New York Post
보수 성향
스포츠
부정적Knicks’ NBA Finals Game 3 report card: Jalen Brunson’s inefficient series finally hurts
Report card from the Knicks’ 115-111 loss over the Spurs in Game 3 of the NBA Finals on Monday night at the Garden.
"INEFFICIENT" · 총 10건
필터 보기현재 지수
49.5
0 = 부정 우세
50 = 중립
100 = 긍정 우세
최근 7일 기준 82,216건을 분석한 결과, 뉴스 심리지수는 49.5(균형)입니다. 긍정 10,224건(12.4%)·중립 59,393건(72.2%)·부정 12,599건(15.3%)이며, 중립 비중이 뚜렷하게 높습니다. 성향 지수는 종합 20.1(보수 경향)입니다.
Report card from the Knicks’ 115-111 loss over the Spurs in Game 3 of the NBA Finals on Monday night at the Garden.
Green hydrogen has been hyped as a silver bullet solution to decarbonize hard-to-abate sectors like steelmaking and industrial shipping for decades now. It can be combusted at high heats like fossil fuels, but leaves behind nothing but water vapor when it burns, rather than world-heating greenhouse gas emissions. So why does green hydrogen still represent less than one percent of all hydrogen production in the United States? The simple answer is that producing green hydrogen is expensive, and usually an inefficient use of the clean energy resources…
The government must stop shifting the cost of weak revenue mobilisation onto households and the corporate sector and instead offer targeted tax relief to offset the burden imposed in recent years, including a reduction in the petroleum levy. While support for the most vulnerable remains necessary given high poverty levels, sustained job-creating growth is vital. It is unreasonable to tax a monthly income of Rs50,000, which falls below the amount required for a family’s subsistence. To make the tax regime more logical and equitable, the income tax threshold should be raised to Rs1.5 million per annum (Rs125,000 per month) from the current Rs600,000. The tax slabs and rates should then be recalibrated accordingly to preserve progressivity while providing meaningful relief to low-income earners. At the same time, there is little justification for imposing a super tax on the already compliant corporate sector while large segments of the economy — including many services, retail and wholesale trade, real estate, and farm landowners — continue to remain undertaxed or effectively enjoy a tax holiday. With inflation once again edging upward, the persistently high petroleum levy is adding to the cost pressures across the economy. The levy needs to be rationalised and gradually reduced to levels comparable with regional averages to provide much-needed relief to consumers and businesses alike. ‘Attempting to extract more taxes from an already stressed private sector is likely to generate frustration and resentment rather than meaningful additional revenues’ Measures to broaden the tax base by effectively bringing big property owners and traders into the federal tax net, while ensuring that provinces adequately tax agricultural income and other undertaxed service providers, could not only offset the revenue loss from providing relief to overburdened taxpayers but also generate substantial additional revenues. “A wider and more equitable tax base would improve compliance, reduce distortions, and strengthen fiscal sustainability without placing further pressure on already heavily taxed segments of society,” said a retired Federal Board of Revenue officer. Meanwhile, revenue targets should be set realistically, considering the near-stagnant state of the economy, where economic growth is barely keeping pace with population growth. Under these circumstances, greater emphasis should be placed on reducing wasteful administrative spending and rationalising the costs of an oversized and inefficient state apparatus. “Sizeable increase in tax revenues is rarely achieved in a low-growth environment,” observed a tax expert who requested anonymity. “Attempting to extract more taxes from an already stressed private sector is likely to generate frustration and resentment rather than meaningful additional revenues. It could further undermine business confidence, discourage investment, and deepen the economic slowdown at a time when the country can least afford it.” The government will need to use the budget to convince the public that it is not only cognisant of the mounting economic pressures on households and businesses but is also committed to addressing rising poverty and inequality, while facilitating the private sector for accelerating GDP growth. More importantly, it must demonstrate a credible strategy to lift growth to the levels capable of generating sufficient productive employment for the country’s expanding workforce and improving living standards on a sustained basis. The spending patterns witnessed during Eid, where a small segment of society reportedly spent millions on sacrificial animals, in a country where half the population remain below or near the poverty line, underscored the widening gap between the affluent and the struggling majority. Growing frustration among the youth over limited economic opportunities, coupled with widening income and wealth disparities, is increasingly viewed as a source of political and social risk not only for the government of the day but also for the country’s fragile democratic order and broader institutional framework. Some observers caution that unless the upcoming budget sends a clear signal that the government is committed to expanding opportunities, reducing barriers to upward social mobility, and addressing economic exclusion, public discontent could intensify. Failure to tackle these underlying grievances may further erode trust in institutions and increase the risk of social unrest. “We dread a Bangladesh-like situation if mounting economic grievances remain unaddressed. Our platforms are not merely advocating the interests of businesses; we are also urging the government to safeguard the economic rights of citizens and provide tax relief to the middle class,” remarked a leading Karachi-based business leader while explaining the budget proposals submitted to the government. The reference was to the 2024 turmoil in Bangladesh, widely referred to as the “July Uprising”, a massive, student-led movement that toppled Prime Minister Sheikh Hasina’s government. Many analysts view it as a reminder of how economic pressures, perceptions of nepotism and inequality, and limited opportunities can amplify public discontent and trigger wider political instability. Official estimates place Pakistan’s poverty rate at 28.9 per cent of the population. However, a recently released report by the Social Policy and Development Centre paints a bleak picture, suggesting poverty incidence at 43.5pc in 2024-25, with urban poverty rising at a faster rate. The report also points to a widening income gap. According to its findings, inequality increased by 12pc between 2018-19 and 2024-25, with deterioration more pronounced in urban centres. Members of Prime Minister Shehbaz Sharif’s economic team were approached for their views on the concerns raised in this report. While some chose not to comment ahead of the budget, the responses of others had not been received by the filing deadline. Published in Dawn, The Business and Finance Weekly, June 1st, 2026
Country: Democratic Republic of the Congo Source: ELRHA Author Jennifer O’Keeffe, Augustin Gang Karume and Paul Spiegel This blog series accompanies the Mortality Estimation Systems Innovation Partnership (SIP), supported by UKHIH-Elrha, which brings together diverse partners to strengthen how mortality data is collected, interpreted, and used across humanitarian crises. Earlier blogs in this series highlighted why excess mortality measurement is critical for understanding crisis severity, as well as exploring how to maximise local and national actors' leadership in the mortality estimation ecosystem. In this third blog, we turn to Eastern Democratic Republic of the Congo, where Rebuild Hope for Africa and the Johns Hopkins Center for Humanitarian Health share how their work is making mortality estimation more accurate, accessible, and feasible for national actors best placed to do this work, even in the most challenging settings. “As an indicator, a mortality rate tries to evaluate the size and scale of a crisis in a single metric.” The Public Health Aspects of Complex Emergencies and Refugee Situations, 1997, Michael Toole, Ronald Waldman In 2023, the Humanitarian Congress in Vienna released a statement saying, "The humanitarian imperative is an absolute moral obligation to save lives and alleviate human suffering on the basis of need, without discrimination”. Yet**,** when resources are constrained, allocation is often based on geopolitical interests, media coverage, or how relatable a population may be to high-income donor countries. In short, human lives are valued differentially. The disconnect is not theoretical. In 2022, Rebuild Hope for Africa (RHA) led a nationwide mortality survey in the Central African Republic which estimated up to 5% of the population had died during the previous year. Despite the scale of these findings, the study received little media attention and did not lead to meaningful changes in donor policy. In conflict-affected settings, various, often compounding, factors make primary data collection difficult or impossible. These include forced displacement, insecurity, system failures, poor infrastructure, limited capacity, and restricted access. In practice, mortality is often not measured at all. And as threats to healthcare workers grow, international agencies have become understandably risk averse, collecting data only safer, accessible areas, where death rates are usually lowest. Without reliable data, decision makers and responders depend on fragmented sources and non-robust estimates. The result is a biased and misleading picture of crisis severity, that often portrays crises as less severe than they are. The magnitude of these biases and their effects on decisions by humanitarian actors, governments, and donors who rely on such data, remain largely unexamined. Our partnership between Rebuild Hope for Africa (RHA) and the Johns Hopkins Center for Humanitarian Health (CHH) is working to change this. Eastern Democratic Republic of the Congo - An Unquantified Crisis Few places demonstrate the challenges of mortality estimation more than the Democratic Republic of the Congo (DRC), one of the world’s most enduring humanitarian crises. The crisis worsened drastically in January 2025 when the country suffered a devastating double shock: the abrupt withdrawal of USAID funding and a violent military offensive by the Rwandan-backed rebel group M23. The M23 seized large swathes of territory, killing and displacing an unknown number of people in the process. With the departure of many international agencies and a vacuum in humanitarian response, the population has been left vulnerable to the worst effects of the conflict. A year later, the true human cost remains unknown. We recognise that without reliable data, it becomes even harder to mobilise the support that people living in Eastern DRC urgently need. Placing Data and Decision-Making in Congolese hands Augustin Gang Karume, one of the authors of this blog, was born and raised in Eastern DRC, where he still lives and works today. In 2008, he founded RHA to place data and decision-making back in Congolese hands. He understood then that national actors are the future of sustainable humanitarian response. Rooted in the community and living with the long-term consequences of decision-making, national actors have a strong incentive to prioritise community needs over institutional agendas. Using local networks and knowledge, they are the best equipped to conduct primary data collection in insecure settings. While international actors have scaled back amid funding austerity, national organisations like RHA have remained in place, continuing to work for and within their communities. These actors are also proving to be far more cost-effective and efficient. Without international overhead, they can often deliver results at a fraction of the cost of international organisations. As an example, RHA’s 2022 nationwide mortality survey in the Central African Republic, cost a total of 50,000 USD, whereas a single district SMART survey may cost upwards of 15,000 USD*. National actors are the first responders in nearly all crises and remain present long after international attention and funding fade. Bridging Local Leadership with Technical Expertise With funding from the UK Humanitarian Innovation Hub’s Systems Innovation Partnership, we are bridging RHA’s local leadership with technical expertise from the CHH, combining community trust with advanced epidemiological and statistical training. Together RHA and CHH are collaborating on a study to assess potential biases in mortality estimation through both primary data collection and innovative use of statistical approaches. We’re working to make mortality estimation more accurate, credible, and efficient, with the intent to apply the findings across humanitarian settings. In the primary data collection component, our study is comparing three different methods of mortality estimation: a retrospective household survey, rapid key informant listing, and a full census. Using a common reference population and recall period, the study aims to identify where biases arise, quantify which deaths are missed, and assess relative performance of a light-, medium- and resource-intensive approach to mortality measurement. In the statistical component, we are applying innovative use of established causal and design-based methods to assess biases. We are testing the utility and feasibility of these methods to answer questions like: to what extent are hard to capture deaths, such as neonatal and violent deaths, systematically missed; can fewer survey clusters still provide estimates precise enough for decision making; and can analytical adjustments be used to address known biases? We are also supporting localisation by building field-ready guidance tools designed to make mortality estimation more accessible to operational actors. These tools include an algorithm to help teams choose a method, an operational readiness checklist, and a guide to data validation, triangulation, interpretation. Our aim is to make mortality estimation practicable in even the most challenging settings, without compromising quality. As the best-placed actors to assess mortality, we hope to pilot the guidance with national actors in the DRC and elsewhere to ensure it is user-friendly, actionable, and scalable for use in any crisis. Looking Ahead: Making Mortality Count Without credible mortality data, humanitarian response risks being inefficient, inequitable, and disconnected from reality. We cannot respond appropriately to crises we do not understand. When those with the greatest capacity to measure mortality have the least stake in the results, the system fails. The best way to ensure efficiency and effectiveness is to place local organisations at the centre. Connecting local expertise with technical knowledge offers a path toward a fairer humanitarian sector, where the reality of a crisis is described by those living through it. *2017 estimate adjusted for inflation.
As the summer travel season starts to take off, FAA Administrator Bryan Bedford tells CBS News he has confidence in the system, despite hundreds of FAA facilities being run on decades-old technology.
The United Kingdom has big economic problems. Growth is largely stagnant, inefficient spending on bloated welfare programs is out of control, and taxation as a percentage of GDP is at the highest level since World War II. But things seem set to get even worse. Wes Streeting, a governing Labour Party parliamentarian and front-runner to […]
Twisha Sharma Death Updates: The case came under scrutiny as the Madhya Pradesh Police faced criticism for an inefficient probe into the case, while Giribala Singh was accused of exerting undue influence.
When it comes to AI models, size matters. Even though some artificial-intelligence experts warn that scaling up large language models (LLMs) is hitting diminishing performance returns, companies are still coming out with ever larger AI tools. Meta’s latest Llama release had a staggering 2 trillion parameters that define the model. As models grow in size, their capabilities increase. But so do the energy demands and the time it takes to run the models, which increases their carbon footprint. To mitigate these issues, people have turned to smaller, less capable models and using lower-precision numbers whenever possible for the model parameters. But there is another path that may retain a staggeringly large model’s high performance while reducing the time it takes to run an energy footprint. This approach involves befriending the zeros inside large AI models. For many models, most of the parameters—the weights and activations—are actually zero, or so close to zero that they could be treated as such without losing accuracy. This quality is known as sparsity. Sparsity offers a significant opportunity for computational savings: Instead of wasting time and energy adding or multiplying zeros, these calculations could simply be skipped; rather than storing lots of zeros in memory, one need only store the nonzero parameters. Unfortunately, today’s popular hardware, like multicore CPUs and GPUs, do not naturally take full advantage of sparsity. To fully leverage sparsity, researchers and engineers need to rethink and re-architect each piece of the design stack, including the hardware, low-level firmware, and application software. In our research group at Stanford University, we have developed the first (to our knowledge) piece of hardware that’s capable of calculating all kinds of sparse and traditional workloads efficiently. The energy savings varied widely over the workloads, but on average our chip consumed one-seventieth the energy of a CPU, and performed the computation on average eight times as fast. To do this, we had to engineer the hardware, low-level firmware, and software from the ground up to take advantage of sparsity. We hope this is just the beginning of hardware and model development that will allow for more energy-efficient AI. What is sparsity? Neural networks, and the data that feeds into them, are represented as arrays of numbers. These arrays can be one-dimensional (vectors), two-dimensional (matrices), or more (tensors). A sparse vector, matrix, or tensor has mostly zero elements. The level of sparsity varies, but when zeroes make up more than 50 percent of any type of array, it can stand to benefit from sparsity-specific computational methods. In contrast, an object that is not sparse—that is, it has few zeros compared with the total number of elements—is called dense. Sparsity can be naturally present, or it can be induced. For example, a social-network graph will be naturally sparse. Imagine a graph where each node (point) represents a person, and each edge (a line segment connecting the points) represents a friendship. Since most people are not friends with one another, a matrix representing all possible edges will be mostly zeros. Other popular applications of AI, such as other forms of graph learning and recommendation models, contain naturally occurring sparsity as well. Beyond naturally occurring sparsity, sparsity can also be induced within an AI model in several ways. Two years ago, a team at Cerebras showed that one can set up to 70 to 80 percent of parameters in an LLM to zero without losing any accuracy. Cerebras demonstrated these results specifically on Meta’s open-source Llama 7B model, but the ideas extend to other LLM models like ChatGPT and Claude. The case for sparsity Sparse computation’s efficiency stems from two fundamental properties: the ability to compress away zeros and the convenient mathematical properties of zeros. Both the algorithms used in sparse computation and the hardware dedicated to them leverage these two basic ideas. First, sparse data can be compressed, making it more memory efficient to store “sparsely”—that is, in something called a sparse data type. Compression also makes it more energy efficient to move data when dealing with large amounts of it. This is best understood by an example. Take a four-by-four matrix with three nonzero elements. Traditionally, this matrix would be stored in memory as is, taking up 16 spaces. This matrix can also be compressed into a sparse data type, getting rid of the zeros and saving only the nonzero elements. In our example, this results in 13 memory spaces as opposed to 16 for the dense, uncompressed version. These savings in memory increase with increased sparsity and matrix size. In addition to the actual data values, compressed data also requires metadata. The row and column locations of the nonzero elements also must be stored. This is usually thought of as a “fibertree”: The row labels containing nonzero elements are listed and linked to the column labels of the nonzero elements, which are then linked to the values stored in those elements. In memory, things get a bit more complicated still: The row and column labels for each nonzero value must be stored as well as the “segments” that indicate how many such labels to expect, so the metadata and data can be clearly delineated from one another. In a dense, noncompressed matrix data type, values can be accessed either one at a time or in parallel, and their locations can be calculated directly with a simple equation. However, accessing values in sparse, compressed data requires looking up the coordinates of the row index and using that information to “indirectly” look up the coordinates of the column index before finally reaching the value. Depending on the actual locations of the sparse data values, these indirect lookups can be extremely random, making the computation data-dependent and requiring the allocation of memory lookups on the fly. Second, two mathematical properties of zero let software and hardware skip a lot of computation. Multiplying any number by zero will result in a zero, so there’s no need to actually do the multiplication. Adding zero to any number will always return that number, so there’s no need to do the addition either. In matrix-vector multiplication, one of the most common operations in AI workloads, all computations except those involving two nonzero elements can simply be skipped. Take, for example, the four-by-four matrix from the previous example and a vector of four numbers. In dense computation, each element of the vector must be multiplied by the corresponding element in each row and then added together to compute the final vector. In this case, that would take 16 multiplication operations and 16 additions (or four accumulations). In sparse computation, only the nonzero elements of the vector need be considered. For each nonzero vector element, indirect lookup can be used to find any corresponding nonzero matrix element, and only those need to be multiplied and added. In the example shown here, only two multiplication steps will be performed, instead of 16. The trouble with GPUs and CPUs Unfortunately, modern hardware is not well suited to accelerating sparse computation. For example, say we want to perform a matrix-vector multiplication. In the simplest case, in a single CPU core, each element in the vector would be multiplied sequentially and then written to memory. This is slow, because we can do only one multiplication at a time. So instead people use CPUs with vector support or GPUs. With this hardware, all elements would be multiplied in parallel, greatly speeding up the application. Now, imagine that both the matrix and vector contain extremely sparse data. The vectorized CPU and GPU would spend most of their efforts multiplying by zero, performing completely ineffectual computations. Newer generations of GPUs are capable of taking some advantage of sparsity in their hardware, but only a particular kind, called structured sparsity. Structured sparsity assumes that two out of every four adjacent parameters are zero. However, some models benefit more from unstructured sparsity—the ability for any parameter (weight or activation) to be zero and compressed away, regardless of where it is and what it is adjacent to. GPUs can run unstructured sparse computation in software, for example, through the use of the cuSparse GPU library. However, the support for sparse computations is often limited, and the GPU hardware gets underutilized, wasting energy-intensive computations on overhead. Petra Péterffy When doing sparse computations in software, modern CPUs may be a better alternative to GPU computation, because they are designed to be more flexible. Yet, sparse computations on the CPU are often bottlenecked by the indirect lookups used to find nonzero data. CPUs are designed to “prefetch” data based on what they expect they’ll need from memory, but for randomly sparse data, that process often fails to pull in the right stuff from memory. When that happens, the CPU must waste cycles calling for the right data. Apple was the first to speed up these indirect lookups by supporting a method called an array-of-pointers access pattern in the prefetcher of their A14 and M1 chips. Although innovations in prefetching make Apple CPUs more competitive for sparse computation, CPU architectures still have fundamental overheads that a dedicated sparse computing architecture would not, because they need to handle general-purpose computation. Other companies have been developing hardware that accelerates sparse machine learning as well. These include Cerebras’s Wafer Scale Engine and Meta’s Training and Inference Accelerator (MTIA). The Wafer Scale Engine, and its corresponding sparse programming framework, have shown incredibly sparse results of up to 70 percent sparsity on LLMs. However, the company’s hardware and software solutions support only weight sparsity, not activation sparsity, which is important for many applications. The second version of the MTIA claims a sevenfold sparse compute performance boost over the MTIA v1. However, the only publicly available information regarding sparsity support in the MTIA v2 is for matrix multiplication, not for vectors or tensors. Although matrix multiplications take up the majority of computation time in most modern ML models, it’s important to have sparsity support for other parts of the process. To avoid switching back and forth between sparse and dense data types, all of the operations should be sparse. Onyx Instead of these halfway solutions, our team at Stanford has developed a hardware accelerator, Onyx, that can take advantage of sparsity from the ground up, whether it’s structured or unstructured. Onyx is the first programmable accelerator to support both sparse and dense computation; it’s capable of accelerating key operations in both domains. To understand Onyx, it is useful to know what a coarse-grained reconfigurable array (CGRA) is and how it compares with more familiar hardware, like CPUs and field-programmable gate arrays (FPGAs). CPUs, CGRAs, and FPGAs represent a trade-off between efficiency and flexibility. Each individual logic unit of a CPU is designed for a specific function that it performs efficiently. On the other hand, since each individual bit of an FPGA is configurable, these arrays are extremely flexible, but very inefficient. The goal of CGRAs is to achieve the flexibility of FPGAs with the efficiency of CPUs. CGRAs are composed of efficient and configurable units, typically memory and compute, that are specialized for a particular application domain. This is the key benefit of this type of array: Programmers can reconfigure the internals of a CGRA at a high level, making it more efficient than an FPGA but more flexible than a CPU. The Onyx chip, built on a coarse-grained reconfigurable array (CGRA), is the first (to our knowledge) to support both sparse and dense computations. Olivia Hsu Onyx is composed of flexible, programmable processing element (PE) tiles and memory (MEM) tiles. The memory tiles store compressed matrices and other data formats. The processing element tiles operate on compressed matrices, eliminating all unnecessary and ineffectual computation. The Onyx compiler handles conversion from software instructions to CGRA configuration. First, the input expression—for instance, a sparse vector multiplication—is translated into a graph of abstract memory and compute nodes. In this example, there are memories for the input vectors and output vectors, a compute node for finding the intersection between nonzero elements, and a compute node for the multiplication. The compiler figures out how to map the abstract memory and compute nodes onto MEMs and PEs on the CGRA, and then how to route them together so that they can transfer data between them. Finally, the compiler produces the instruction set needed to configure the CGRA for the desired purpose. Since Onyx is programmable, engineers can map many different operations, such as vector-vector element multiplication, or the key tasks in AI, like matrix-vector or matrix-matrix multiplication, onto the accelerator. We evaluated the efficiency gains of our hardware by looking at the product of energy used and the time it took to compute, called the energy-delay product (EDP). This metric captures the trade-off of speed and energy. Minimizing just energy would lead to very slow devices, and minimizing speed would lead to high-area, high-power devices. Onyx achieves up to 565 times as much energy-delay product over CPUs (we used a 12-core Intel Xeon CPU) that utilize dedicated sparse libraries. Onyx can also be configured to accelerate regular, dense applications, similar to the way a GPU or TPU would. If the computation is sparse, Onyx is configured to use sparse primitives, and if the computation is dense, Onyx is reconfigured to take advantage of parallelism, similar to how GPUs function. This architecture is a step toward a single system that can accelerate both sparse and dense computations on the same silicon. Just as important, Onyx enables new algorithmic thinking. Sparse acceleration hardware will not only make AI more performance- and energy efficient but also enable researchers and engineers to explore new algorithms that have the potential to dramatically improve AI. The future with sparsity Our team is already working on next-generation chips built off of Onyx. Beyond matrix multiplication operations, machine learning models perform other types of math, like nonlinear layers, normalization, the softmax function, and more. We are adding support for the full range of computations on our next-gen accelerator and within the compiler. Since sparse machine learning models may have both sparse and dense layers, we are also working on integrating the dense and sparse accelerator architecture more efficiently on the chip, allowing for fast transformation between the different data types. We’re also looking at ways to manage memory constraints by breaking up the sparse data more effectively so we can run computations on several sparse accelerator chips. We are also working on systems that can predict the performance of accelerators such as ours, which will help in designing better hardware for sparse AI. Longer term, we’re interested in seeing whether high degrees of sparsity throughout AI computation will catch on with more model types, and whether sparse accelerators become adopted at a larger scale. Building the hardware to unstructured sparsity and optimally take advantage of zeros is just the beginning. With this hardware in hand, AI researchers and engineers will have the opportunity to explore new models and algorithms that leverage sparsity in novel and creative ways. We see this as a crucial research area for managing the ever-increasing runtime, costs, and environmental impact of AI.
Closing small boat stations has proven difficult. Leaving them unchanged is operationally inefficient. These units are enduring parts of the Coast Guard’s force structure, yet their full potential is not always realized. This article proposes a model to better align their mission with national priorities.During the recent Senate confirmation hearing for the next commandant of the U.S. Coast Guard, senators raised a wide range of global maritime concerns, including Arctic competition, cyber threats targeting ports, migration pressures, and increasingly severe storms. Yet the hearing repeatedly returned to a far more local subject: the Coast Guard units in senators’ backyards.Coast Guard The post Presence or Capacity? The Coast Guard Can Have Both Through Small Boat Stations appeared first on War on the Rocks.
MUMBAI: Liquidity risk is increasing for Indian-based real-estate developers, as non-bank financial institutions (NBFI; including housing finance companies) are shying away from lending to the sector, said Fitch Ratings.Developers that rely on refinancing from NBFIs, particularly those with weak financial profiles, will be affected the most should conditions persist. The availability of unencumbered assets among large developers may be of limited use, as NBFIs are looking to shed their already-high exposure to the sector, especially to large borrowers.NBFIs have disproportionately increased their share of real-estate sector credit in the previous few years, owing to heightened risk aversion by banks; banks have been cutting exposure due to their own funding challenges that began in late 2018, which have become more acute in the previous few months; domestic bank exposures fell to 2.3% of loans in the financial year ending March 2019 from 2.8% in 2015-16.NBFIs are now also shying away from refinancing maturing debt of even large, proven developers to limit concentration risk to the sector. This is pushing developers towards alternative funding channels, such as private equity. The availability of such funding could be more limited than the value of maturing debt and may only be available to established developers with sufficient unpledged assets. It would also come at a higher cost. We believe banks may still consider exposure to quality real estate, but overall exposure continues to decline.Developers that are focused on high-end projects may face higher risk, as sales of such projects have slowed in the last two years. We believe these developers would be wary of taking sharp price corrections on unsold inventory to boost sales, except in extreme circumstances, as this could diminish the value of unsold inventory and weaken collateral cover for existing lenders.In addition, any boost in sales would be temporary. Meanwhile, developers with substantial exposure to affordable housing may still benefit from marginal access to lenders in light of healthy pre-sales growth, supported by India's substantial housing deficit and government incentives for buyers via the credit-linked subsidy scheme as well as for developers, including tax deductions and grant of infrastructure status, which entitles companies to some benefits and concessions.The government has announced measures to improve NBFI-sector liquidity, but their efficacy remains to be seen. For example, we believe the government's July 2019 announcement to provide a first-loss guarantee of 10% on securitised assets issued by NBFIs to banks could ease funding pressure for NBFIs in the short term. However, the provision refers only to financially sound issuers and there is a lack of clarity about the duration of the guarantee and the definition of what comprises a 'financially sound' entity. In addition, most of the actions by the authorities to alleviate the liquidity squeeze will benefit the largest and least risky NBFIs and is unlikely to address the pressure on the more property focused players.Defaults by two NBFIs - Infrastructure Leasing & Financial Services Ltd (IL&FS) in September 2018 and Dewan Housing Finance Corporation Ltd (DHFL) in June 2019 - have contributed to the sector-wide liquidity squeeze, as investors have become more risk averse. Banks' low appetite for lending to real-estate developers is evidenced by the usually high risk weights attached to such loans. These are due to developers' typically low credit ratings amid high leverage, making exposure to the sector an inefficient use of banks' already-limited capital.Substantial bank recapitalisation to increase lending capacity could benefit NBFIs as well as real-estate developers, subject to the banks' risk appetite. Although a structural improvement in NBFI asset books would take time. Nonetheless, even under better conditions we expect NBFI's to tighten credit standards, with developers facing funding pressure until there is a broader improvement in their operations, with better end-user demand and pricing support.