competition, controlling for the regulatory framework, not always reduces the dispersion in markups but may improve productivity: the ability of a firm to adjust to competitive pressures can increase productivity without necessarily causing market share to shrink, but firms that cannot adjust to competition (laggards) lose profits and may eventually exit the market. Adding to previous research on the importance of ICT to growth, we provide empirical evidence that through regulation more competition and higher growth create competitive pressure. How ICT contributes to growth is well-established (for a literature review see Dedrick et al. (2003) and Draca et al. (2006)); for one thing, it lowers the cost of exchanging information. ICT impacts productivity in the direct production of ICT goods and services through spill-over or network effects in other industries that use ICT goods and services (Alam et al. (2008)). It also contributes to the stock of capital (e.g., network towers, fiberoptic cables) and supports worker upskilling, enhancing labor productivity through learning opportunities opened up by online platforms. Spill-over or embedded channels make it more complicated to identify the role of ICT in productivity because they are not directly measured, and often cannot be seen. It is therefore not surprising that the ICT contribution to growth has historically been under- estimated (Van Reenen (2010); Hempell (2005); Biagi (2013); Byrne & Corrado (2017)). ICT has been found to matter significantly to the rest of the economy . Not only does it make workers more productive (Brynjolfsson & Hit (2010)), it also gives small firms access to external markets, which improves competition and makes it possible to import skills (Dutz et al. (2018)). It enhances consumer welfare by reducing inefficiencies in the deliv- ery of goods and services (Biagi (2013)) and improves the quality of goods and services produced. ICT enables countries to leapfrog the stages of deployment and hence growth by adopting frontier technologies without incurring the massive costs of first deploying out- dated technologies (see examples provide by Hilbert & Katz (2003) - such as adapting 5G 1 technologies without first having to adopt 3G). However, ICT can also disrupt traditional markets. First-mover firms that adopt ICT early may increase their market share (De Ridder (2019)), which happens when firms are able to substitute away from labor (reduce marginal costs) and produce outputs using ICT technologies (one-time fixed costs). Though this generates aggregate short-term growth benefits, however, in the long run it may reduce growth if new firms are unable to compete with incumbents due to major upfront fixed costs (Aghion et al. (2019)). The length and efficacy of the ICT contribution to growth depends critically on the dif- fusion process such as uptake and learning (Hilbert & Katz (2003)). Fast diffusion may imply a quick growth spurt but not necessarily a prolonged one. Slow diffusion might imply a smaller but longer growth process. In present-value terms fast and slow diffu- sion can be equal unless continual innovations are part of the process. It is important to note, however, that growth depends on adequate infrastructure, hardware, and software. Weak regulations, whether misguided or intentional, can cause problems on both the de- mand side (diffusion and adoption) and the supply side (development and deployment) of infrastructure, hardware,and software. This paper focuses on the supply side effects of competition policy. Historically, in most countries in Latin America and the Caribbean (LAC) the public sector provided broadband and ICT goods and services, on the theory that (1) the fixed cost was simply too large for private sector participation and (2) these were in any case considered public goods (Hilbert & Katz (2003)). Today economies of scale and in in some countries greater technological achievements have moved from the public provision model to one where some aspects of ICT are being privatized. Private initiatives can fill in gaps in public provision and address inequalities in access to knowledge and connectivity; however, although in theory regulation might try to promote private initiatives, in practice it may end up creating inefficiencies because of the mix of both inadequate institutional capacity (e.g., a lack of technical skills and inefficient procedural processes), and political will (e.g., inertia from historical views regarding the role of big companies and the state, and often myopic political preferences and corruption). While the relationship between the use of ICT and firm performance is positive, com- petitive pressures complicate it (Iacovone et al. (2016)). ICT network development and the resultant productivity gains could be accelerated by removing regulatory bottlenecks. Examples of regulatory constraints include online services that countries tax differently. In- ternational players may gain an unfair competitive advantage in terms of pricing, and hence demand, relative to local producers of similar products and services (ITU (2017)). Some regulations can suppress growth by deterring important investments (Dutz et al. (2018); Alam et al. (2008); Cette et al. (2017)). Highly regulated ICT services may also lead to suboptimal allocation of resources in the upward value chain (Paterson et al. (2013)). As 2 ICT services linked to manufacturing become stronger, for example, so does the impact of regulation. Van Reenen (2010) shows that countries whose labor and product markets are highly regulated tend to be less productive, mainly because the regulatory hurdles discour- age investment in ICT; as a result, the ICT potential is far below that of peer countries that impose fewer, or more supportive, regulations. Privacy laws are also very contentious. Laws governing the use of data, how to target advertisements, and user consent are continuously changing. Campbell et al. (2015) show that stricter privacy laws are one reason why the EU has fewer internet advertising firms than the US: there is less incentive to enter a market when the expected revenues are lower. Developing infrastructure often entails large fixed costs. Land rights, access rights (via licenses or permits), sharing of infrastructure, and bidding for spectrum are important to competition both within markets but across users of infrastructure in other markets. Com- panies that win tenders, as in cases when there are natural monopolies, publicly funded monopolies, or when little product substitutes exist, might be tempted to abuse their dom- inance (despite concessional limiting abuse in awarded contracts) by charging high prices for use of the infrastructure which limits competition at a cost to consumers. Unregu- lated companies may also price-discriminate, which also creates uncompetitive outcomes. Restrictions on foreign ownership of infrastructure are another distortion that limits firm entry and technology spillovers (Lee & Shy (1992))and suppresses ICT penetration (Katz & Jung (2021)). Strong patent laws can restrict imitation of technologies, which height- ens the market power of firms that innovated early; some regulations limit the number of distributors for certain software or hardware (e.g., chip devices) using quotas, licenses, or permits, which is another way regulation can deter firm entry; some regulations may be a burden for producers requiring important input components with high tariffs. In terms of software, some regulations may prohibit distribution rights (think of operating systems or sales of commercial software); poorly drawn regulations might not protect user data on digital platforms (think of directed advertising); or economies of scale where the first mover has an advantage in data collection and processing, which discourages new entrants from competing. Finally, certain regulations might not break up network externalities, such as tie-in sales where a company might create frontier hardware technology but allows only its own less effective software to operate on it. These are just a few examples of competition limiting cases that apply in different markets for ICT products. The contribution of this paper is to study the impact on productivity from regulatory and enforcement changes that affect competition in the product market. We present evi- dence that pro-competition regulation can reduce distortions, especially if it is efficiently enforced. We present a stylized model for understanding some of mechanisms that drive the impact of competition policy on productivity within the ICT sector and the spill-over to other sectors. 3 In what follows, the paper describes the methodology for this study and the empirical results. We then review the deployment and uptake of ICT in LAC and compute measures of relative distortions. Section 3 describes economic framework that allows us to gauge how removing distortions can lead to competition and productivity improvements in the ICT sector. Section 4 looks at the empirical links of de jure and de facto policies on productivity outcomes in Peru. It maps firm-level data to regulations and enforcement data at the four-digit level where markets are delineated as segments of the ICT sector. Collection of regulation and enforcement data is important for studying how regulations may improve growth and reduce market distortions. 2 ICT Distortions in LAC The uptake of computers and the internet is evidence of the importance of ICT for both work and personal activities. Computer and internet penetration rates have increased across the globe. According to the International Telecommunication Union (ITU), in 2020 the share of LAC households that own a computer ranged from a high of 68% in Uruguay to a low of 17% in Honduras (ITU (2020)). The rates are slightly higher for access to the internet (households that do not own a computer may own a cellphone with internet capabilities; see Figures 1 and 2). ICT adoption varies widely in LAC: Uruguay, Argentina, and Brazil have the highest adoption rates; Honduras, Nicaragua, and Guyana the lowest (Figure 3). Unfortunately, the benefits of ICT are not shared between and within countries in LAC. For instance download speeds in some countries are slow relative to peers and developing countries generally (Figure 4). Slow uptake of ICT could be the result of several factors, including the current devel- opment context and the regulatory environment of each country. The ITU’s ICT Regulatory Tracker provides comparable scores for both regulatory and competition frameworks (ITU (2020)). The Tracker traces private vs. public ownership, the degree of competition for broadband, basic and leased line services, and whether the law recognizes the concept of significant market power. These are classified as G1, which cannot be considered a competition model, to G5, which features independent regulators, a pro-competition en- vironment, and working together across agencies and firms to improve the deployment of ICT. The ITU scores each country according to these criteria. Table 1 summarizes the set of scores for Peru as an example of the different measurements. LAC seems representative of a competitive environment. Figure 5 ranks the competi- tiveness of ICT from a monopoly (score = 4) to one that is fully competitive (score = 1). Because these scores are based on self-reporting, they do not necessarily relate to outcomes. Despite seeming to be pro-competition, in terms of costs LAC is more discouraging. The 4 ITU collects data on connection fees, broadband wireless charges, broadband tax rates, business telephone installation fees, broadband caps, monthly subscription charges and broadband speeds. From these measures we compute a cost index using simple principal components. An increase in the index implies that costs are rising over time. Note that data are standardized across countries and hence an increase should reflect idiosyncratic increases in costs that imperfectly controls for inflation, quality changes or upgrades. To construct this index, all the input variables should have the same sign - so that the inter- pretation of an increase implies a higher cost relative to a global mean (see Appendix A, Table 41). The estimated indices are summarized in Figure 6. Costs have been rising over time for all countries in the sample. These costs, associated with higher fees, and lower comparative speeds indicate entry barriers and hence market concentration. Although rising total costs could also indicate quality improvements, the measures here are mainly associated with regulatory costs. A second measure to compute gaps in the input market uses the statistic from Bartelme & Gorodnichenko (2015) and computes the deviations from a frontier market using input- output tables. This measure tracks distortions in input markets by sector, on the assumption that an undistorted economy functions like a market leader. A further assumption is that all countries use the same intermediate goods technology. The distortion is then computed as a ratio of inputs in LAC countries relative to a frontier country in the provision of ICT goods (here the Republic of Korea). A higher ratio implies that the input distortion has risen relative to the frontier economy. The undistorted share of intermediates of country i in sector j in period t (bi,j,t ) is related to the distorted share (γi,j,t ) via a scalar that captures distortions (τi,j,t ):1 bi,j,t τi,j,t = γi,j,t (1) Given that input-output tables are based on observed variables (i.e., they do not distin- guish between distorted and undistorted states); the size of the distortion is then expressed relative to a frontier country (e.g., the United States or Korea). We extract the relative dis- tortions using IO data: KOR j bi,j,t τi,t =1− KOR (2) j bi,j,t 1 n The firm profit maximizing problem reads as max Pj,t Yj,t −wNj,t −rKj,t − i=1 (1+ τi,j,t )Pi Mi,j,t , Kj,t ,Nj,t ,Mi,j,t α 1−γj n γ where the Cobb-Douglas production function is Yj,t = Kj,t ((Aj,t Nj,t )α ) i,j i=1 Mi,j,t , where N is labor, K capital stock, M intermediate inputs, P, w, r the price of output Y , labor and capital respectively. The first P Mi,j,t γi,j order condition for the intermediate input is bi,j,t = iPj Y j = τi,j . The undistorted economy (i.e., without the tax wedge) is equal to γi,j,t . The wedge in this setup reduces the input share coefficient. 5 If a country’s distortion is larger than that of the frontier country, then τ > 0, since by design it produces less. Figure 7 summarizes the estimated distortions by industry in Peru, Colombia, Chile, Costa Rica, Mexico and Brazil. Except for Mexico, all the countries have significant distortions in ICT relative to Korea - in the case of Costa Rica up to 40%. Two aspects should be noted from these estimates: (1) Output distortions have been rising over time - our hypothesis is that they limit entry, make markets more concentrated and have slowed productivity growth. (2) Input distortions in the ICT sector are high relative to frontier economies, but remained constant over the sample period. These higher distortions are correlated with low spending by ICT firms on research and development (R&D) relative to middle-and high-income countries (Figure 8) and with low patent registrations (Figure 9). 3 ICT Misallocation and Productivity This section presents a framework to examine the role of ICT production and competition in aggregate productivity. It builds on Edmond et al. (2015) who analyzed the impact of competitive gains from international trade with varying markups and that where there is extensive misallocation, greater competition reduces markup distortions by creating com- petition and hence reduces productivity losses. In an extension, Edmond et al. (2018) decompose markup costs into three distinct channels: (1) Aggregate markups act as a tax on output. Competition within a market is low when markup dispersion is higher among firms in that market. (2) The result is a misallocation of factors of production. This is often associated with reallocation of production from firms with high labor shares to ones with low shares and (3) it leads to inefficient entry. An interesting result of the Edmond et al. (2018) framework calibrated using U.S. cen- sus data is that while competition reduces the markup of incumbents, it also reallocates market share toward the largest incumbent firms, leaving the net effect of markups un- changed. We recast Edmond et al. (2015) in the domestic market as two sectors, ICT and non- ICT and three types of firms: final good producers, traditional intermediate goods firms and ICT intermediate goods firms. 3.1 Final Goods Producers A single consumption good is produced in a perfectly competitive industry using inputs y (s) from a continuum of sectors: θ θ −1 θ −1 Y = y (s) θ ds (3) 6 where θ > 1 is the elasticity of substitution between sectors. Each input y (s) is produced from both traditional and ICT intermediate inputs, xi and zi : γ nx (s) nz (s)   γ −1 γ −1 γ −1 y (s) =  xi (s) γ + zi (s) γ  (4) i=1 i=1 where γ > θ is the elasticity of substitution across inputs within a particular sector s and α is the share parameter. The number of traditional and ICT intermediate good producers in sector s, nx (s) and nz (s), are determined in equilibrium by the free entry condition. 3.2 Intermediate Good Firms There are two types of intermediate goods firms in each sector s: traditional and ICT. A traditional intermediate good firm i in sector s produces output using labor (l) according to the following production function: xi (s) = ax x i (s)li (s) (5) Similarly, an ICT intermediate good firm i in sector s uses the production function: zi (s) = az z i (s)li (s) (6) where ax z i and ai are firm-level productivity. Section 3.3 discusses firm entry and productiv- ity draw. Final good producers buy both traditional and ICT intermediate inputs at prices px i (s) and z pi (s), and sell a final good at price P . A final good producer chooses intermediate inputs xi (s) and zi (s) to maximize: nx (s) nz (s)   max P Y −  px i (s)xi (s) + pz i (s)zi (s) ds  (7) xi (s),zi (s) i=1 i=1 The first order conditions yield the demand functions: −γ −θ px i (s) p(s) xi (s) = Y, (8) p( s) P −γ −θ pz i (s) p(s) zi (s) = Y, (9) p(s) P The aggregate and sectoral price indices can be written as 1 1−θ 1−θ P = p(s) ds , (10) 1 nx (s) nz (s)   1−γ 1−γ 1−γ  p( s) =  px i (s) + pz i ( s) . (11) i i 7 Intermediate goods firms engage in Cournot competition within a sector. Taking the wage w as given, a traditional intermediate good firm chooses a quantity to maximize its profit: w max px i ( s) − xi ( s ) , (12) xi (s) ax i ( s) and, likewise, an ICT intermediate good firm chooses a quantity to maximize its profit: w max pz i (s) − zi (s) (13) zi (s) az i ( s) where the demand for xi (s) and zi (s) is given by Equations (8) - (11). The maximization problems imply that intermediate good firms charge a markup over marginal cost εx i (s) w px i (s) = x x , (14) εi (s) − 1 ai (s) εz i ( s) w pz i ( s ) = z z , (15) εi (s) − 1 ai (s) where εx z i (s) > 1 and εi (s) > 1 can be written as:  −1 1−γ px i ( s) 1 px (s) 1−γ 1 εx i (s ) =  + 1− i , (16) p(s) θ p(s) γ  −1 1−γ 1−γ pz i (s) 1 pz i (s) 1 εz i ( s ) =  + 1− . (17) p(s) θ p(s) γ px 1−γ pz (s) 1−γ i ( s) p(s) and pi(s) are the firms’ share of sectoral revenue, and εx z i (s) and εi (s) are the demand elasticity faced by firms. As γ > θ, firms with larger shares of sectoral revenue face lower demand elasticity and can charge higher markup. The markup a firm charges is an increasing convex function of its market share. At one extreme, when there is no competition, a pure monopolist charges a markup of θ/(θ − 1). On the other extreme, in a perfect competition environment, an infinitesimal firm charges a markup of γ/(γ − 1). The more firms enter the market, the lower the markup is. 3.3 Free Entry Condition and Equilibrium An intermediate good firm pays a sunk cost f x or f z that allows it to draw a sector s (to- gether with sector productivity) and idiosyncratic productivity. Firm productivity ax i (s) and z x z ai (s) are a product of a sector-specific component (ψ (s) and ψ (s)) and an idiosyncratic component (ν x (s) and ν z (s)): aj j j i (s) = ψ (s)νi (s); where j ∈ {x, z } (18) (19) 8 The sector-specific component is drawn independently and identically distributed from a Pareto distribution with shape parameter ξ > 0. Within a sector s, the idiosyncratic component νij (s) ≥ 0 is drawn from a discrete distribution {0, 1, ν ¯j (s)} with probability { 1 − pj j j j j l (s) − ph (s), pl (s), ph (s)}. An entrant draws a random variable νi (s) that determines if it is successful. With probability pj j l (s) + ph (s), the entrant is successful and begins operat- ing. With probability pj h (s), the entrant is successful and becomes a high-productivity firm. With the discrete distribution, in equilibrium each sector is characterized by λ = { px x z z x z x x z z l , ph , pl , ph , ψ , ψ , nl , nh , nl , nh }, (20) where nj j l and nh j ∈ {x, z } are the number of low-productivity and high-productivity firms, respectively. The expected profit of an entrant is j πe = pj j ′ j j ′ j l πl (λl ) + ph πh (λh ) − wf (λ) dF (λ); where j ∈ {x, z } (21) where λ′l is equal to λ except that nj j ′ l is replaced by nl + 1, and λh is equal to λ except that nj j j j h is replaced by nh + 1. πl and πh are an operating profit for low productivity firms and high productivity firms as described in Equation (12) and (13), respectively. The number of traditional and ICT entrants N x and N z are pinned down by the free entry conditions: x πe (N x ) = 0 (22) z πe (N z ) = 0 (23) Lastly, there is a representative consumer that inelastically supplies one unit of labor and consumes the final good. The labor market clearing condition is then (nx x x x z z z z x x z z l (s)ll (s) + nh (s)lh + nl (s)ll (s) + nh (s)lh (s) + N f + N f ) ds = 1, (24) where nj j l (s) and nh (s) j ∈ {x, z } are the number of low-productivity and high-productivity firms, respectively. The market clearing condition for the final good is simply C = Y . 3.4 Markup and Productivity Loss Aggregate productivity of the economy is defined as: Y A = ˜ (25) L    −1 1 xi ( s ) 1 zi (s)   =   nx i ( s) x + nz i (s) z ds , (26) i∈l,h ai (s) Y i∈l,h ai ( s ) Y 9 ˜ is the aggregate amount of labor net of fixed entry costs. where L Aggregate markup is aggregate price P divided by aggregate marginal cost w/A and can be written as: PA µ = (27) W    −1 1 px i (s)xi (s) 1 pz i (s)zi (s)  =   nx i ( s ) x + nz i (s ) z ds (28) i∈l,h µ i ( s ) P Y i∈l,h µ i (s ) P Y where µx z i (s) is the markup of a traditional intermediate good firm i of sector s and µi (s) is the markup of an ICT intermediate good firm i of sector s. Therefore, the aggregate productivity can be expressed as a function of markup:   1 −θ θ −1 µ(s) A= a(s)θ−1 ds (29) µ where the sector-level markup µ(s) is p(s)a(s) µ( s ) = , (30) W and the sector-level productivity is   1 −γ −γ γ −1 γ −1 µx i (s) γ −1 µz i ( s) a(s) =  nx x i (s)ai (s) + nz z i (s)ai (s)  , (31) i∈l,h µ(s) i∈l,h µ( s ) In the optimal allocation equilibrium, there is no markup dispersion µx z i (s) = µi (s) = µ(s) = µ ∀i, s, and the aggregate productivity is 1 θ −1 ∗ θ−1 A = a(s) ds (32)   1 γ −1 a∗ (s) =  nx x i (s)ai (s) γ −1 + nz z i (s)ai (s) γ −1  (33) i∈l,h i∈l,h 3.5 Calibration The model is calibrated to reproduce several stylized features from the Peruvian firm sur- veys for 2007-17. We assume there are five sector types. With probability Ωk , an entrant is assigned the sector type k . The sector type determines the probability that an entrant draws specific productivity and the distribution of competitors it will face. For simplicity, 10 we assume that the probability of becoming a high or a low productivity firm is the same for traditional and ICT firms and across sector types. Values of the following parameters are chosen to minimize the distance between model moments and counterparts in the Pe- ¯z , pl , ph , fx , fz , {Ωk }5 ¯x , ν ruvian data: ξ, ν k=1 , γ, θ . ¯x , pl , ph , fx govern the productivity distribution within and between sectors. Parameters ξ, ν ¯z and fz determine the market share of ICT firms and aggregate fraction firms that under- ν take ICT activities. In the data, a firm is identified as an ICT firm if it operates in the ICT sector defined in the ISIC Rev.4.2 {Ωk }5 k=1 are estimated to match the distribution of firms across 5 sectors: primary, manufacturing, wholesale trade, retail trade and services. The relationship between the elasticities θ and γ is determined by equations (14)-(17), from which we can rewrite a firm’s inverse markup as 1 γ−1 1 1 = − − ωi (s), (34) µi (s) γ θ γ where ωi (s) is the firm’s market share, and the inverse markup is simply the firm’s labor share. Letting α0 and α1 be the regression results of equation (34), we can estimate: −1 1 α1 γ−1 θ= − (35) γ α0 γ We then calibrate θ and γ using the estimated aggregate markup of 10% for Peru and equation (35). Table (2) and (3) report calibration results. 3.6 Quantitative Exercise Figure 10a shows the distribution of markups µi (s) in our baseline model for traditional and ICT intermediate goods firms, pooling over all sectors. The baseline model has an aggregate markup of 11.15%, an average markup of 11.07% and a standard deviation of log markup of 0.0008. Markups are more dispersed among ICT firms. Firms with more a larger market shares also exhibit higher markups (Figure 10b). With competition among ICT producers , ICT firms with small market shares can charge higher markups than firms that produce traditional intermediate goods. Reducing entry barriers (via a reduction in fixed costs, such as permit fees, licenses and competitive bidding) reduces market concentration. By lowering the dispersion in markups, sector-level productivity rises. Aggregation implies a net productivity gain, though only a small one. 2 The activities in the ICT sector can be grouped into ICT manufacturing industries (2610, 2620, 2630, 2680), ICT trade industries (4651, 4652, 4741) and ICT services industries (5820, 6110, 6120, 6130, 6190, 6201, 6202, 6209, 6311, 6312, 9511, 9512). 11 The model provides insights into the link between entry barriers, markups, and pro- ductivity. It mainly implies that reducing in entry barriers (e.g., through pro-competitive policies) reduces the dispersion in markups, which then generate less concentrated ICT markets. The link to sectoral and aggregate productivity depends on the degree of price dispersion and the elasticities of substitution. 4 Empirical Results The previous section highlighted the causal link between the regulatory environment, dis- tortions and productivity. This section focuses on microeconomic evidence by mapping exogenous variations in competition regulations (the sectoral laws) and competition en- forcement data to firm level data. We estimate the impact of regulations and enforcing regulations on productivity and markups for different ICT sectors (hardware, infrastruc- ture, telecommunications and software). The results show that competition enforcement by either Indecopi (the National Insti- tute for the Defense of Free Competition and the Protection of Intellectual Property) or Osiptel (the Supervisory Body of Private Investment in Telecommunications) has a positive impact on revenue total factor productivity (TFPR). Assuming that we account for vari- ations in markups correctly, this indicates that there is a real productivity improvement. Interestingly, and as Edmond et al. (2018) would have expected, this result holds mainly for the most productive firms, the leaders of the productivity distribution. It appears that middle-aged, but more productive firms benefit most from enforcement of competition regulations, however, inconsistent validation by robustness checks raises doubts about the significance of firm age. Markups seem to not be affected by competition enforcement cases. When isolating each ICT branch, we observe sector-specific patterns. 4.1 Data Description Firm-level data from the Annual Economic Survey (EEA) is collected by the National Insti- tute of Statistics (INE) for 2007-17. The EEA draws its sample from a directory of formal firms with annual sales above 150 Tax Units,3 based on administrative tax records. The survey stratifies the sampling frame by economic activity and firm size proxied by annual sales to ensure that it is nationally representative of Peruvian firms at the four-digit ISIC code level. For each year, given their sectoral importance, firms with annual sales above a certain threshold are selected using forced inclusion sampling (inclusion forzosa) to form 3 150 Tax Units amounted to 517,500 in 2007 and 607,500 in 2017. The tax is called Unidad Impositiva Tributaria. 12 a panel. Note, however, that market entrance and exit can only be approximated; a firm may no longer be observed if its sales have fallen below the selection threshold or if it has actually left the market. Similarly, market size can only be approximated because the panel is drawn from a selection from the universe of firms, rather than from the whole market. We look specifically at the ICT sector, which we separate into hardware, software, telecom- munications, and infrastructure subsectors.4 The survey collected 2,955 observations from ICT firms over the 11-year period. Of these firms, about 34% were active in markets where competition was enforced. In other words, of 940 distinct firms from all ICT sectors, 57% were subject to competition enforcement in their sub-sector. The largest number of firms produced hardware, which accounted for 48% of total ICT sales in the 2007–17 period; telecommunications accounted for 22% of sales, software for 17%, and infrastructure for 14%. Apart from Hardware, the variations in markup were fairly constant over the sample period (Figure 11). Most firms surveyed were established in the 2000s because of the inclusion criterion of large firms. Except for telecommunications, ICT employs a large share of labor in value- added. Telecommunications and infrastructure have the highest shares in concentration (prox- ied by the four-firm ratio) and markups, calculated from variable profits. 5 This is to be expected given the large entry costs, economies of scale and scope. ICT regulation and enforcement data are drawn from the Peruvian regulatory author- ities: the Ministry of Transport and Communications (MTC), Osiptel for telecommunica- tions and infrastructure, and Indecopi for other regulations pertaining to software and hardware. The MTC is the technical authority for defining and enforcing policy; one of its goals is to connect economic agents nationwide. It has multiple functions, but we focus on its role in the design, coordination, and evaluation of regulations for infrastructure and telecommunications; the evaluation and granting of licenses; and the enforcement and supervision of spectrum assignment. Osiptel is a specialized, autonomous, and decentralized public body tasked with pro- moting market competition. It regulates tariffs, regulates access, supervises, and enforces competition and technical norms for telecommunications. It therefore evaluates markets and carries out investigations both on its own initiative or in response to third-party com- plaints, resolving disputes as well as addressing anti-competitive behavior by imposing corrective measures or sanctions. Indecopi is charged with promoting and defending competition in all other sectors. It can evaluate and approve corporate mergers in any sector and adjudicate cases that might 4 For infrastructure we use ISIC 4321 and 4220. 5 Πt (Pt −Ct )Qt 1 Markups (µ) can be calculated using profits (Π), sales (P Q) and costs (CQ): Pt Q = Pt Qt =1− µ . 13 infringe competition. It is also charged with removing unnecessary regulatory barriers and simplifying administrative procedures through ex post assessment of norms. The applicable regulations are laws and decrees issued between 2007 and 2017 that specifically target the ICT sector and either facilitate or complicate competition. Facilitating regulations improve the administrative efficiency of entering or operating in an ICT mar- ket, improving competition by facilitating both market entrance and resource allocation, and potentially contributing to productivity growth. Conversely, regulations that impose additional administrative requirements or make current ones more complex are likely to generate the opposite effect, hindering market entrance, limiting resource movement, and thus deterring productivity growth. However, it could also be argued that new regulations can also address market failures because an explicit purpose of some is to improve ser- vice quality. In this sense, they could raise quality or productivity standards for entering a market or for doing business and remaining in the market, which would translate into higher average market productivity. Possibly, each new regulation could boost or depress productivity—the complexity it adds makes it hard to determine whether it promotes or restricts competition. To disentangle the layers of complexity, we look at how enforcement of any type of regulation affects productivity and markups: First, we study the behavior of firms that were affected by either an increase or a reduction in regulatory requirements. We then do the analysis separately by isolating observations that were affected only by one type of regulation. We match changes in regulations and their enforcement over time to the firm-level panel. Regulatory cases are matched to firm data by looking at the market affected, defined by the 2-digit ISIC sector and by the year of implementation. We compare the behavior of these firms to that of firms in the same ICT branch that were not subjected to any new regulation, controlling for firm, sector, and year characteristics through fixed effects—if other sector-specific regulations had been introduced, their effect will not be confounded because it is controlled by the fixed effects. The regulatory variables are coded as dummies where 1s imply a change in regulation as defined above. Appendix A.2 presents a compila- tion table and a summary of important ICT regulatory changes with reference to dates and justifications of the legal changes. For interpretation purposes, the regulations identified are matched to the OECD competition checklist to assess whether they affect the number of suppliers, the ability of suppliers to compete, the incentives of suppliers to compete, or the choices and information available to customers. Osiptel and Indecopi enforcement data capture the number of investigations of anti- competitive behavior that resulted in a fine or a corrective measure in each year and sector. Additionally, for telecommunications, the data set contains the fines by company name, which are then mapped to a corresponding ISIC code. Each resolution that found infringe- 14 ments to competition law defines the market affected at the 4-digit ISIC level and the year for which the correction of competition was enforced. Most cases did not specify a partic- ular region, which suggests that the anti-competitive behavior was not limited to a narrow location, so each case is assumed to have national impact. Moreover, the limited number of observations per region in the firm database reduced the possibility of studying the few cases that had a region-specific scope. When a resolution applied to a sub-region rather than the whole national market, our coefficients would be picking up a diluted, underesti- mated effect that would represent a lower bound of what the actual effect is. We assume that competition enforcement is exogenous to firm performance. Osiptel and Indecopi are independent and autonomous bodies that receive cases initiated exter- nally by complaint from any physical or moral persons and internally based on monitoring and control organs. This ensures that competition enforcement is applied nationwide re- gardless of firm or regional idiosyncrasies. Violation of competition law needs to be legally confirmed, following pre-defined requirements, for a case to be accepted. Rulings can be appealed to independent courts. All these elements support the assumption that interven- tions are exogenous. The information about de jure and de facto regulations was collected from the digital legislative archive from OSIPTEL and INDECOPI, following the criteria that the degree or resolution had to be relevant (i.e. pertain to at least one of the ICT branches of hardware, software, telecom or ICT infrastructure), timely (i.e. had to be published and implemented in the 2007-2017 period) and had to be related to competition (e.g., encouraging firm entry, reducing misallocation and administrative barriers that hinder innovation). 4.2 Empirical Approach The methodology described above requires that productivity and markups be identified consistently if the economic impacts of regulation and enforcement are to be accurately characterized. A widely used methodology to estimate productivity is Ackerberg et al. (2015) and a popular extension for estimating markups is De Loecker et al. (2020). Recent critiques of both methodologies require a change in assumptions or a very specific data set. As an example, Bond et al. (2020) show that it is theoretically impossible to recover markups if the firm-level data do not contain separate entries for quantities and prices. Another possible problem arises in the presence of demand shifters, which may render estimates biased (Doraszelski & Jaumandreu (2019)). On these terms, the data sets used here would be criticized; we do not observe price and quantity data separately (only price times quantity) and we need to account for several demand shifters. Sampi et al. (2021) recently addressed some of these criticisms and produced unbiased policy estimates in the presence of demand shifters as well as missing price data. The methodology, which entails several deviations from Ackerberg et al. (2015) 15 is agnostic about the production function; it controls for demand shifters, is a workaround when firm-level price data are missing and controls for marginal costs, which would render previous regression equations mis-specified. The demeaned (controlling for fixed effects) system of equations is summarized as: µ µ − log SXjt = τt + τ1 z1jt + τ2 z2jt − log(γ0 (log Kjt − log Njt ))γ + ϵµ jt (36) log Qjt = log Fjt + ϕt + ω (log Qjt−1 − log Fjt−1 − ϵjt−1 ) + ϕω z 1 1jt + ϕω ω 2 z2jt + ϵjt + ϵjt In addition to demeaning the data we include time fixed effects (τt , ϕt ) for each period t, and interact them with sectors and states to control for sectoral and provincial shifts. We include two demand shifters: a dummy for the competition regulation laws (z1jt ) that affect each firm j and one for competition enforcement (z2jt ). The dummies are coded as a 1 for when there is an enforcement case or a pro competition law is enacted, and for the years thereafter. The first equation is the markup equation, but written in terms of the revenue share (SXjt ). The output elasticity is approximated by log(γ0 (log Kjt − log Njt ))γ where Kjt is the stock of capital and Njt is the number of employees. The curvature of the elasticity relies on the estimated γ parameters. The second equation describes output, which is equal to the unknown functional form for the production function Fjt , which is approximated by the fitted values from a Taylor series expansion, F (Kjt ; Ljt ; Mjt ), where M is intermediate inputs. The system is jointly estimated using quasi-maximum likelihood (QLME) where the correlation (endogeneity) between error terms is directly estimated and hence controlled for (E (ϵµ ω jt (ϵjt + ϵjt )) = σ ̸= 0). This controls for various other problems that might arise in the data. For instance, when price data are not observed, quality improvement will affect both prices and productivity. Demeaning the system prevents us from recovering the value of γo , thus making it impossible to back out the value of the output elasticity (therefore markups) when the DGP process is a CES production function with elasticity of substitution significantly different from zero, or any other type of unknown production function. However, the approximation in Equation 36 will provide enough information on the curvature of the output elasticity regardless of its value in levels, thus the system in Equation 36 should produce consistent estimates for τ µ and ϕω , independent of the form of the production function. Given that the system is demeaned, the impact of the regulation and enforcement variables is related to the dispersion in the markup around the mean markup of the ICT sectors. This relates to the previous section where aggregate productivity is a function of the markup dispersion of ICT firms. Note that the same demand shifters (z ) enter both the markup and the output equation with different parameters. The regulatory variable is a signal to markets to change behavior 16 but also to indicate that a more competitive environment is desirable. A weak signal should not affect competition and productivity. The enforcement coefficient summarizes the direct and indirect implications of infringing. Here, too, if enforcement is weak then significant impacts should not necessarily be expected. We add additional fixed effects regressions to aid the narrative. Specifically, we esti- mate the impact of our competition variables on real sales, the real profit rate (real value less variable cost divided by real sales) and marginal costs (proxied by real unit labor costs). Even if our estimates of productivity and markups are insignificant, we may still see changes in costs or sales. Sales, for example, may go up even if productivity does not. If market demand for ICT services increase due to quality changes, for example, it may lead to an increase in the retail prices, which also push up sales, regardless of whether more competition is generated (e.g., lower markup dispersion) or productivity has increased. It is therefore important that the estimator used is robust to changes in market demand. If productivity and markups do not change but sales increase, that can only be due to a change in consumer demand, which we proxy using the sectoral price elasticity of de- mand, or a reduction in marginal costs, which we approximate using unit labor costs. To approximate quality we require a measure of aggregate demand. Our quality measure follows Khandelwal et al. (2013) and is equal to sales (P Y ) less the price elasticity of demand β multiplied by price. Although we do not observe prices in our data, we can approximate it with our demeaned output elasticity, demeaned wage share and demeaned marginal cost (i.e., log(P ) = log(εY − W N PY ) − W ). The output elasticity is simply equal to Y K ε = γ log log N . Once we construct our price proxy, we estimate the price elasticity of demand as log(Y ) = β log(P ) + F E controlling for several fixed effects (FE). The qual- ity measure derived here controls for markups (both firm and retail level) and changes in marginal cost. This is relevant for some ICT markets such as telecommunications. While competition regulation might not induce more entrants into the telecommunications sector (e.g., due to high fixed entry costs), it may improve the quality of outputs and thus increase sales. 4.3 Results We apply equation 36 to separate ICT subsectors and to the ICT sector as a whole. The results are summarized in several tables. Our results compare the QLME estimates to Ackerberg et al. (2015) (ACF) for productivity and De Loecker et al. (2020) (DLW) for markups. Because ACF and DLW are standard methodologies for estimating productivity and markups, they function as comparison cases. Each column of results is indexed. The tables of ICT and its associated sectors summa- rize the impacts of regulations (de jure) and enforcement (de facto). The first two columns summarize the impacts on productivity using the QLME and ACF methodologies. The third 17 and fourth columns summarize the impact of competition policy on markups using QLME and DLW. The remaining columns summarize the impact of competition policy on profits (the profit ratio), deflated sales, employment, real marginal costs, real intermediate costs per worker, product quality and TFP covariance (or between effects). Several robustness checks were performed on the aggregate impacts. We produce tables that differentiate the impacts of enforcements and regulations on the productivity distribu- tion of firms in a given market as well as accounting for firm age. Our sample consists of firms in the same market that were affected by an intervention (treatment group) and those that were not (control group). We cannot fully control for multi-product firms; however, by segmenting ICT into subsectors we get closer to homoge- neous product groupings. We repeat the exercise for both enforcement outcomes and regulatory changes. The direct estimates of ϕ measure the impact of regulation and enforcement on TFPR but not necessarily quantity total factor productivity (TFPQ). However, since we are also estimating impacts on the dispersion in markups, when all else is held constant the difference between TFRP and the markup approximates the impact on TFPQ; this requires that marginal costs remain fixed. Aggregate ICT The first estimates are for aggregate ICT sector. The impact of enforcement reduces the markup dispersion (note that the result is not statistically significant) (Index 3 and 4 of Table 6). TFPR is positive at the aggregate ICT level, but once again not statistically µ significant. The net impact on TFPQ, holding all else constant, is positive (ϕw 1 < τ1 ). This result seems robust regardless of methodology. The impact of enforcement on productivity varies by productivity decile. Both the QLME and ACF methodologies show that TFPR significantly increases for the top 25% of produc- tive firms (which we refer to as the leaders). TFRP for the lower 75% of firms is not statistically significant (Index 1 and 2 of Table 7). This result suggests that strong escape competition effects are at play, where leaders in the market upgrade to escape competi- tion while less productive firms are unable to adjust to competitive forces. The effects on the dispersion in markups tend to not be significant except for firms in the bottom 25 percentile—competitive pressure leads to churning at the lower end of the productivity dis- tribution. This supports the escape-competition channel in the sense that leaders are able to maintain market share in spite of competition and competitive forces create churning mainly for productivity laggards. When the data are further disentangled into firm age and productivity percentiles it appears that leaders again adjust better than laggards to competitive forces (Index 1 and 8 of Table 8). In fact, the results are significant using both QLME and ACF methodologies for all age profiles; the adjustment is largest for middle-aged firms in business for 6–15 years. 18 The impact on the dispersion in markups is generally not significant, except that firms aged 25 and above have higher markups using DLW (Index 8 of Table 9). Regulation policies seem to generate results similar to those of enforcement (Index 1 and 2 of Table 10). They increase productivity regardless of the methodology applied. Interestingly, those that we coded as complexifying generate significant productivity im- provements. In line with our earlier hypothesis, complexifying regulations are related to quality upgrading. Unlike enforcement, regulations that make processes more complex also increase dispersion in markups, although this is not a robust finding across method- ologies. This result can mainly be explained by improvements in quality upgrading leading to higher profits, without any remedial action in terms of marginal costs. As we disentangle these aggregate effects by looking at the firm productivity profiles, we find that leaders in the productivity distribution are more responsive to any regulation in terms of TFPR (Index 1 and 2 of Table 11). In other words, the more productive a firm is compared to its market peers, the more it benefits from regulation. The introduction of any regulation is associated with an increase in TFPR 26%-42% for firms in the 85th percentile compared to that of firms in markets unaffected by any regulation. These results appear to be driven by middle-aged firms, above 5 and below 15 years (Index 1 to 8 of Table 14). The results hold equally well, though they are slightly smaller, for firms whose regu- latory requirements have been simplified. Here a new layer of information is introduced: Laggards in productivity decrease product quality. Apparently low-productivity firms are unable to escape competition when it becomes easier for new firms to enter the market (Index 1 and 2 of Table 12). Disaggregated by age, the most productive middle-aged firms see a robust rise in TFPR while the oldest firms (more than 25 years) instead incur TFPR losses (Index 1 to 8 of Table 16). It is not always more experienced firms that innovate to escape competition. Competitive pressure on older firms seems to lead to productiv- ity losses until ultimately they exit the market; meanwhile, productive middle-aged firms retain or increase market share. The results also hold if we look only at firms whose regulatory requirements became more complex, and they seem to be larger (Index 1 and 2 of Table 13). This type of regula- tion stimulates higher product quality, which enables firms to earn larger profits—evidenced by higher markups (though this has not been confirmed by a DLW robustness check) and more profit measured as the log of price cost margins. It is also possible that in meeting the higher quality standards imposed by the new regulations, firms finance those efforts by charging more, which enables them to raise product quality. Disaggregation by age suggests that older firms are better able to translate more com- plex regulatory requirements into TFPR gains (Index 1 to 8 of Table 18). Markups seem to rise significantly for older market leaders (Table 19). Only experienced and productive firms increase markups, which suggests that they acquire more market power when con- 19 fronted by competitive pressure, unlike firms that are subject to simplifying regulations. Because older firms may have access to more funds and knowledge to conform to new industry standards, they can adapt more easily to complexity. However, when regulation is simplified, the source of competitive pressure changes: new firms can enter a market and compete because barriers to entry are lower. Complexifying and simplifying regulations thus impact productivity differently according to firm age. We also interact the enforcement and regulatory dummies, which allows us to study how much enforcement of competition by Indecopi or Osiptel is enhanced or eroded when a new regulation is introduced. This tells us about how the regulatory environment is improved on paper as well as by actively enforcing competition policy in the market. It also indicates the perverse effects of doing one but not the other, and the extent to which they could cancel each other out: enforcing competitive regulations significantly increases aggregate productivity, but the impact on the dispersion of markups is generally not signif- icant (Index 1 and 2 of Table 20). From the results for aggregate ICT we can infer that competitive pressure supports pro- ductivity, especially for middle-aged firms, across the entire productivity spectrum for com- petitive policies that are enforced. Leaders that can adapt can escape competitive pressures by boosting productivity without necessarily losing market share (as proxied by markup dispersion). The type of regulation also matters: while both complexifying and simplifying types can boost productivity, they are conditional on firm experience, as measured by age. Simplifying regulations support quality upgrading for young productive firms; complexify- ing regulations boost the productivity of more experienced firms. ICT Sectors The Infrastructure Subsector: The few regulations affecting ICT infrastructure mainly apply to bids for the backbone network (see the appendix). Perhaps regulators do not pri- oritize having more players in infrastructure purely to heighten competition, given the high costs of entry into the subsector - i.e., the focus is on competition for the market, and bids are the instrument to promote competition in downstream markets. Instead, they attempt to regulate the natural oligopoly of infrastructure expansion. Thus, to support competition, regulators have focused on promoting access of new firms to existing infrastructure so that they can provide services, e.g., through infrastructure-sharing agreements with network neutrality. A notable example of infrastructure regulation enacted in 2015 reduces administrative barriers to entry: It allows automatic approval of permits so that low-infrastructure projects do not have to wait for construction permits to be approved before starting operations; permits are verified a posteriori through a random selection of projects to be reviewed. While there have been several complaints of anti-competitive behavior, such as abuse of dominance and price-fixing, most decisions have been for the defendant. 20 The regression results for markups are not statistically meaningful, although the point estimates suggest that TFPR increases with competition policy and enforcement (Index 1 and 2 of Table 21). The only meaningful impact is on quality upgrading due to enforce- ment across deciles in the productivity distribution. The youngest and most productive firms significantly raised both TFPR and markups as a result of regulations simplifying re- quirements and enforcement outcomes (Tables 22 and 24). Interestingly, infrastructure is the only sector where innovation upgraded the quality of the service (as measured by our indicator defined above) because of enforced competition, as indicated by an increase in intangible investments and service quality. The more productive a firm, the larger the increase in service quality. The quality measure represents consumer preferences, so im- provements in competition regulation translate into welfare gains for consumers because innovation leads to higher-quality infrastructure. The Telecommunications Subsector: Of the many regulations that apply to telecom- munications. most relate to transparency and promotion of neutrality and non-discrimination principles. Although fixed costs were already high, several regulatory changes raised fees for compliance, which would make it harder for new companies to enter the market. Some enforcement actions resulted in firms being admonished but other firms incurred monetary fines because of unfair competition and misleading information. Similar to the effect ob- served for the whole ICT sector, competition enforcement is associated with a significant increase in TFPR for the most productive, middle-aged firms (Tables 26 and 27). Un- like general ICT trends, telecommunication firms whose TFPR rose apparently also raised markups, particularly firms aged 16 to 25 years (Table 28). Introduction of a new regula- tion is associated with a rise in TFPR and a decrease in markups for market leaders (Table 29).While statistically significant, these results are only confirmed by standard method- ologies when the effects are broken up by age. Robust results for TFPR rises appear only for young firms and for reduced markups only for middle-aged firms (Table 30). However, there are not many observations, which calls for caution in interpreting these results, which should be taken simply as indicative. The Hardware Subsector: Among the most significant policy changes affecting ICT hardware was making available for public procurement competitive private sector offer- ings in an electronic catalog and reducing the costs of switching mobile telephone opera- tors while keeping the same mobile device. Most enforcement findings relate to misleading advertising. Competition enforcement has had hardly any effect on hardware firms (Tables 31-35). except that the least productive firms lowered their product quality, or that con- sumers opt out from these laggards. This situation poses a threat mainly to middle-aged low-productivity firms, which respond by lowering their markups, perhaps in an attempt to offset decreases in quality. Perhaps middle-aged firms that are not close to the hardware technology frontier struggle to compete when pressure builds. However, the limited effects 21 of enforcing hardware competition may also be linked to the reduced size and dynamism of the hardware sector in Peru.6 The Software Subsector: Most enforcement rulings related to software were based on abuse of dominance and price-fixing—typical rulings to discourage monopoly power. Most of the important regulatory changes relate to tax deductions for spending on R&D; better protection of data; and protecting rights to freely choose an internet service provider; thus, the focus is less on competition and more on consumer welfare. Regulation and enforcement have negligible impact on both markup dispersion and productivity in the software market (Tables 36-40). The only significant impact is on wages. A hypothesis not tested here is that the software sector, which is typically made up of small firms at the household or individual level; the sector also has more informal workers than other ICT sectors, and leads to increased demand for laborers without sacrificing profits. Interestingly, the impact of enforcement reduces both markup dispersions and TFPR, with TFPQ being positive when all else is held constant. Enforcement also has little impact on sales and marginal costs but considerably more impact on profits. The reduction in profits correlates with the reduction in markup dispersion. In other sectors we saw markups rise as consumer demand for quality increases, perhaps signaling low elasticity in that market. Here instead we have that consumer preferences decrease as markups rise, pointing to higher elasticity or product substitutability in software products. Unlike in the other sectors, for software we can study the interaction between de facto and de jure regulation that raise the complexity of product market regulations. As a result, we observe an increase in markups and in the profit ratios of market leaders. It may be that leaders just adjust to new regulatory standards by decreasing the number of employees to reduce input costs, or that they simply charge higher prices. The findings suggest that competition policy and enforcement have mainly improved the productivity of the firms that were already most productive (leaders) and those able to escape competition. It does not appear that heightening competition was a major concern of competition policy. However, if the results are productivity enhancement and quality up- grading, that implies a net welfare benefit. The insignificant markup result, however, sug- gests strong escape-competition effects; Leaders escape competitive pressures by increasing productivity or upgrading quality. They retain their market share—especially older, more productive firms. The impacts of competition regulation and enforcement on productivity also hold by sector, though regulation has more nuanced effects. 6 The ICT hardware sector in Peru is very limited in comparison to its South American peers. In 2020, among South American countries, Peru accounted for only 10 percent of computer exports https://oec. world/en/profile/hs/computers 22 5 Conclusion Though ICT is firmly integrated into business operations and personal use, in LAC it is not yet as optimized as in high-income countries. When competitive forces materialize, ICT can stimulate growth spurts by improving productivity. Competitive forces create strong incen- tives for leaders to escape the effects of competition, mainly by heightening productivity and upgrading quality. The direct and spillover effects of ICT productivity increases have meaningful implications for aggregate productivity. Poorly designed regulations and poorly enforced competition policy may not create a truly competitive environment or make a sec- tor more productive. This paper demonstrates that while countries in LAC in have been targeting pro-competitive regulations, distortions still exist. Because the business environment has not yet noticeably reduced these distortions, there is room for more productivity gains. To explore the channels and consequences of competition regulation and enforcement we used a heterogeneous firm model that distinguishes between ICT and non-ICT firms, and within those groups productive from less productive firms. The model accounts for a dispersion in markups across ICT firms. Competitive forces (e.g., lower barriers to en- try) reduce the dispersion of markups within the ICT sector and ICT sub-sectors and lead to a general increase in the market share of ICT firms. Ultimately, an increase in compe- tition raises productivity within both ICT and non-ICT groups, thus improving aggregate productivity. The theoretical findings are supported by firm-level empirical analysis. Data for firms are combined with data on ICT competition regulations and enforcement. The data are split further into ICT subsectors—infrastructure, hardware, software, and telecommunications. Ultimately, de jure and de facto competition does improve productivity, but primarily in firms that were already the most productive in support of the escape-competition hypothe- sis. This result is consistent across firm age groupings. Interestingly, the impact on markup dispersion is not significant across age and productivity groups. Older productive firms have both higher productivity and higher markups when regulations are made more com- plex; younger productive firms benefit from regulations that simplify processes. In several cases, low-productivity firms confronted by competitive pressures have even lower produc- tivity and markups. This suggests that competition occurs leads to churning at the lower end of the productivity distribution, while market leaders, defined as the top 25 percent in that distribution, are able to escape competition and maintain or increase their market shares. The subsector results reveal that the types of regulation and enforcement data col- lected relate to the regulator’s implicit objectives. The objections for infrastructure and telecommunications, as defined by the laws and enforcement rulings, are not necessarily 23 to generate more market entrants, but to mimic competition for the market by stipulating concessions that generate benefits for downstream industries such as connecting industries using the internet backbone. As a result, these sectors see significant quality upgrading and productivity improvements. Where the laws are clear and linked to generating competi- tive outcomes, as are several in the software sector, markups in those industries fall. The results are more ambiguous for the other subsectors because the intensity of enforcement and design of the laws focus on transparency, procedures, and misleading advertising; the result is a complex set of outcomes. Our analysis suggests that the efforts of competition policy to generate competitive forces result in productivity gains—which raises the question: would economic growth be even greater if agencies are adequately staffed, roles more clearly delineated, and enforce- ment broadened? Future work should entail a study on the effects of specific regulatory interventions to distill more specific policy recommendations than the generalized con- cept of promoting competition, such as those initiatives undertaken by the government to enable ex ante regulatory assessments.7 7 see https://www.gob.pe/22241-analisis-de-impacto-regulatorio-ex-ante-air 24 References Ackerberg, D. A., Caves, K. & Frazer, G. (2015), ‘Identification properties of recent produc- tion function estimators’, Econometrica 83(6), 2411–2451. Aghion, P., Bergeaud, A., Boppart, T., Klenow, P. J. & Li, H. (2019), A theory of falling growth and rising rents, Technical report, National Bureau of Economic Research. Alam, A., Casere, Paloma, A. C. & Udompsaph, C. (2008), Unleashing prosperity: Produc- tivity growth in eastern europe and the former soviet union, Staff Report 45627, World Bank. Bartelme, D. & Gorodnichenko, Y. (2015), Linkages and economic development, NBER Working Paper No. 21251, National Bureau of Economic Research. Biagi, F. (2013), ICT and productivity: A review of the literature, JRC No. 2013/09, JRC. Bond, S., Hashemi, A., Kaplan, G. & Zoch, P. (2020), Some unpleasant markup arithmetic: Production function elasticities and their estimation from production data, NBER Work- ing Paper No. 27002, National Bureau of Economic Research. Brynjolfsson, E. & Hit, L. (2010), ‘Beyond computation: Information technology, orga- nizational transformation and business performance’, Journal of Economic Perspectives 4, 23–48. Byrne, D. & Corrado, C. (2017), ‘ICT prices and ICT services: What do they tell us about productivity and technology?’, Finance and Economics 2017/015. Discussion Paper. Campbell, J., Goldfarb, A. & Tucker, C. (2015), ‘Privacy regulation and market structure’, Journal of Economics & Management Strategy 24(1), 47–73. Cette, G., Lopez, J. & Mairesse, J. (2017), ‘Upstream product market regulations, ICT, r&d and productivity’, Review of Income and Wealth 63, 68–89. De Loecker, J., Eeckhout, J. & Unger, G. (2020), ‘The rise of market power and the macroe- conomic implications’, The Quarterly Journal of Economics 135(2), 561–644. De Ridder, M. (2019), ‘Market power and innovation in the intangible economy’. URL: https://www.repository.cam.ac.uk/handle/1810/332313 Dedrick, J., Gurbaxani, V. & Kraemer, K. L. (2003), ‘Information technology and eco- nomic performance: A critical review of the empirical evidence’, ACM Computing Surveys 35(1), 1–28. 25 Doraszelski, U. & Jaumandreu, J. (2019), Using Cost Minimization to Estimate Markups, CEPR Discussion Papers 14114, C.E.P.R. Discussion Papers. URL: https://ideas.repec.org/p/cpr/ceprdp/14114.html Draca, M., Sadun, R. & Van Reenen, J. (2006), Productivity and ICT: A review of the evidence, CEP Discussion Papers dp0749, Centre for Economic Performance, LSE. Dutz, M., R.K., A. & Packard, T. (2018), The jobs of tomorrow” technology, productivity and prosperity in Latin America and the Caribbean, Staff report, World Bank. Edmond, C., Midrigan, V. & Xu, D. Y. (2015), ‘Competition, markups, and the gains from international trade’, American Economic Review 105(10), 3183–3221. Edmond, C., Midrigan, V. & Xu, D. Y. (2018), How costly are markups?, Technical report, National Bureau of Economic Research. Hempell, T. (2005), ‘What’s spurious, what’s real? measuring the productivity impacts of ICT at the firm level’, Empirical Economics 30, 427–464. Hilbert, M. R. & Katz, J. (2003), ‘Building an information society: A Latin American and Caribbean perspective’. Iacovone, L., Perreira-Lopez, M. & Schiffbauer, M. (2016), Competition makes IT better: Evidence on when firms use IT more effectively, Policy Research Working Paper 7638, World Bank. ITU (2017), ‘The challenge on managing digital content’. Paper for the ITU-TRAI regulatory roundtable. ITU (2020), ‘Regulatory tracker’. data retrieved from ITU, https://www.itu.int/net4/ itu-d/irt/. Katz, R. & Jung, J. (2021), ‘The impact of policies, regulation, and institutions on ICT sector performance’, International Telecommunications Union . Khandelwal, A. K., Schott, P. K. & Wei, S.-J. (2013), ‘Trade liberalization and embed- ded institutional reform: Evidence from chinese exporters’, American Economic Review 103(6), 2169–95. URL: https://www.aeaweb.org/articles?id=10.1257/aer.103.6.2169 Lee, F. C. & Shy, O. (1992), ‘A welfare evaluation of technology transfer to joint ventures in the developing countries’, The International Trade Journal 7(2), 205–220. 26 on-Aznar, A., Sellner, R., Brandl, M.-B. & Robinson, C. (2013), ‘Regula- Paterson, I., Rinc´ tion of services industries and ICT diffusion: Accounting for upstream and downstream linkages’, SERVICEGAP Discussion Papers (DP23). Sampi, J., Jooste, C. & Vostroknutova, E. (2021), Identification properties for estimating the impact of regulation on markups and productivity, Policy Research Working Paper Series 9523, The World Bank. URL: https://EconPapers.repec.org/RePEc:wbk:wbrwps:9523 Van Reenen, J. (2010), The economic impact of ICT, Iterim Report 2007-0020, Smart. 27 % 10 20 30 40 50 60 70 80 90 0 100 KSA CHE NDL DEU KAZ CAN POL USA MLT AUS MYS FRA HRV JPN AZE ROU RUS KOR BEL EGY URY MAR ARG CHL TUN 28 TUR CRI BRA MEX ECU DZA JAM PAN COL BOL PER Source: ITU data, author calculations. VNM DOM BWA PRY PHL ZAF Figure 1: Households with Access to Computers THA IDN SLV HND GHA PAK IND KEN BGD ETH RWA % 0 100 10 20 30 40 50 60 70 80 90 KSA CAN KOR SWE BEL NDL USA JPN DEU AUS MYS CHL MLT URY KAZ ARG RUS FRA HUN AZE MAR POL BRA CRI ROU HRV 29 THA TUR DOM PRY MEX EGY TUN CHN VNM ZAF COL Source: ITU data, author calculations. PER ECU PAN BWA DZA BOL GHA Figure 2: Households with Access to the Internet SLV IDN PHL HND IND HND KEN RWA PAK BGD ETH Figure 3: ICT Adoption Rates Source: ITU data, author calculations. Note: Yellow and brown bars correspond to 2014 and 2016, respectively for non LAC countries. 30 Figure 4: Download Speeds Source: ITU. 31 Figure 5: Competition Framework for a Select Number of LAC Countries Source: ITU. 32 Figure 6: Distortions Index ARG BRA CHL COL CRI MEX PER 3 2 1 Distortion Index 0 −1 −2 2005 2008 2010 2012 2015 Source: ITU data, author calculations. Note:This figure is based on data from ITU. The index is generated using standard principal components on data standardized across countries. The variables included are connection fees, broadband wireless charges, broadband tax rates, business telephone installation fees, broadband caps, monthly subscription charges, and broadband speeds. 33 Figure 7: Estimated Magnitude of Distortions, 2005-15 Distortion (IND) relative to KOR BRA CHL COL CRI MEX PER 0.3 Distortion (AGR) relative to KOR 0.4 0.2 0.2 0.0 −0.2 0.1 −0.4 2005 2008 2010 2012 2015 2005 2008 2010 2012 2015 Distortion (SRV) relative to KOR Distortion (ICT) relative to KOR 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 0.0 0.0 2005 2008 2010 2012 2015 2005 2008 2010 2012 2015 Source: OECD data, author calculations. Note: Values range from 1 to -1. A value larger than 0 implies that the country has relatively more input distortions than the frontier country chosen. A value less than 0 implies that a country has fewer distortions than the frontier country. Unlike in the previous figure, input distortions have been falling in ICT while output distortions have been rising. 34 Figure 8: Firm R&D Spending on ICT Source: OECD data, author calculations. 35 Figure 9: ICT Patent Registrations Source: Own calculations using OECD Note: tan and orange bars correspond to 2014 and 2016, respectively for non LAC countries. 36 Figure 10: Markup Distribution Markup and revenue share 0.4 Non-ICT ICT 0.35 0.3 Market share (%) 0.25 0.2 0.15 0.1 0.05 0 1.11 1.111 1.112 1.113 1.114 1.115 1.116 1.117 1.118 1.119 1.12 Firm markup (a) Markup distribution (b) Markup and market share 37 Figure 11: Variations in Markups using Profit Ratios Source: Annual Economic Survey, author calculations. 38 Table 1: The ITU’s ICT Regulatory Tracker Regulatory framework Regulatory mandate Regulatory regime Competition framework G1 Consolidated with policy maker/regulator Business as usual Doing as we have always done State-owned monopoly G2 Separate agency First wave of regulatory reform Doing more Liberalization G3 Separate agency, autonomous in decision making Advanced liberalization of ICT sector Doing the right things Partial competition G4 (85/100) Separate agency with enforcement power (18/20) Adjacent issues become core mandate (11/22) Doing the things right (28/30) Full competition (28/28) G5 Separate agency as part of a network of partner regulators Active collaboration across the board Doing things together Intra-modal competition Note: Number in parenthesis are scores for Peru. 39 Table 2: Parameter Values Parameter Description Values ζ Sectoral productivity 0.51 ¯x ν High productivity (intermediate) 1.16 ¯z ν High productivity (ICT) 1.29 pl Probability of low productivity 0.37 ph Probability of high productivity 0.18 fx Fixed cost of entry (traditional) 0.06 fz Fixed cost of entry (ICT) 0.12 γ Substitution within sector 10.1 θ Substitution between sector 1.34 40 Table 3: Model and Data Moments Moments Data Model Mean market share of top 5% (within sector) 65 % 66% Market share of top 10% 69% 54% Market share of top 25% 84% 85% Market share of top 50% 94% 93% Market share of ICT firms 7.3% 7.7% Fraction of ICT firms 4.1% 4.0% Aggregate markup 10% 11% Median number of firms 1077 1080 41 Table 4: Scenarios Scenario Productivity Gain Number Share of ICT (Relative to Baseline) of Firms ICT Firms Market Share Baseline 0% 4,896 4.0% 7.7% 5% reduction in ICT entry cost 0.1% 4,922 4.6% 8.7% 50% reduction in ICT entry cost 0.9% 5,262 5.3% 10.2% 5% reduction in entry cost 1.1% 5,379 4.3% 8.5% 50% reduction in entry cost 7.0% 9,197 3.1% 6.2% 5% reduction in ICT price (subsidy) 0.5% 4,970 8.7% 24.2% 5% reduction in ICT price (control) -1% 4,438 6.1% 17.7% 42 Table 5: Summary Statistics of ICT Firms Hardware Software Infrastructure Telecommunications Number of firms (2017) 126 77 30 34 Median year of establishment 1998 2000 1999 2000 Wage bill/VA (mean) 64.6% 64.4% 67.5% 66.0% Four-firm ratio (2017) 32.7% 34.7% 50.1% 51.1% Sales/Total ICT (2007-17) 48% 17% 14% 22% Median markup (using profit shares) 1.29 1.20 1.19 1.24 43 Table 6: Aggregate ICT: Enforcement impacts TFPR Markup Profits Sales Employees Wages Intermediate Cost Product quality Intangibles QMLE ACF QMLE DLW 1 2 3 4 5 6 7 8 9 10 11 Any intervention 0.0601 0.0628 -0.0015 0.0286 -0.0238 -0.0332 -0.068 0.0807 0.0468 -0.0756*** -0.0426 Obs. total 2014 2953 2014 2903 2953 2014 2953 2951 2953 2951 1734 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Note: Each column is numbered and corresponds to the index references in text. FE stands for Fixed Effects. 44 Table 7: Aggregate ICT: Enforcement impacts on Productivity Distribution TFPR Markup Quality QMLE ACF QMLE DLW 1 2 3 4 5 15 percentiles -0.173** 0.0113 -0.117 0.0583 -0.377*** 25 percentiles -0.0691 0.0849 -0.140* 0.0194 -0.180** 35 percentiles -0.0632 0.045 -0.130* -0.275* -0.114 45 percentiles 0.014 -0.0425 -0.0723 -0.167 -0.0712 55 percentiles 0.0972 0.0842 -0.0371 0.134 -0.0126 65 percentiles 0.0757 0.244** -0.0499 0.175 -0.00569 75 percentiles 0.184** 0.241** 0.0463 0.0122 -0.0512 85 percentiles 0.248*** 0.413*** 0.075 0.162 -0.0746 95 percentiles 0.186** 0.337** -0.0385 0.112 -0.0473 Obs. total 2013 2953 2013 2903 2951 Firm FE Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 45 Table 8: Aggregate ICT: Enforcement Impacts on Productivity by Age 5 years 15 years 25 years above 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.232 0.735 -0.0571 -0.241 -0.225 0.31 -0.633** 0.168 25 percentiles -0.0282 0.255 -0.00992 -0.00412 -0.163 0.115 -0.288 0.00403 35 percentiles 0.02 0.26 -0.0506 -0.0465 -0.0457 0.0166 -0.461 0.175 45 percentiles -0.0603 0.182 0.072 -0.0696 -0.0856 -0.0441 0.184 -0.22 55 percentiles -0.12 -0.203 0.137 0.145 0.0562 0.024 0.206 0.146 65 percentiles 0.0598 0.243 0.0755 0.272** 0.0822 0.0499 0.0298 0.253 75 percentiles 0.442 0.449 0.138 0.254** 0.154 0.0698 0.321 0.0563 85 percentiles 0.557** 0.236 0.197* 0.437*** 0.239 0.14 0.215 1.190** 95 percentiles 0.568* 0.74 0.164 0.261 0.201 0.0423 -0.236 1.484** Obs. total 2013 2953 2013 2953 2013 2953 2013 2953 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 9: Aggregate ICT: Enforcement Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.0787 0.24 -0.0915 0.0338 -0.147 0.0774 -0.306 -0.0943 25 percentiles 0.0132 0.0321 -0.13 0.068 -0.173 -0.0821 -0.441 0.358 35 percentiles -0.0006 -0.348 -0.147 -0.149 -0.104 -0.372 -0.341 0.534 45 percentiles -0.0495 -0.311 -0.0589 0.00217 -0.0868 -0.241 -0.164 -0.728 55 percentiles -0.0881 -0.436 -0.0453 0.298** 0.0104 -0.0467 -0.116 0.172 65 percentiles -0.0081 0.141 -0.0761 0.22 0.00343 0.0189 -0.154 -0.101 75 percentiles 0.277 -0.0333 -0.00897 0.0384 0.0785 -0.0883 0.0728 0.513 85 percentiles 0.195 -0.156 0.0468 0.203 0.114 -0.108 -0.0889 1.289** 95 percentiles 0.411 -0.0811 -0.0381 0.168 -0.0498 -0.238 -0.595* 1.716** Obs. total 2013 2903 2013 2903 2013 2903 2013 2903 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 46 Table 10: Aggregate ICT: Regulation Impacts TFPR Markup Profits Sales Employees Wages Intermediate Cost Product quality Intangibles QMLE ACF QMLE DLW 1 2 3 4 5 6 7 8 9 10 11 Any regulation 0.0622* 0.0713* 0.0237 0.117* 0.144 -0.0787 -0.0626 -0.0306 -0.0592 -0.0718*** 0.341* Simplifying 0.0163 0.0765 0.0161 0.240*** 0.06 -0.191 -0.232* -0.0778 0.0121 0.0852*** 0.157 Complexifying 0.173*** 0.159** 0.183*** 0.253 0.326* -0.114 -0.194 -0.14 -0.0724 0.195*** 0.189 Obs any 2014 2953 2014 2902 2953 2953 2953 2953 2953 2951 1734 Obs Simplifying 995 1786 995 1773 1786 1786 1786 1785 1786 1785 924 Obs Complexifying 662 1352 662 1347 1352 1352 1352 1351 1352 1351 711 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 47 Table 11: Aggregate ICT: Any Regulation Impacts on Productivity Distribution TFPR Markup Quality QMLE ACF QMLE DLW 1 2 3 4 5 15 percentiles -0.0807 -0.174* -0.0869 -0.268* -0.0992 25 percentiles 0.057 -0.213** 0.00128 -0.131 -0.0273 35 percentiles 0.0668 -0.099 -0.0416 -0.234* -0.0244 45 percentiles 0.0913 -0.0265 -0.0329 -0.192 -0.00744 55 percentiles 0.0832 0.14 -0.0786 -0.0654 -0.0216 65 percentiles 0.154** 0.0332 -0.0318 -0.128 -0.00247 75 percentiles 0.166** 0.118 0.00728 0.0953 0.0289 85 percentiles 0.232*** 0.350*** 0.0826 0.218 -0.0288 95 percentiles 0.284*** 0.176 0.0811 0.247 0.000794 Obs. total 2013 2953 2013 2902 2951 Firm FE Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 48 Table 12: Aggregate ICT: Simplifying Regulation Impacts on Productivity Distribution TFPR Markup Quality QMLE ACF QMLE DLW 1 2 3 4 5 15 percentiles 0.0031 -0.0627 -0.0135 -0.106 -0.210** 25 percentiles 0.0886 -0.184 0.0319 -0.0862 -0.191** 35 percentiles 0.104 0.0471 -0.00855 -0.0005 -0.171* 45 percentiles 0.170** -0.109 0.0487 0.0845 -0.142* 55 percentiles 0.129 0.153 0.00983 0.148 -0.131 65 percentiles 0.183** 0.0423 0.0674 0.0697 -0.144* 75 percentiles 0.186** 0.0314 0.101 0.124 -0.0786 85 percentiles 0.214*** 0.281** 0.109 0.235 -0.0984 95 percentiles 0.275*** -0.0736 0.137* 0.0645 -0.0711 Obs. total 995 1786 995 1773 1785 Firm FE Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 49 Table 13: Aggregate ICT: Complexifying Regulation Impacts on Productivity Distribution TFPR Markup Quality QMLE ACF QMLE DLW 1 2 3 4 5 15 percentiles -0.13 0.501 -0.158 1.033*** -0.144 25 percentiles 0.366 -0.0344 0.202 -0.209 0.316 35 percentiles 0.333 -0.181 0.244 -1.080* 0.211 45 percentiles 0.0881 -0.174 0.0311 -0.936 0.0174 55 percentiles 0.211 0.0695 0.0865 -1.381 0.112 65 percentiles 0.31 0.318 0.104 -0.106 0.229 75 percentiles 0.360** 0.171 0.123 0.552 0.223 85 percentiles 0.521*** 0.205 0.415*** 0.217 -0.11 95 percentiles 0.545*** 0.511* 0.268 2.075* 0.0262 Obs. total 662 1352 662 1347 1351 Firm FE Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 14: Aggregate ICT: Any Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.0392 -0.476* -0.205** -0.0388 0.0246 -0.188 0.419 -0.167 25 percentiles 0.105 -0.278 -0.0374 -0.0886 0.0938 -0.0257 0.26 -0.994*** 35 percentiles 0.122 -0.105 -0.0261 0.181 0.0691 -0.142 0.428* -0.663** 45 percentiles 0.218 0.122 0.0108 0.0995 0.0952 -0.00661 0.28 -0.498* 55 percentiles 0.256 0.14 0.000925 0.19 0.0889 0.0695 0.318 -0.0525 65 percentiles 0.273 0.19 0.0995 0.0381 0.127 0.0502 0.358 -0.606 75 percentiles 0.645*** 0.384 0.0333 0.0591 0.199 0.266** 0.246 -0.171 85 percentiles 0.639*** 0.132 0.200** 0.392*** 0.174 0.351*** 0.082 -0.104 95 percentiles 0.623*** 0.143 0.255** 0.370** 0.164 0.395* 0.325 -0.892 Obs. total 2013 2953 2013 2953 2013 2953 2013 2953 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 50 Table 15: Aggregate ICT: Any Regulation Impacts on Markups by Age 5 years 15 years 25 years above 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.0721 -0.435 -0.185** -0.391** 0.0787 0.102 -0.125 -0.266 25 percentiles 0.085 -0.88 -0.108 0.0582 0.161 0.0549 -0.0366 -0.681 35 percentiles 0.101 -0.25 -0.151 -0.182 0.0687 -0.186 0.056 -0.251 45 percentiles -0.084 -0.0479 -0.0934 -0.137 0.102 -0.0789 -0.0205 -0.439* 55 percentiles -0.0852 -0.277 -0.0818 0.096 -0.0746 -0.167 -0.0545 -0.193 65 percentiles -0.0717 -0.373 -0.0182 0.0615 -0.05 -0.229 -0.014 -0.375 75 percentiles 0.121 -0.145 -0.047 0.0241 0.0355 0.0278 0.0965 0.588 85 percentiles 0.142 0.278 0.0825 0.0181 0.0482 0.376 0.0852 0.323 95 percentiles 0.158 -0.237 0.0856 0.173 0.0867 0.698* -0.102 -0.00226 Obs. total 2013 2902 2013 2902 2013 2902 2013 2902 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 16: Aggregate ICT: Simplifying Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.108 -0.936** -0.0549 0.489* 0.134 -0.113 -0.232 0.0896 25 percentiles 0.0977 -0.47 0.0892 0.00467 0.138 -0.137 -0.126 -0.177 35 percentiles 0 -0.0845 0.0852 0.112 0.0964 -0.00231 0.123 0.0252 45 percentiles 0.133 -0.179 0.149 -0.135 0.179 0.128 0.436 -0.344 55 percentiles 0.15 -0.373 0.066 0.271** 0.233 0.282* 0.303 -0.322 65 percentiles 0.144 -0.337* 0.147 0.298** 0.261 0.155 0.337 -0.426* 75 percentiles 0.393** -0.491*** 0.109 0.0862 0.208 0.266** 0.182 0.0772 85 percentiles 0.426** -0.249 0.196* 0.414*** 0.294* 0.288* -0.496* -0.372* 95 percentiles 0.403** -0.588** 0.240** 0.232* 0.277 0.649*** 0.0996 -1.547* Obs. total 995 1786 995 1786 995 1786 995 1786 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 51 Table 17: Aggregate ICT: Simplifying Regulation Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.0194 -0.0654 0.0352 0.0334 -0.071 -0.186 -0.361 0.0493 25 percentiles -0.0123 0.689** 0.0793 0.0333 0.0201 -0.320* -0.181 -0.371 35 percentiles 0 0.457 0.0221 -0.0316 -0.0623 -0.164 -0.132 0.487** 45 percentiles -0.0718 0.259 0.0686 -0.172 0.0981 0.179 0.13 0.0884 55 percentiles -0.083 -0.0411 0.0102 0.301** 0.0733 -0.0342 0.0962 -0.318 65 percentiles -0.0861 -0.0823 0.115 0.28 0.0603 -0.131 0.119 -0.306 75 percentiles 0.084 -0.339 0.111 0.254* 0.0779 0.106 0.12 0.0405 85 percentiles 0.143 0.502 0.175* 0.137 0.138 0.114 -0.548** -0.235 95 percentiles 0.073 -0.385 0.170* 0.137 0.134 0.424 -0.0748 -0.386 Obs. total 995 1773 995 1773 995 1773 995 1773 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 18: Aggregate ICT: Complexifying Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.495 - -0.366 -0.0374 -0.246 0.235 2.521*** 1.368*** 25 percentiles 0 -0.396*** 0.0294 0.299 0.963 0.283** 0.692* -0.969*** 35 percentiles 0 0.308 0.102 -0.117 0.271 -0.253** 0.630* -0.206 45 percentiles 0 0.712 0.0474 -0.169 0.0432 0.0455 0.22 -0.265 55 percentiles 0 0.712 0.226 0.0255 0.215 -0.449 0.215 0.841 65 percentiles 0 0.712 0.208 0.287 0.381 0.0771 1.316 0.841 75 percentiles 1.025** - -0.0254 0.0829 0.708** 0.206 1.913 0.227 85 percentiles 1.084** 0.711*** 0.405** 0.156 0.329 -0.00604 0.634** 0.734 95 percentiles 0.451 0.711*** 0.426* 0.974*** 0.451 -0.514*** 1.259** 0.734 Obs. total 995 1773 995 1773 995 1773 995 1773 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 52 Table 19: Aggregate ICT: Complexifying Regulation Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.493 - -0.322 0.166 -0.058 1.133*** 0.894 1.442*** 25 percentiles 0 -3.582*** -0.0362 0.928 1.079 0.868** 0.379 -1.830*** 35 percentiles 0 1.191*** -0.193 -1.03 1.210** -2.222*** 0.453 -0.867 45 percentiles 0 -2.864*** -0.105 -0.727 0.134 -1.536** 0.195 -0.37 55 percentiles 0 -2.864*** 0.117 -1.214 -0.0987 -1.05 0.325 4.982*** 65 percentiles 0 -2.864*** 0.216 -0.0485 -0.162 -0.258 2.042 4.982*** 75 percentiles -0.164 - 0.236 -0.193 -0.108 -0.124 2.312** 2.176** 85 percentiles -0.347 0.673* 0.488*** -0.183 -0.074 -0.751** 0.559** 5.190*** 95 percentiles 0.0333 0.673* 0.387* 1.888 -0.16 0.0864 0.495 5.190*** Obs. total 662 1347 662 1347 662 1347 662 1347 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 20: Aggregate ICT: Interacting Impacts on Productivity Distribution TFPR Markup Quality QMLE ACF QMLE DLW 1 2 3 4 5 15 percentiles -0.09 -0.282* -0.0652 0.0389 -0.234** 25 percentiles -0.0272 -0.0181 -0.0864 -0.0268 -0.172* 35 percentiles 0.0323 -0.0136 0.118 -0.105 -0.067 45 percentiles 0.0685 0.0385 -0.0157 0.00691 0.000937 55 percentiles 0.0489 0.282** 0.0205 -0.091 -0.12 65 percentiles 0.0957 0.267** 0.0318 -0.0804 0.0353 75 percentiles 0.176** 0.291** 0.137* 0.244* 0.0917 85 percentiles 0.314*** 0.472*** 0.147** 0.29 0.139** 95 percentiles 0.325*** 0.236 0.116 0.166 0.148* Obs. total 2013 2953 2013 2901 2951 Firm FE Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 53 Table 21: Infrastructure: Enforcement Impacts TFPR Markup Profits Sales Employees Wages Intermediate Cost Product quality Intangibles QMLE ACF QMLE DLW 1 2 3 4 5 6 7 8 9 10 11 Any intervention 0.441* -0.362 0.0257 0.524*** 0.217 0.161 0.139 -0.825*** 0.174 -0.0863*** 0.232*** Obs. total 172 352 172 349 352 352 352 352 352 352 192 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 54 Table 22: Infrastructure ICT: Enforcement Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.947 0.00514 0.103 -1.314 -0.325 0.267 -1.112 0.267 25 percentiles 1.435 -14.07*** 0.249 -1.302 -0.365 0.937 -1.112 1.93 35 percentiles 1.435 -2.181 0.195 -1.027 -0.397 0.679 0.329 1.93 45 percentiles 1.367 -1.518** 0.232 -1.714* -0.381 -0.352 0.567 1.949 55 percentiles 2.99 -1.028 0.402 -1.805 -1.052 -0.824 0.567 1.949 65 percentiles 2.446 -1.147 0.222 0.307 -0.828 -0.947 0.0808 1.949 75 percentiles 2.368** 0.191 0.308 0.176 0.00465 0.0501 -1.096 1.009 85 percentiles 2.389** 0.729 0.514 -0.191 0.267 0.0501 -1.514 1.009 95 percentiles 2.448** 0.729 0.47 -0.191 0.264 -0.265 -1.514 1.009 Obs. total 172 352 172 352 172 352 172 352 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 55 Table 23: Infrastructure ICT: Enforcement Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 2.270* 5.116 0.123 2.437 -0.525 -0.0379 -0.94 1.462 25 percentiles 1.696* 16.92*** 0.156 0.244 -0.503 1.632* -0.94 8.209*** 35 percentiles 1.696* 23.16 0.116 -0.78 -0.67 2.862 -0.327 8.209*** 45 percentiles 0.396 2.355** 0.251 0.886 -0.548 -0.128 -0.0549 5.700** 55 percentiles 0.211 0.618 0.313 3.153* -1.096* -1.158 -0.0549 5.700** 65 percentiles 0.306 0.419 0.212 0.875 -0.852 -1.359 -0.325 5.700** 75 percentiles 1.515 1.188 0.00627 0.591 -0.699 -1.219 -1.122 1.08 85 percentiles 1.639* 1.22 0.136 1.302 -0.642 -1.219 -1.246 1.08 95 percentiles 1.644* 1.22 0.237 1.302 -0.64 -0.741 -1.246 1.08 Obs. total 172 349 172 349 172 349 172 349 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 24: Infrastructure ICT: Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.131 -0.739 0.0633 -0.798 0.243 -0.207 -0.983 -0.366 25 percentiles -0.0712 -2.670** 0.134 -0.527 0.21 0.0763 -0.0517 -0.324 35 percentiles -0.0882 1.41 0.199 0.578 0.264 -0.0851 0.45 -0.0764 45 percentiles 0.233 -1.231* 0.116 0.715 0.306 -0.349 0.451 -0.076 55 percentiles 0.488 -1.241* 0.133 0.266 0.309 -1.016 -0.00514 -0.121 65 percentiles 0.433 -1.054* -0.118 -0.101 0.339 -1.177* -1.087 -0.354 75 percentiles 2.112** 0.836 -0.121 -0.529 0.393 -0.616 0.343 1.347** 85 percentiles 2.146** -0.855 0.331 -0.0285 0.351 -0.232 0.343 1.347** 95 percentiles 2.146** -0.855 0.331 -0.242 0.355 -0.347 0.457 1.347** Obs. total 172 352 172 352 172 352 172 352 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 56 Table 25: Infrastructure ICT: Regulation Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.562 1.324* 0.082 1.533** 0.186 0.361 -0.624 1.143 25 percentiles 0.677 2.251* 0.0815 0.782 0.253 0.94 -0.28 1.954* 35 percentiles 0.677 9.504*** 0.0472 -0.34 0.349 0.545 0.0939 0.849 45 percentiles 0.0623 2.058 0.171 0.0559 0.355 0.465 0.0938 0.851 55 percentiles 0.256 2.181 0.145 0.748 0.395 0.247 -0.202 0.921 65 percentiles 0.245 1.546 -0.192 0.616 0.442 1.189 -0.854 1.808* 75 percentiles 1.769* 2.9 -0.223 1.151 0.351 1.384 0.0134 1.304*** 85 percentiles 1.903* -1.611 0.0509 0.987 0.135 0.743 0.0134 1.304*** 95 percentiles 1.903* -1.611 0.0509 1.729** 0.239 1.084 0.055 1.304*** Obs. total 172 349 172 349 172 349 172 349 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 57 Table 26: Telecommunications: Enforcement Impacts TFPR Markup Profits Sales Employees Wages Intermediate Cost Product quality Intangibles QMLE ACF QMLE DLW 1 2 3 4 5 6 7 8 9 10 11 Any intervention 0.287** 0.355*** 0.181 0.234 0.308 -0.0307 -0.253 0.293 0.212 -0.0463 -0.0649 Obs. total 250 391 250 390 390 390 390 390 390 390 241 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 58 Table 27: Telecommunications ICT: Enforcement Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.484 2.037*** 0.0675 -0.939 0.0888 1.127 0 0.963* 25 percentiles -0.683 1.948*** 0.323 -0.235 0.417 0.719 0 -4.817*** 35 percentiles -0.486 0.385 0.247 -0.353 0.402 0.0364 -0.442 7.841*** 45 percentiles -1.125 1.161 -0.105 -0.505 0.416 -0.164 0.00493 -4.817*** 55 percentiles -1.046 1.188 -0.329 -0.241 0.543 -0.0524 0.00493 - 65 percentiles -0.592 1.184 0.00414 -0.148 0.488 -0.502 0.155 - 75 percentiles 0.087 0.448 0.318 -0.0635 0.606* -0.156 0.155 - 85 percentiles -0.67 -0.241 1.088*** 0.861** 0.606* 0.276 0.155 - 95 percentiles -1.125 1.157 0.914*** 0.745** 0.543 0.363 0.155 - Obs. total 250 391 250 391 250 391 250 391 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 59 Table 28: Telecommunications ICT: Enforcement Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.328 0.157 0.338 2.612** -0.25 0.195 0 2.051 25 percentiles 0.178 0.0254 0.479 0.0135 -0.246 0.9 0 1.918*** 35 percentiles 0.265 -0.461 -0.256 -0.369 -0.236 -0.453 -0.985 -0.201 45 percentiles 0.244 0.217 -0.702 -0.769** -0.213 -0.666 0.236 1.918*** 55 percentiles 0.166 0.194 -0.71 0.311 0.221 0.337 0.236 - 65 percentiles 0.522 0.134 -0.679 0.945 0.221 0.28 0.432 - 75 percentiles 0.208 0.119 -0.396 0.268 0.882** 0.444 0.432 - 85 percentiles 0.435 -2.848 0.425 0.896 0.882** 0.680** 0.432 - 95 percentiles 0.244 0.157 0.554 0.489** 0.221 0.177 0.432 - Obs. total 250 347 250 347 250 347 250 347 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 29: Telecommunications ICT: Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 2.941*** 4.938*** -0.132 0.113 0.57 -2.144* 0 2.623* 25 percentiles 2.561*** 4.377*** -0.423 -0.578 1.272 -1.968** 1.136 2.623* 35 percentiles 2.709*** 2.421 -0.274 -1.332 0.846 -1.343* 1.136 -25.29 45 percentiles 2.648*** 2.217 -0.484 -1.814 0.201 -0.685 -0.541 -25.29 55 percentiles 2.534*** 4.027*** -0.627 -1.632 0.0478 -0.399 0.463 - 65 percentiles 2.011** 4.024*** -0.359 -0.232 0.0956 -1.13 0.463 - 75 percentiles 2.842*** 4.021*** -0.308 -1.102 -0.721 -1.951** -1.509 -2.961*** 85 percentiles 2.814*** 0.833 0.797** -0.64 -0.663 0.016 -1.509 -2.961*** 95 percentiles 2.777*** 0.0602 0.941** -1.55 0.445 0.016 -1.509 - Obs. total 250 391 250 391 250 391 250 391 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 60 Table 30: Telecommunications ICT: Regulation Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.334 12.45*** -1.052** -4.852** 1.097 0.105 0 -0.0967 25 percentiles 0.432 3.993 -0.951* -0.446 0.73 0.831 0.557 -0.0967 35 percentiles 0.185 -4.046** -0.953* -0.544 0.533 3.384** 0.557 8.389 45 percentiles 0.00861 -1.791* -1.110* 0.977 -0.1 0.154 2.081 8.389 55 percentiles -0.109 -0.0125 -1.170* 2.748** -0.541 -0.947 1.363 - 65 percentiles 0.0635 -0.1 -0.622 0.482 -0.208 0.111 1.363 - 75 percentiles 0.15 3.009 -0.990** 0.307 0.000575 0.168 -2.522 2.044* 85 percentiles 0.0907 2.063 -1.456*** 0.637 0.0563 0.547 -2.522 2.044* 95 percentiles 0.0396 -0.939 -1.560*** 0.321 0.0451 0.547 -2.522 - Obs. total 250 346 250 346 250 346 250 346 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 61 Table 31: Hardware: Enforcement Impacts TFPR Markup Profits Sales Employees Wages Intermediate Cost Product quality Intangibles QMLE ACF QMLE DLW 1 2 3 4 5 6 7 8 9 10 11 Any intervention 0.0244 0.0164 -0.0402 -0.0357 -0.0913 0.0127 -0.0383 0.0739 0.0944 -0.0718*** -0.0346 Obs. total 1174 1626 1174 1624 1626 1626 1626 1626 1626 1626 908 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 62 Table 32: Hardware ICT: Enforcement Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.0291 0.134 -0.101 -0.273 -0.0772 -0.0388 -0.720** 0.733* 25 percentiles 0.0665 -0.25 -0.0101 0.0273 -0.0521 -0.0496 -0.336 1.483 35 percentiles 0.134 0.166 0.0446 -0.0668 0.0728 0.0116 -0.417 1.605* 45 percentiles -0.0488 0.0635 0.022 -0.175 0.0605 -0.0606 -0.372 0.239 55 percentiles 0.00773 -2.247*** 0.0108 0.0771 0.118 0.184 -0.0407 0.675 65 percentiles 0.135 0.079 -0.0403 0.204 0.13 0.164 0.00139 0.739 75 percentiles -0.0966 0.337 0.059 0.213 0.119 0.0277 0.778 1.39 85 percentiles -0.0553 -0.15 0.0252 0.384*** 0.0197 -0.231* 0.939 1.181** 95 percentiles 0.294 0.663 -0.0975 0.176 0.283 -0.338** -0.709 3.036*** Obs. total 1174 1626 1174 1626 1174 1626 1174 1626 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 63 Table 33: Hardware ICT: Enforcement Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.19 -0.196 -0.237** -0.504* 0.00776 0.236 -0.252 0.177 25 percentiles -0.0441 -0.14 -0.134 0.0567 -0.133 -0.357 -0.396 1.052 35 percentiles -0.113 0.288 -0.13 -0.309 0.0213 -0.518* -0.316 1.331** 45 percentiles -0.111 0.427 -0.103 -0.0776 0.0924 0.168 -0.308 0.78 55 percentiles -0.181 -1.464*** -0.115 0.092 0.093 0.331 -0.184 0.698 65 percentiles -0.0884 0.454 -0.165 0.173 0.191 0.335 -0.0143 0.289 75 percentiles -0.233 0.792 -0.0781 -0.14 0.117 0.279 0.824 1.558 85 percentiles -0.213 -0.0176 -0.124 0.0935 0.0428 -0.521* 0.787 1.727 95 percentiles -0.0455 0.362 -0.254** -0.106 0.264 -0.624** -0.939** 5.798** Obs. total 1174 1626 1174 1626 1174 1626 1174 1626 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 34: Hardware ICT: Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0.239 -0.343 -0.0713 0.0906 0.214 -0.271 0.219 0.0559 25 percentiles -0.0161 0.115 0.145 0.598 0.199 -0.155 -0.00552 -0.456 35 percentiles 0.13 -0.837 0.0273 0.628* 0.0897 -0.288 0.388 -0.402 45 percentiles 0.144 -0.343 0.0476 0.367 0.0852 0.484 0.06 -0.371* 55 percentiles 0.326 -0.629 -0.244 0.233 0.154 0.148 0.432 0.399 65 percentiles 0.259 -1.169* 0.241 -0.278* 0.21 0.115 0.0925 0.494 75 percentiles 0.778* -1.799*** 0.13 -0.33 0.111 0.317 -0.0479 0.0734 85 percentiles 0.846* -0.804* 0.18 0.512 0.177 0.337 -0.417 -0.363 95 percentiles 0.662 -2.052*** 0.159 0.522 0.249 0.312 -0.146 -2.012*** Obs. total 1174 1626 1174 1626 1174 1626 1174 1626 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 64 Table 35: Hardware ICT: Regulation Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.502 -0.436 -0.239 0.0407 0.131 0.187 -0.121 -0.478** 25 percentiles -0.592 0.889 -0.154 0.656** 0.21 -0.0468 -0.0568 -0.0147 35 percentiles -0.486 0.157 -0.0919 0.576* -0.11 -0.332 0.00717 0.721** 45 percentiles -0.313 0.062 -0.113 -0.0972 0.0981 0.619** -0.119 -0.0688 55 percentiles -0.138 -0.488 -0.293* 0.141 0.095 0.304 0.152 0.0589 65 percentiles -0.174 -1.089** 0.0367 -0.419** 0.0619 0.365 0.339 -0.089 75 percentiles -0.059 -1.756*** 0.0567 -0.247 0.126 0.378 0.31 -0.0336 85 percentiles 0.0536 -0.542 0.137 0.244 0.0871 0.580** -0.329 -0.248 95 percentiles -0.0904 -1.971*** 0.0477 -0.111 0.218 0.598** -0.441 -1.574*** Obs. total 1174 1626 1174 1626 1174 1626 1174 1626 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 65 Table 36: Software: Enforcement Impacts TFPR Markup Profits Sales Employees Wages Intermediate Cost Product quality Intangibles QMLE ACF QMLE DLW 1 2 3 4 5 6 7 8 9 10 11 Any intervention -0.119 -0.00053 -0.140* 0.0983 -0.0599 -0.279* -0.0497 -0.0607 -0.223 -0.0723*** -0.165 Obs. total 386 584 386 583 584 584 584 584 584 584 393 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 66 Table 37: Software ICT: Enforcement Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0 1.199 -1.076 0.23 0.155 -0.0348 -0.174 - 25 percentiles 1.354 - -0.241 0.109 1.812 0.136 0 -0.442*** 35 percentiles 0 -0.906*** -0.990*** 0.0993 0.248 -0.193 0 -0.442*** 45 percentiles 0 -0.906*** -0.197 -0.149 0.318 0.109 0 0.463 55 percentiles 0 - -0.172 -0.303** 0.377 -0.365 0 - 65 percentiles 0 - -0.161 -0.479 -0.931 0.0109 5.25 -0.389 75 percentiles 0 -9.074* 0.468 -1.135** 0.534 0.119 -0.132 - 85 percentiles 1.779* -5.181*** 0.214 0.243 0.562 0.416*** 0 2.112*** 95 percentiles 0 -6.511*** 0.4 0.915** 0.424 0.397*** 0 - Obs. total 385 584 385 584 385 584 385 584 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 67 Table 38: Software ICT: Enforcement Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles 0 0.544 0.000312 0.434 -0.242 0.397 0.411 - 25 percentiles 0.474 - -0.384 -0.385 -0.914 -0.563 0 1.026*** 35 percentiles 0 -7.820*** -0.861*** 1.708 -0.0826 -1.161* 0 1.026*** 45 percentiles 0 -7.820*** -0.339 1.899 0.581 -0.651 0 -1.428 55 percentiles 0 - -0.156 1.400*** 0.315 -1.351* 0 - 65 percentiles 0 - -0.02 0.31 -0.616 0.126 2.514 -4.569*** 75 percentiles 0 -38.06*** 0.710** -2.639*** 1.104 1.416 -0.126 - 85 percentiles -0.936 -5.070* 0.726*** -0.497 0.488 0.977 0 3.157*** 95 percentiles 0 -2.497 0.428 0.202 0.349 0.963 0 - Obs. total 385 583 385 583 385 583 385 583 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. Table 39: Software ICT: Regulation Impacts on Productivity by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -1.485 -0.0403 -0.312 -0.295 -0.19 0.173 0.484 0.27 25 percentiles -1.554* -0.0142 0.203 -0.0233 0.612 0.405 1.058* 3.098*** 35 percentiles -1.514 1.417* 0.166 -0.480* 0.101 0.405 0.902* 0.781** 45 percentiles -0.721 0.0589 0.042 -0.909*** -0.083 0.226 0.736 -0.0843 55 percentiles -1.674 -0.744 0.0501 -0.147 0.175 0.478 0.0284 - 65 percentiles -1.699 0.024 0.123 0.237 0.178 0.033 2.223 -0.311 75 percentiles 0.408 0.508*** -0.329 -0.223 0.131 0.271 0.987 4.905 85 percentiles 0.0182 -0.499 0.0306 -0.167 0.299 0.843** 0.0252 0.703 95 percentiles -1.442 3.712*** 0.216 -1.588* 0.286 0.800** -0.692 13.93*** Obs. total 385 385 385 385 584 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 68 Table 40: Software ICT: Regulation Impacts on Markups by Age 5 years 15 years 25 years More than 25 years QMLE ACF QMLE ACF QMLE ACF QMLE ACF 1 2 3 4 5 6 7 8 15 percentiles -0.987 -0.449 -0.117 -0.831** -0.444 0.29 0.588 -1.268 25 percentiles -1.037 -2.546*** 0.00157 1.012 0.53 0.219 0.387 5.060*** 35 percentiles -0.985 -0.0171 -0.207 -0.189 0.441 -0.139 0.482 0.498 45 percentiles -0.537 -0.559 -0.16 -0.511 -0.0533 -0.25 0.306 0.479 55 percentiles -0.877 -3.717 -0.0562 -0.371 0.0253 0.985 0.213 - 65 percentiles -0.886 -0.779 -0.00061 0.235 0.182 0.385 1.801 -1.033** 75 percentiles -0.426 0.888*** 0.0274 -0.339 -0.169 -0.37 1.274 5.621* 85 percentiles -0.398 1.645 0.367** -0.433 -0.0799 1.214 0.188 1.647** 95 percentiles -0.792 3.945*** 0.0892 -1.75 -0.146 2.480*** -0.294 13.17*** Obs. total 385 583 385 584 385 584 385 583 Firm FE Yes Yes Yes Yes Yes Yes Yes Yes Sector*Year FE Yes Yes Yes Yes Yes Yes Yes Yes State*Year FE Yes Yes Yes Yes Yes Yes Yes Yes *** p < 0.01, ** p < 0.05, * p < 0.1. 69 Appendix A Appendices A.1 ICT costs 70 Table 41: Costs Combared, Brazil, Colombia, Mexico and Peru Country BRA COL MEX PER Mean Broadband Cap 0 0 0 0 Mean Broadband Connection Fee (USD) 0 22.26 0 73.65 Mean Broadband Speed (Mbit/s) 0.76 0.61 3.00 0.70 Mean Broadband Subscription Fee (monthly) 54.13 50591.50 305.63 89.30 Mean Broadband Tax Rate 40.15 16.00 16.00 18.00 Mean Connection Fee (USD) 30.33 18.32 108.08 107.66 Mean Installation Fee (telephone USD) 30.33 53.24 112.82 138.21 Mean International Bandwidth (Mbit/s) 942666.67 258145.50 603403.97 84177.67 SD Broadband Cap 0 NaN 0 0 SD Broadband Connection Fee (USD) NaN NaN 0 NaN SD Broadband Speed (Mbit/s) 0.35 0.15 2.83 0.42 71 SD Broadband Subscription Fee (monthly USD) 34.41 22780.86 75.12 42.02 SD Broadband Tax Rate NaN 0 0 NaN SD Connection Fee (USD) 12.33 25.91 10.41 45.70 SD Installation Fee (telephone) 12.33 NaN 9.06 8.16 SD International Bandwidth (Mbit/s) 1057100.44 222911.88 816373.77 99716.22 A.2 Peruvian ICT Competition Laws summarized A.2.1 Compilation of ICT competition laws 8 8 While not used in the analysis, we subjectively classify the potential of pro competition rules to be strong and weak. If a law encourages clearly market entry, or benefits for connecting industries, then those would be categorized as strong. 72 Table 42: Summarized competition laws ICT branch Publication Name Disposition Competition OECD- Effect checklist Telecom op- 2002 Resolution Obligation to publish the rates for public or private calls for Pro, low C2, D1 erators No. 048-2002- bids or negotiations CD/OSIPTEL Telecom op- 2006 Resolution Sets the formula for fixing rates regarding access and shared Pro, low B1 erators NO. 008-2006- use of public use infrastructure for the provision of public CD/OSIPTEL telecommunications services Hardware 2008 Legislative De- Calls private sector offers on specific products and services, Pro, A3 cree N.1018 and the most competitive ones are selected to be included in high(market an electronic catalog for public procurement entrance) Telecom 2008 Resolution No. Mobile number portability. Establishes procedures, economic Pro, high A3, D2 operators; 044- 2008- and technical conditions for the operation of mobile number (market Hardware CD/OSIPTEL portability. entrance) 73 ICT general 2008 Legislative de- General Law of Free Competition Pro, high all cree 1034 ICT general 2008 Legislative de- General Law on Unfair Competition Pro, high all cree 1044 Telecom op- 2008 Legislative De- Obliges an operator that holds the category of Major Provider Pro, high A3 erators; In- cree 1019 in a certain market to share its infrastructure at a market (market frastructure price. entrance) Telecom op- 2008 Resolution Complements Legislative Decree 1019 by setting procedures Pro, medium A3 erators; In- NO. 020-2008- and deadlines for sharing infrastructure from the major (market frastructure CD/OSIPTEL provider. Facilitates competition by allowing operators with entrance) lesser infrastructure to use the network of other operator to provide their service. Telecom op- 2009 Supreme Obliges an operator that holds the category of Major Provider Pro, medium A3 erators; In- Decree N° in a certain market to resell its traffic at a reasonable price. (market frastructure 002-2009-MTC entrance) Software 2009 Decree N.005- Prohibition to contact individuals registered in a list to pro- Against, A3, B2 2009/COD- mote products and services, yet does not cover the use of in- medium INDECOPI dividual information to target ads on digital platforms. Access (market to the list requires registration with a registration fee entrance) Telecom op- 2010 Supreme The evaluation of concession renewal becomes more complex Against, A2 erators Decree 036- (includes a factor linked to the number of violations commit- medium/high 2010-MTC ted, increasing the unpredictability to renew concessions). (market en- The concession renewal methodology generates several ad- trance) ministrative procedures for both OSIPTEL and the companies Software 2012 Resolution Complements Decree N.005-2009/COD-INDECOPI by includ- Against, low A3, B2 N° 159-2012- ing into the list microentrepreneurs, removes registration fee INDECOPI/COD and removes the requirement for costumers to register every two years, setting now a permanent registration Infrastructure 2012 Law N° 29904 Facilitates infrastructure deployment for broadband services: Pro, high A3 regulates the construction of the National Backbone Fiber Op- (market tic Network. entrance) 74 Internet 2012 Resolution Telef´ onica del Per´ u and companies of its economic group are Pro, high A3 Access No. 132-2012- set as Major Provider in Fixed Internet (wholesale market): (market Provider CD/OSIPTEL they must submit a Basic Sharing Offer and a Basic Resale entrance) Offer regarding traffic for the Fixed Internet service and for their public telecommunications services Telecom op- 2012 Resolution Consolidates the technical, economic, legal and procedural Pro, low A2, A3 erators; In- No. 134-2012- provisions on which the supervised negotiation scheme is frastructure CD/OSIPTEL based. Allows operators to freely agree on their intercon- nection relations and OSIPTEL to guarantee an adequate and timely interconnection. Telecom op- 2013 Supreme Public bid for infrastructure expansion made Osiptel respon- Against, high A3, B1, erators; In- Decree 014- sible for setting uniform unflexible tariffs for the use of in- (price distor- B2, B4 frastructure 2013-MTC frastructure by operators. Tariffs for each megabyte per sec- tions) ond (Mbps) are applied nationwide and independent from the level of usage. Internet 2013 Supreme Public service concessionaires with relevant infrastructure Pro, high A3, A4 Access Decree 014- will facilitate access and use to public telecom service con- (market Provider; 2013-MTC cessionaires for the deployment of their networks. Respect of entrance) Telecom Network Neutrality. 75 A.2.2 De-jure competition regulation framework in Peru The regulatory framework is actively being reviewed to strengthen competition law. With major changes in 2008, 2015 and 2018, anti-competitive practices and unfair competition have been further regulated across all sectors, with stronger role from the competition authorities to investigate and sanction anti-competitive behavior, along with the ability to carry out studies aimed at removing bureaucratic barriers. More recently, merger control law was strengthened in January 2021 (Law N 31112), as merger notifications to the competition authority became mandatory to cover all sectors as opposed to only the energy sector. The competition authority may also act ex officio if it is suspects dominant position risks. Obligations affecting the telecommunications and infrastructure sector are most often generated by the legislative branch through a Law or Legislative Decree (a norm with the rank of Law), which needs to be approved and regulated by a Supreme Decree issued by the MTC. The MTC might sometimes require Osiptel to issue a complementary norm, addressing missing technical details for the implementation of the obligation. Osiptel might also directly generate regulatory obligations through a Resolution of the Board of Directors, which is then approved and regulated by the MTC through a Supreme Decree. This process might cause a gap between the moment of issuance of the obligation via Law and the its entry into force. A.2.3 Infrastructure regulation Without adequate infrastructure, the deployment and use of hardware and software is limited. Regulations in Peru favor competition, but some impediments, such as high entry barriers still exist. The infrastructure landscape (ownership of network towers and cables) are dominated by only a few firms. Infrastructure was mainly owned by the state, until private sector participation was encouraged in the 1990s. Privatization started in 1992 to sell the state-owned enterprises ENTEL, in charge of local telephony across the country except for Lima, and Compa˜ n´(i)a Pe- ruana de Tel´ onica efonos, focused on the capital. Three main bids and two years after, Telef´ Internacional de Espa˜ na won the concession contract which involved specific network ex- pansion and service quality improvement targets. These investment commitments have been a feature of most Peruvian privatizations. Open market policy was put in place from 1998, inviting private sector participation in the development of infrastructure and the de- ployment of network services. Since the early 2000’s, the market has remained controlled under the hands of the same players: Telef´ onica (now called Movistar, with around 80.74% of the market share for fixed broad band and 40.7% for mobile broadband), followed by Claro (America Movil, with up to 20% of the market share for fixed broad band and 32.4% 76 for mobile broadband), Nextel (now called Entel) and Viettel (also known as Bittel). Note that there was some dynamism in the market with the entry of Entel and Bittel. These four players have their own infrastructure, plus there is another small opera- tor called Inkacel that uses third party infrastructure. Recent efforts strive to promote infrastructure-sharing for telecommunication operators development as well as for internet providers through Mobile Virtual Operators regulation. Telefonica is the largest owner of infrastructure, and since 2012 it was determined as Major Provider in Fixed Internet, oblig- ing it to share its infrastructure at a market price (Resolution No. 132-2012-CD/OSIPTEL). Viettel is the only one among the four that owns the totality of the fiber it uses and does not rent it to third parties. Nevertheless, some infrastructure is still owned by the government. For instance, the SOE Electropuno SAA or Machupichu SA 9 own electric towers or medium voltage transmission lines, which are for rent under infrastructure sharing agreements. The ”backbone network” has been the latest most important infrastructure project en- couraging private sector deployment of infrastructure and provision of telecom services to under served areas. The public bid introduced in 2012 (Law N. 29.904) and approved by Supreme Decree in 2013 (N. 014-2013-MTC), aims at connecting all provinces and dis- trict capitals. The MTC was set in charge of the project design as well as to determine the applicable tariffs. The design was twofold; connecting provincial capitals first, then re- gional projects would expand the network to district capitals. A Private-Public-Partnership co-financing mechanism was put in place, such that the concessionary firm would be in charge of the network construction, maintenance and operation. The firm would be a neutral operator, that is, offering the service of transporting data by other operators, but not offering telecom service to final demand. The State, in exchange, guaranteed a fixed income level for the bid winner, independent to the actual sales level. That is, the state would cover the difference between the current sales and the fixed income level by disburs- ing quarterly payments; thereby de-linking performance to payment and disincentivizing optimal behavior from Azteca. The MTC determined a uniform price for each megabyte per second (Mbps), which would be a nationwide tariff independent from the level of usage and was determined before the realization of the public bid. This level was set without a regulatory procedure, yet it was set with the operational support from Ospitel - the regulatory entity of telecom- based on data and estimates provided by the MTC (e.g. cost and demand estimates), which in retrospect reflected an outdated market reality since these did not contemplate neither the adoption of mobile internet nor the arrival of 4G. The bid opened in May 2013 with the disclosure of its rules, and finished in 2014, won by the Mexican company Azteca Comunicaciones, apparently the only bidder, which agreed 9 https://www.osiptel.gob.pe/repositorioaps/data/1/1/1/par/contrato-infra-americamovil-electropuno2012 contrato-infraestructura-CLARO-Electropuno-2012.pdf. 77 to set a maximum and unique tariff at the national level (US$23 per MB), independently from the level of mbps usage. By the time it started deploying its services, starting opera- tions from late 2016, other operators had already entered the market, offering lower tariffs and capturing the demand that was originally thought to be captured by the bid winner. The arrangement had not considered the possibility that other private operators would build their own networks and expand their services. At the end, private operators were indeed Azteca customers, but only on small stretches as they used their own infrastructure for most of the routes. The entrance of new players might have been incentivized by the fact that Azteca was only able to provide one service with one fee; whereas competitors enjoyed flexibility to attract demand by offering discounts per volume level or per contract duration time. Therefore, counter intuitively, the lack of flexibility due to over-regulation to the bid winner incentivized competition. A.2.4 Rules regarding the use of broadband Broadband expansion appears to be highly regulated and controlled by the regulatory agen- cies, yet there is a clear effort to promote broadband development through reinforced com- petition. Instances of it include the Law for the promotion of broadband and the construc- tion of the backbone network in 2012, infrastructure-sharing in 2013 and virtual operator measures in 2016. The latter sets a stable framework to allow operators without network infrastructure to enter markets through contracts based on the principles of neutrality, non- discrimination, equal access, non-exclusivity, free and fair competition. In order to achieve universal access to broadband across districts, instead of developing a SOE, the MTC sets conditions when renewing licenses related to standards on quality and coverage level of service. An example was the condition for Telefonica to provide more than 12,000 free Internet broadband connections for public institutions such as education, healthcare and security. Spectrum issuance is done via licenses. The ITU defines three models for assigning spec- trum. (i) Command and control model, which implies allocation according to a primary market, based on administrative allocations; (ii) Market-oriented model, which includes the organization of auctions, secondary markets and sharing arrangements; and, (iii) Gen- eral use model, which implies full liberalization (limited to non-interference principles) in the use of spectrum. The MTC originally adopted the first model of allocation based on the primary mar- ket. In 2019 it recognized the need to update it and adapt it to technological change and growing demand. The MTC therefore introduced the Supreme Decree N. 015-2019-MTC to regulate the leasing of spectrum following the second model of secondary market. It is expected that the temporary leasing would allow optimization of the spectrum usage, in- creased flexibility in its allocation and thereby market development. Such decree presents 78 a possible barrier in Article 14, by establishing the obligation to set an Operations Support System between the leessor and the tenant. This enables the MTC to monitor the spectrum user obligations and raises administrative costs for the parties involved. There is also a fee for the use of radioelectric spectrum aimed at financing the cost of supervision and control of the spectrum. This is considered as a duplicated cost, since there is also a fee paid to the MTC for commercial operation, while the largest burden of spectrum supervision is born by Ospitel. Municipalities seem to be the main actors imposing restrictions, arising from a lack of coordination between Ospitel, the MTC and local governments, but efforts have been undertaken to tackle this. 10 For instance, district level restrictions on the deployment of base stations seem to be linked to the low levels of base station density per capita compared to the rest of the region, Asia and Europe. Given the high level of autonomy enjoyed by local authorities, the approach taken to cut down on local bureaucratic barriers has been to encourage regulatory improvements. For instance, Ospitel published in 2015, 2017 and 2018 the District Connectivity Index, rewarding municipalities that did not impose any burden on infrastructure deployment (e.g. Linked to municipal authorizations and requisites), to facilitate infrastructure expansion. Instances of good practices measured in the index are the adaptation of local administrative procedures for the deployment of telecom infrastructure; municipal management through the perception of the four mobile operators on the municipal facilities for infrastructure deployment; and service quality. Negative externalities may arise from other public entities generating further obliga- tions. The Ministry of Environment requires environmental impact assessment of any broadband deployment project over 200 meters. This applies to most broadband deploy- ment projects, and besides the monetary cost, it delays work of around four months ap- proximately, until the certificate is delivered. For instance, deployment of fiber optic lines within the city of Lima need to be evaluated to verify that the line does not overlap with any buffer zone of a natural protected area (Directorial Resolution 321-2017-MTC/16 321- 2017-MTC/16). Nevertheless, the environmental benefits might offset the negative exter- nalities of the project delays. The sector appears to be historically over-regulated, and frequent changes seek to im- prove its efficiency but reduce the predictability of the regulatory framework. As an exam- ple, the Conditions of Use and Portability Regulations has changed three times in a same year, and is the regulation linked with the largest number of sanctions over the last two decades. Sectoral fees paid by operators highly exceed those paid in Europe, Ecuador and Colombia (the independent study only takes those two LAC countries as regional compara- tors). Fees entail a contribution for regulation (”Aporte por regulacion”) paid to Ospitel 10 2014 GSMA interview to Head of legal and regulatory affairs of the Association for the Promotion of National Infrastructure (AFIN). 79 (0.5% of annual sales), a commercial operations rate paid to the MTC (0.5% of annual sales) and a contribution to the Telecommunications Investment Fund ”FITEL” (1% of an- nual sales). This is on the top of a Radio Spectrum Use Fee, which purpose is to cover the cost of monitoring and controlling the radio spectrum. A.2.5 The granting of special licenses Special licenses are required to deploy telecom infrastructure, but efforts have been taken to speed up the process and paperwork burden (e.g. Law N. 30228 from 2014, approved by the supreme decree D.S. 003-2015-MTC of 2015). Previously, permits were required to be approved before undertaking any infrastructure work, and only if no response was obtained after 30 days did the principle of positive silence apply. The new law applies the principle of automatic approval, where infrastructure projects submit their permits and can start undertaking work without requiring such approval. The verification of the authenticity of documents is done at posteriori through a random selection of projects to be examined, imposing a large fine in case of infraction. The new law does not change fees and states that processing fees should correspond to the real cost born by the administration to issue the permit. License renewal for the use of spectrum has been an opportunity for the government to achieve its goal of expanding universal service access without recurring to a SOE, though its long and non-linear process has brought uncertainty for future renewals. License tend to cover terms of around 20 years, and since the privatization of the sector in the late 1990s, the main license renewal took place in 2011 for Telefonica, regarding the 850 and 1900 MHz bands for another 18 years. The renewal process took 18 months because, despite having already negotiated the terms of contract, the election of a new administration in Peru brought about new objectives and terms had to be renegotiated. The renewal did not involve a one-time payment of a fee but rather many conditions, considered as harsh and unprecedented by Telefonica.11 Among them are the commitment to invest 1.2 billion USD over 5 years; free internet in government institutions such as schools, hospitals, police stations; and expansion of coverage to all district capitals, including the rain forest area which is currently under-served. 12 A.2.6 On the rules regarding tie-in sales Companies used to restrict access to telecom services from the same brand phone devices, until MNP was implemented in 2010 and reinforced in 2014, along with measures for unlocking mobile devices from fixed plans as well as for unsubscribing from an operator 11 https://www.rcrwireless.com/20130123/carriers/telefonica-peru-finds-harsh-accepts-governments-condit 12 GSMA Licence renewal in LATAM 2014. 80 with a click. It only applies to Peruvian companies, and a lack of compliance is subject to fines.13 This has allowed a reduction in exchange costs and contributed to the improvement of competition in the mobile market, according to the Competition commission, Ospitel. Ospitel has had only one case against one of the main telecom company which was selling internet data services conditional on buying also their telephone services. OSPITEL won the case and a rule to prevent this was implemented from 2013. The procedure initiated in 2011 against Telef˜ u S.A.A. for the alleged commission of abuse onica del Per´ of dominant position, under the modality of tie-in sale: conditional sale of ADSL internet service to the purchase of its fixed telephony service. Such tie-in sale strategy was in place since 2001, yet it particularly became problematic from 2007 when demand for fixed telephone lines stagnated and internet demand continued to rise. This abuse of dominant position might have been linked to negative consequences from a scheme setting maximum levels of tariffs for telephone services that was still in place around 2007. This scheme was incentivizing the creation of tie-in sales, where companies were able to cut down costs on one product by compensating it with higher prices from the tied products sold. OSPITEL declared the case to be grounded in evidence, imposing a fine of 492.21 UITs and ordering a corrective measure consisting in offering the sale of the ADSL in- ternet service ”only”. The Dispute Resolution Court confirmed the OSPITEL’s resolution, setting the fine at 407 UITs in 2011/2012.14 The telephone company filed a contentious- administrative complaint against the resolution, calling for its nullity. It seems that the fine was finally imposed in 2017.15 A.2.7 The treatment of international companies There does not seem to be any preferential treatment for foreign companies over national ones, overall. Free Trade Zones and Special Treatment Zones (which must be established by law), such as the of Amazonia or high Andean zones, enjoy the exoneration of some taxes (e.g. Sales tax) as well as preferential customs duties on the import of goods destined for those areas or which enter through them.16 While these benefits aim at incentivizing for- eign direct investment, they do not discriminate between national and foreign companies. One bureaucratic barrier preventing FDI in these special zones is the Article 71 of the Con- stitution, which establishes that within 50 kilometers of the border, foreign investment is not allowed for acquisition or possession. This has posed a problem for special zones such as the city of Tacna, that aims to attract FDI yet gives preferential treatment to national 13 https://www.osiptel.gob.pe/noticia/osiptel-desbloqueo-celulares-usuarios-nacional-internacional. 14 https://www.osiptel.gob.pe/documentos/81309-expediente-n-0052011 and https://www. osiptel.gob.pe/Archivos/Publicaciones/memoria_osiptel_web_2012.pdf. 15 Procedure N. 005-2011-CCOST/LC :https://www.osiptel.gob.pe/noticia/ ndp-sancionan-movistar-incumplir-prohibicion-ventas-atadas-servicios. 16 https://home.kpmg/content/dam/kpmg/pe/pdf/Inversiones-en-Per%C3%BA-2019ES.pdf. 81 investments due to this constitutional barrier.17 Peru has bilateral Double Taxation Agreements that gives preference to certain coun- tries. Spain for instance does not have such an agreement, so companies that are not domiciled in Peru and that provide services to Peruvian companies would be subject to a 15-30% income tax, if the service is considered as technical assistance. This means that a Peruvian company would have to retain the 15-30% income tax if it were to get services from a non-domiciled company that does not belong to the list of countries with which Peru has such agreements, thereby giving preference to companies from certain countries over others. A.2.8 Competition regarding hardware Peru has an entity called ”Peru Compras”, created in 2008 by the Legislative Decree N.1018 within the Ministry of Economics and Finance, in charge of making sure public procure- ment is efficient, transparent and competitive. It organizes calls for the private sector to make bids on specific products and services (including all hardware purchases); these are reviewed, and the most competitive ones are selected to be included in an electronic cat- alog for public procurement, within Framework Agreements (”Acuerdos Marco”). These catalogs have a list of benefits for both public administration as well as for suppliers.18 To prevent tie-in sales, Peru approved in 2013 the Mobile number portability (MNP), which enables mobile telephone users to retain their mobile telephone numbers when changing from one mobile network carrier to another (Supreme decree N.003-2007, Res- olution N. 166-2013-CD/OSIPTEL). While it is not a restriction, it incentivizes consumers to change chip-reading devices, further reinforced in 2014 with a new scheme that allows processing the MNP within 24 hours, fully automated. There are not many barriers to the sale of chip-reading devices, but the main restriction was imposed in 2012 and in 2017 whereby sellers were requested to verify the buyer’s identity before selling the device in order to prevent the sale of stolen phones (Resolution N. 138-2012-CD/OSIPTEL , Supreme Decree N. 023-2014-MTC)19 . That involved operators paying a fee to access the National Registry of Identification and Civil State in order to make such biometric verification, besides the costs of operating the verification at points of sale. The group AFIN estimated the cost overrun to be around $39 million USD between 2013 and 2018. Similarly, to prevent mobile phones to be stollen, operators became required to register mobile phones sold in a white list, creating a supplementary burden (Supreme decree N. 009-2017-IN). 17 https://gestion.pe/economia/zofratacna-huawei-grandes-tecnologia-limitadas-invertir-tacna-270557-not ?ref=gesr. 18 https://www.perucompras.gob.pe/acuerdos-marco/cuales-son-los-beneficios.php. 19 https://hiperderecho.org/2019/06/identificacion-biometrica-obligatoria/. 82 A further constraint for the sale of cellular devices is the certification requirement from the MTC. In 2020, the MTC announced that such devices need to be re-certified with new GSMA TAC letters (what do the acronyms mean), align with a system for cell broadcast diffusion, and that prior certificates were no longer valid. This appears to restrict imports of cheap, lower quality cell phones that do not meet the requirements (e.g. OEM phones). The MTC justifies this as necessary to protect the security of users and to block highly falsifiable mobile devices.20 A.2.9 Competition rules regarding software A first set of regulations related to software was promulgated in 2003 to determine the rules for public administration use of software (DS 013-2003-PCM), followed by technical norms for efficient use of software in public administration. The first one is approved in 2004 by the Ministerial Resolution No. 179-2004-PCM. It describes the structure of the software’s life cycle and its processes but does not specify details on how those processes should be implemented. Another technical norm in 2009 comes to address this gap. It is only until 2015 that these technical norms are compiled and approved by a directorate resolution. The compilation of the processes for software development clarifies each step and could be considered as a regulatory improvement for transparency and a first step towards the simplification of regulatory burden. There seems to be no specific regulation for software industries development, but rather for promoting investment in scientific or technological innovation. For instance, the Na- tional Policy Document for the Development of Science, Technology and Technological Innovation from CONCYTEC lists relevant regulations and none relate directly to software but rather overall to the promotion of investment in technology and science. Within such framework, the main regulation that seems to affect costs is a tax reduction for investment in scientific, technological and innovation research (Law N. 30056; article 37; from 2013). In the same vein, in 2016, another law complements this effort by strengthening the incen- tive through subsidies for investment in R&D, innovation, etc. The subsidy comes in the form of tax deduction that can go up to 215% of expenses incurred in projects from 2016 to 2019 (Law N. 30309). The competition agency registers all patents in a database free for access.21 The database query allows to filter patents using IPC codes.22 For instance, for the class G06 within ”Com- puting, calculating or counting”, there were 498 invention patents submitted between 1953 20 http://www.perusmart.com/mtc-restringe-importacion-celulares-baratos-peru/. 21 https://www.indecopi.gob.pe/web/invenciones-y-nuevas-tecnologias/ buscadores-de-patentes. 22 https://www.wipo.int/classifications/ipc/ipcpub/?notion=scheme&version=20200101& symbol=none&menulang=en&lang=en&viewmode=f&fipcpc=no&showdeleted=yes&indexes=no&headings= yes¬es=yes&direction=o2n&initial=A&cwid=none&tree=no&searchmode=smart. 83 and 2019. Among these, only 86 related to the subclass ”digital computing or data pro- cessing systems or methods specially adapted for administrative, commercial, financial, managerial, supervisory or forecasting purposes”, and were submitted between 2005 and 2019. For ”Information storage” (G11), 29 invention patents were submitted between 1953 and 1996, whereas the subclass ”ICT specially adapted for specific application fields” has had no submissions. This database could be used for instance as a performance indicator of the promotion of competition and investment in scientific, technological and innovation research through tax reduction and then subsidy implementation in 2013 and 2016. A report from the Spanish Embassy in Peru finds that the largest consumer of software solutions in Peru is the financial sector, followed by HR departments, laboratories, phar- maceutical companies, and mining and energy sector, given that most of Peru’s exports are related to mining. It seems that, in 2013, 85% of companies in software sector in Peru were SMEs, mainly multinationals implanted in the country, such as Oracle, IBM, Microsoft, HP, Epson or Dell. A.2.10 Digital platforms and data collection Individual data rights are referred to in the Peruvian Constitution of 1993, regarding the protection of individual data and the conditions to access it when public entities are the supplier of information, thereby omitting the role of private sector as a collector or supplier of individual data. The Constitution allows individuals to ask any public entity to share in- formation, with the exception of personal information. It also states that communications, telecommunications or its instruments can only be open, investigated and interfered if mandated by a judge. The first reference to individual data collected by private entities in Peruvian legislation comes in 2001 (Law N. 27489) and refers to information related to credit reports gathered by financial institutions, therefore not necessarily covering non-financial personal informa- tion. In 2009, a law restricts the use of individual information for advertising and marketing, but not the collection of private data. It creates a register of phone numbers and emails, such that it becomes forbidden to contact registered individuals through their personal information to promote products and services (Decree N.005-2009/COD-INDECOPI). This refers to telephone calls, text messages and emails, therefore not including the use of individual information to target advertisement on digital platforms. Companies wishing to access the list to know which persons not to contact have to register and renew their registration every two years, thereby creating a bureaucratic barrier for the advertisement service industry. Personal data use and collection from both private and public entities is regulated 84 through the publication of a data protection Law N.297333,23 published in 2011 and ap- proved in 2013 (Supreme Decree N.003-2013-JUS). The law allows entities to collect data that is relevant and pertinent for an explicitly stated purpose, in a legal and transparent way, under consent of the individual. It forbids the use of data for purposes other than those stated explicitly, unless data becomes anonymized. It regulates the flow of informa- tion across borders such that the third country receiving the information need to have data protection levels adequate to the Peruvian law, if not, the party sending the information becomes the warrantor for data protection according to Peruvian law. The data collector needs to register with the National Authority of Transparency and Access to Public Information.24 This agency issues directives as well as binding technical opinion to draft bills regulating personal data. It exercises administrative, guiding,25 nor- mative, resolution, fiscal and sanctioning functions. Google has already been sanctioned once, as well as the main telephone provider company in Peru, with very large sums, yet the bulk of sanctions are against national private and public bodies with relatively small fines.26 The data protection law was only modified in 2017 by a legislative decree (N.1353)27 to strengthen the Personal Data Protection Regime. It expands the scope to cover more explicitly not only owners of databases containing private information but also to cover those processing such databases. The scope of the National Authority of Transparency and Access to Public Information is also expanded such that it needs to register not only the existence of databases but also information related to the cross-border flows of personal data. When Facebook’s weak data protection policy went viral and Mr. Zuckerberg went to trial in 2018, Peru was also debating further reinforcing data protection measures. It was considering the possibility to ease the ability to revoke individual’s consent to the use of private information at any point in time. A.2.11 The treatment of taxis and user-apps The Congress of Peru started drafting projects of law to regulate the taxi industry by mobile applications since 2016 (e.g. 1505/2016-CR, 2218/2017-CR), in order to combat informal- ity and provide security for passengers. In 2018, it approved a bill that creates a National Registry of taxi companies that operate with such platforms. The registry will cover infor- 23 https://leyes.congreso.gob.pe/Documentos/Leyes/29733.pdf. 24 https://www.minjus.gob.pe/registro-proteccion-datos-personales/. 25 Example of directive for the application of the data protection law: https://www.minjus.gob.pe/ wp-content/uploads/2013/11/Directiva-de-Seguridad-DGPDP.pdf. 26 List of sanctions up to 2019: https://www.minjus.gob.pe/wp-content/uploads/2019/11/ Cuadro-de-publicacio%CC%81n-de-Sanciones-ANPDP.pdf. 27 https://www.gtdi.pe/Actualizacion_LPDP-1-2017. 85 mation of all the taxi companies that operate through applications in the market (eg. Uber or Easy Taxi) as well as information about its drivers, their licenses and their Obligatory Insurance against Traffic Accidents (SOAT) documentation. ”Up to 2018, no legal action has been taken against Uber’s operations. The Provincial Municipality of Callao started the inspection process against the drivers of vehicles performing transport services by means of applications using cell phones, without authorization from the municipality”.28 In 2019 apps are explicitly being targeted for regulation, with two bills being drafted for the taxi and delivery services, one focused on the labor aspect related to social security, the other focused on all taxi services in the municipalities of Lima and Callao (from both traditional and through applications). ”The use of SOAT, Single Card for circulation, valid driver credentials and the use of GPS during trips are some of the requirements that must be met by any authorized operator. In addition, of course, to the census. Once the regulation is approved, a period of adaptation of no less than one year will be given to comply with all the conditions.”29 28 https://repositorio.up.edu.pe/bitstream/handle/11354/2197/Wendy_Tesis_Maestria_208. pdf?sequence=4&isAllowed=y. 29 https://elcomercio.pe/economia/dia-1/apps-en-el-peru-cuales-son-los-retos-pendientes-en-cuanto-a-re ?ref=ecr. 86