Narrow Networks in Marketplace Plans as a Product of Information Salience

There has been a lot of coverage in the news over the last few months about the “narrow networks” of most health insurance plans available in the Marketplaces (AKA Exchanges) [see here, here, and here]. Most plans sold in the Marketplaces will only pay for your care if you use a provider who is a member of the plan’s network. Plans form networks by negotiating with various providers. If a provider is willing to accept a price that the plan offers, the provider is included in the network. If not, the provider is excluded. The carrot of being in a plan’s network (and getting the influx of patients enrolled in the plan) and the stick of being excluded from the network (and losing any patients enrolled in the plan) are how plans bargain down physician and hospital prices and save us all money.

It turns out that in many locations, most Marketplace plans exclude a large portion of doctors and hospitals from their networks. Some have decried this as evil insurers taking advantage of consumers and pocketing all of the extra profits they make by limiting consumer choice. Others have applauded this result, suggesting that insurers are finally competing on price and providing consumers with the low-cost health insurance they want.

I tend to fall in the latter camp, but I think there is an additional issue to consider: the salience of information provided to consumers. In the Marketplaces, plans are differentiated mostly by three factors: Price, cost-sharing, and network. Anyone who has visited one of the state Marketplace websites knows that both price and cost-sharing information are very salient. However, information about a plan’s network is much murkier. Sure, anyone can follow a series of links to an insurer’s (often unclear) website and type in the name of specific providers to determine whether they belong to the insurer’s network. But, for most consumers, especially those with limited experience interacting with specific providers, a list of every doctor in the insurer’s network obviously does not provide a clear picture of the quality of the network.

In economics, this lack of information about network quality is referred to as an information asymmetry. Information is asymmetric because insurers know more about the quality of their network than consumers.  In a famous paper that won him the Nobel Prize in Economics, George Akerlof showed that asymmetric information can produce market failures. In his paper, he used the example of used cars. The argument goes something like this: Suppose 50% of used cars are “lemons” and worth $100 and 50% of used cars are good and worth $1000. If consumers don’t know which used cars are good and which are lemons, they’ll only be willing to pay the average value of a lemon and a used car, $550. If that’s the case, then individuals selling good cars won’t be willing to sell because the value of the good car ($1000) is higher than the price the consumers are willing to pay ($550). Because of this, in the market only lemons are sold.

Now, what do lemons have to do with health insurance plan networks? Well, instead of lemons and good cars, in insurance markets we have wide and narrow networks. If consumers can’t tell the difference between the two when they purchase, they’ll only be willing to pay the average value of a wide and a narrow network. Because wide networks are worth more than this average price that consumers are willing to pay, insurers don’t sell wide networks. In the end, only narrow networks are offered. This is a classic market failure due to asymmetric information a la Akerlof, and it seems like a definite possibility that the Marketplaces are suffering from it (though other forms of asymmetric information like adverse selection are likely to be a factor as well, an issue for a future post).

So, if we have competitive insurance markets, are we just doomed to only be able to choose from narrow, lemon networks? Fortunately, there are solutions to such market failures. The most obvious one is to provide consumers with more information about plan networks. But how can we provide them with more information than a list of every doctor included in the network? This is where the issue of salience comes in. In order to fix this market failure, consumers need more salient information about network size and quality that they can see and easily understand and compare across plans.

How can we provide consumers with such simple and salient information? Well, one way is to copy the method used to provide information about plan cost-sharing in the Marketplaces. In the Marketplaces, plans are divided into one of four cost-sharing metal tiers: Bronze, Silver, Gold, and Platinum, with cost-sharing increasing with the value of the precious metal. Why not provide “network quality tiers” as well? All plans could be categorized as Bronze-Wide, or Platinum-Narrow, or Silver-Medium. Or we could use more flashy names like Awesome, Mediocre, and Lame to describe plan network quality.

Of course, any plan to provide consumers with information on network quality would require us to come up with good ways of measuring network quality. But that sounds like a topic for future research, and good job security for me and other health economists…

New paper on imperfect risk adjustment and sorting in Health Insurance Exchanges

I’ve posted the latest draft of my new paper on the effects of imperfect risk adjustment on prices and sorting in Health Insurance Exchanges (AKA Health Insurance Marketplaces). Go read it!

Health Insurance Premiums in the Exchanges pt. 3

This is the third post in a series I’m writing about premiums in the new health insurance exchanges.

So at the end of my last post, I suggested that it may seem like community rating is unfair to healthy individuals because they will see their premiums go up while sick individuals will see their premiums drop or may be able to get insurance where they otherwise wouldn’t have. I also alluded to the fact that this view of community rating is incomplete. Let’s get into why that’s the case.

The purpose of insurance is to reduce risk. Most of us don’t like uncertainty. When offered $50 or a gamble where 50% of the time we get $100 and 50% of the time we get $0, we’ll take the sure $50. In economics, we refer to this as risk aversion. Insurance increases our utility, or happiness, or satisfaction, or peace of mind, or whatever you want to call it, because it gets rid of uncertainty. For example, if I ask you how much health care you’ll consume next year, you probably have no idea. It could cost you anywhere from $0 to $100,000 or more depending how things turn out. We refer to the variation in the potential amounts you could end up spending next year as risk. If you are risk averse, you will be willing to pay an insurer a health insurance premium in order to get rid of that uncertainty. This will make you better off.

Insurance policies are structured the way they are to decrease the amount of risk you face next year. Most policies do a great job at reducing this risk. However, there is another type of risk that the short-term, annual health insurance contracts so prevalent in the United States don’t help with. To think about this type of risk, let’s start from a world where there is no community rating. This is how things were in many states pre-Obamacare.

When you buy an insurance policy in this world, if you get sick, you’re protected from any large expenditures during the year your contract with the insurer is good for. This is great. It protects you from the uncertainty in your health status during the contract year. But, at the end of that year, you are no longer protected. For example, if you end up getting diagnosed with cancer during the year, you’ll be protected from any costs related to that diagnosis during that year. But, once the year is over, your insurer will likely raise the price of your policy because your expected cost for the next year is now much higher. This is the second type of risk. It is the risk that your health status will change during the year and you will now be charged a higher premium by your insurer for future insurance contracts. Economists refer to this as “reclassification risk.” Your annual insurance policy does nothing to protect you from this type of risk.

Now, in a world where there is community rating, reclassification risk does not exist. This is because your premium cannot depend on your health status. If individuals are risk averse (as most of us are), reducing this risk results in welfare gains, or increases in utility/happiness/whatever you want to call it. So basically, while it may seem like community rating is unfair to healthy people, really, it provides healthy people with a benefit that they probably don’t recognize: Insurance against reclassification risk or getting charged a higher premium in the future because of changes in health status. It is important to recognize that this insurance against reclassification risk benefits individuals whether they end up getting sick or not. This is because individuals don’t like uncertainty, and community rating reduces uncertainty.

So while community rating results in higher premiums for the healthy, it also provides additional insurance benefits. So it clearly benefits the sick who pay lower premiums or who can get insurance now where they couldn’t before, and now it also may provide net benefits to the healthy who pay higher premiums, but now have the additional benefit of being protected from reclassification risk. At this point it is in no way clear that this policy even hurts the healthy, so it is definitely not clear that community rating is unfair to the healthy.

While community rating provides additional insurance benefits, it comes with some costs. I’ll discuss these costs in the next couple of posts. They are very important and could be quite large. As a preview, these costs come in the form of what economists call adverse selection. In the final post, I’ll summarize an excellent new paper that estimates the tradeoffs of these costs and benefits of community rating and provides some implications for policy. Stay tuned!


P.S. Some economists have suggested that another way to protect against reclassification risk is to have longer term health insurance contracts. This is definitely a potential solution to this problem. If the contract length is long enough, reclassification risk is eliminated. This is because your premium never changes. There are a couple of practical issues with this proposal, however. First, because the progress in medical technology is unpredictable and quite rapid, it would be very difficult to set the premium in a way that would provide adequate coverage over the long term. For example if the premium for the long term policy was set at $3000 per year today, but then there was a rapid improvement in technology that resulted in the average cost for an individual being $10000 per year, the premium would have to change. It’s not clear how this would work, especially in the very long term, like lifetime policies. Second, in order for this to work right, an individual would have to be able to make a decision regarding their insurance coverage before anything is revealed about her health status. This means before conception. This is clearly not feasible. There may be other ways to implement this sort of policy and it may in reality be the optimal way to go. However, the current institutional and political realities really preclude this possibility, so I don’t discuss it much here.

Health Insurance Premiums in the Exchanges Pt. 2

This is the second post in a series I’m writing about premiums in the new health insurance exchanges.

Community Rating

To understand why community rating may increase the price of your health insurance policy, you first need to understand how insurers set prices.Let’s start in a market without community rating. If the market is competitive, insurers will set the price of a health insurance policy equal to the expected cost to the insurer of insuring you. The bulk of this expected cost is made up of your expected medical care costs. In addition to these costs, there are other administrative costs, but we’ll ignore these for now because they are probably the same for everyone.

The important thing to note here is that the expected cost to the insurer of enrolling you is likely quite different from the expected cost to the insurer of enrolling your friend who is recovering from cancer. Why is this the case? Well, first let’s talk about what expected cost means. Expected cost is the insurer’s best prediction about how much you would cost them next year. In order to calculate your expected cost, the insurer will use all of the information it has about you. They may get information from prior health insurance claims or from survey questions they ask you to answer before enrolling. Basically, they’ll take a very large group of individuals and divide them into smaller groups based on the information they have about these individuals’ health during year 1. For each small group, they’ll find the average cost in year 2. This value is the expected cost for each member of the small group.

Think for a moment about the small group you might be in. Do you have diabetes? Do you suffer from asthma? Are you a married woman of child-bearing age? Did your mother die of cancer? The answers to these questions will help isolate you in a small group of similar individuals.

Once the insurer calculates your expected cost, he will then set your price for his insurance policy equal to that expected cost. This means that individuals in healthy small groups will have lower prices than individuals in sick small groups. Individuals with diabetes will pay a much higher price than individuals without diabetes.

This is the way things work in an insurance market without community rating. In a market with community rating, everyone pays the same price, no matter what the expected cost of their small group is. That price is now the average expected cost for the entire population.

How much does this matter? If individuals’ expected costs are pretty similar, then there isn’t much difference between a community rated system and a system without community rating. This is also true if insurers can’t really differentiate between groups. For example, if insurers can’t tell the difference between a diabetic and a non-diabetic, they may assign all of these individuals to the same small group and assign them all the same price, even without community rating.

So how much variation is there in expected costs? The short answer is A LOT. The other short answer is the following picture:

cost distn

The picture shows expected health care costs for a sample of individuals likely to enroll in the exchanges from the Medical Expenditure Panel Survey. The expected costs are calculated using a fairly simple prediction algorithm incorporating demographic information along with information about diagnoses from the prior year (for those who are more knowledgeable in this area, the prediction model is a modifed version of the CMS-HCC risk adjustment model). Because the prediction model is fairly simple, the picture shows an underestimate of the variation in expected costs. Insurers can do much better than what I’m showing you here.

There are a lot of individuals clumped to the left side of the distribution because there is not much information about them. Their predicted costs are just the average costs for people their age. The individuals spread out to the right have chronic conditions or other predictable costs. The important thing about this figure is that there is quite a bit of variation. Without community rating, this figure can also be interpreted as showing the variation in annual premiums insurers will charge individuals. Unlucky individuals to the right will have very high premiums. Some individuals that aren’t shown in the picture to make it more readable will have premiums higher than $50,000 per year. Insurers may choose not to even insure them. Others will have very low premiums.

Community rating dramatically changes this picture. Under community rating, there is no premium variation, so premiums are now equal to the average expected costs in the population. In this population, the average is around $3,100. This will cause premiums for individuals with low expected costs to rise and premiums for people with high expected costs to drop. This picture shows changes in premiums due to community rating.

premium changes

Individuals are grouped based on their expected costs and groups with low expected costs are on the left and groups with high expected costs are on the left. It is clear that premiums go up a lot for people with low expected costs but come down a lot for individuals with high expected costs. This is really the purpose of community rating.

This may seem unfair to the healthy people with low expected costs. In the next post, I’ll talk about why that conclusion may not be completely accurate. As a preview, this is because people don’t stay healthy or sick forever. Instead their health status tends to fluctuate meaning that some years, you’ll be in the red bars and some years you’ll be in the blue. This means that community rating effectively provides insurance against changes in health status.

I’ll then follow up with posts on the negative effects of community rating.

Health Insurance Premiums in the Exchanges Pt. 1

There’s been a lot of talk lately about the premiums in the new health insurance exchanges. Some people (like me) feel like the prices are pretty encouraging. Others believe them to be very high and are pretty mad about it, especially if they were previously purchasing insurance in the individual market at lower rates. There seems to be a lot of confusion as to where these new, sometimes higher, prices are coming from. I plan to discuss the sources of price changes in a series of posts. I’m going to attempt to make the posts very accessible, so please let me know if there’s something that’s confusing so I can make it clearer.

I note from the beginning that this entire discussion refers to insurance premiums in the individual market for private health insurance. If you get insurance from your employer and your premiums are rising, this discussion does not apply to you. Your premium increases are also likely to be due to forces other than Obamacare.

Community Rating vs. Additional Benefits

There are two reasons why an individual’s insurance premiums might be rising this year. First, your plan may be changing. More likely than not, your plan is becoming more comprehensive. This is to comply with new regulations requiring plans to have an actuarial value greater than or equal to 60%. A plan’s actuarial value is the portion of health care costs the plan will cover for an average individual. Most individual plans currently fall short of the 60% requirement, so people currently holding individual policies will probably see their plans change to qualify, raising the price.1 This price increase does not come without corresponding benefits, however. Not many people are complaining about having to purchase more comprehensive coverage which they probably couldn’t get last year.

There are also some additional services that Obamacare requires insurers to cover, like coverage for contraceptives and breast pumps for new mothers. These will cause premiums to increase some, but probably not very much.

Second, new regulations require health insurance premiums to be “community rated.” No, this does not mean that everyone in your community has to go somewhere and rate every health plan. It means that plans can’t charge different people different premiums for the same plan. Now, the law does not require perfect community rating. Instead, it allows health plans to vary premiums by age, geography, and smoking status. Really, it just prohibits charging different premiums based on health status, i.e. a health plan can’t charge a sick person more than a healthy person for coverage via the same health plan.

Community rating is the real reason your premiums are increasing. It is also one of the most popular provisions in the health care law. But is it a good policy? I’ll spend the next few posts discussing this policy and its costs and benefits. I hope that by the end you can decide for yourself with full information.

References (gated)

Age-based “Rate Bands” for Health Insurance Premiums

There has been a lot of discussion lately about the premium rating restrictions in the ACA. I want to put some of this discussion in the context of the full set of regulations in the ACA because I feel that most of the discussion focuses on one particular rule and neglects to recognize that the rule interacts with other new policies in ways that could change the conclusions people are making about the potential effects of that policy. First, I’ll give some quick background for anyone who just wants to understand what the new laws about premium setting mean. Then, I’ll try to give a picture of how the rate setting policy interacts with other policies such as risk adjustment which will also go into effect around the country next January.


Currently in most states insurers selling plans on the individual market (i.e. not employer-provided insurance) can charge different people different premiums for the same coverage. They vary these premiums based on expected cost. I’ll give two examples. First, let’s say there’s a pair of twin sisters, both age 35. One of the sisters has asthma, while the other is perfectly healthy. If both sisters attempt to purchase a plan from Blue Cross, Blue Cross is currently able to charge the sisters different premiums for the same plan, and they often do (sometimes they outright reject the sister with asthma). Second, let’s say there are two healthy brothers who were born 20 years apart. Tim is 35 and Tom is 55. Neither brother has any chronic conditions. Again, let’s say both brothers attempt to purchase the same plan from Blue Cross. Again, Blue Cross is able to (and does) charge different premiums for the two brothers based on the expected costs of a healthy 35 year old vs. the expected costs of a healthy 55 year old.

The ACA changes the rules regarding premium setting in the individual market. Charging different premiums based on health (example 1) will be prohibited starting next January. Insurers will also be prohibited from rejecting anyone who wants to purchase coverage. Therefore, our twin sisters will be charged the same premium for the same coverage. The premium paid by the sister with Asthma will likely go down and the premium paid by the healthy sister will likely go up. Charging premiums based on age, however, will still be allowed, though limited to a ratio of 3:1. This means that, as long as the ratio of Tom’s premium to Tim’s premium is less than 3:1, the insurer can continue to charge them the same premiums it was charging before.

Much of the discussion is about the new 3:1 age rate band. Insurers are warning that young people will experience “rate shock” when the law goes into effect because the insurers will have to raise their premiums on the young by a lot to comply with the 3:1 rate band. This is probably true. If we assume perfect competition, premiums are equal to average cost. According to calculations I’ve done using the Medical Expenditure Panel Survey (MEPS) the average health care costs for a 64 year old likely to be purchasing a policy on one of the new state health insurance exchanges are about 9 times the costs of a 21 year old. According to this metric, rate shock will be rather extreme for the young. Insurers argue that due to this rate band policy, the young will not purchase insurance and instead pay the penalty for being uninsured. This will induce adverse selection and cause major problems for the exchanges.

Policy Interactions

I don’t argue that this “rate shock” won’t occur. It probably will, though not to the extent the insurers claim it will because of several factors, mainly that the old and the young are likely to purchase different policies. The young will purchase bronze and silver plans and the old will purchase gold and platinum plans. If only healthy old people buy the lower coverage plans, the age rate bands are unlikely to bind, and the young will still pay similar premiums. However, if this age-based segmentation occurs, adverse selection into the comprehensive plans is likely to be extreme, causing large welfare losses.

This brings me to the real purpose of this post. There is another very important policy in the ACA that deals with this adverse selection problem: Risk adjustment. To give a simplified explanation, risk adjustment is a policy where an insurer sets its own premiums, but a regulator (the exchange) collects those premiums and pools all of the premiums for all of the plans in the exchange into one giant pot. The regulator then reallocates the premiums based on the premium the plan set and on the expected cost of individuals in the plan. The expected cost calculation is based on age, gender, and past health insurance claims. For example, it would predict that a 55 year old male with diabetes would cost much more than a 55 year old male without diabetes. It would also predict that a healthy 55 year old female would cost much more than a healthy 35 year old female. This policy attempts to remove the incentives for plans to select cheap enrollees. It makes all individuals equally attractive to plans (at least based on health status) so that the plans don’t attempt to induce selection using costly methods or by inefficiently rationing services that high-risk individuals demand, such as access to diabetes or mental health specialists (note that if it is efficient to ration access to diabetes specialists, then risk adjustment does not provide additional incentives for plans to inefficiently provide access; if implemented correctly, it just gets plans closer to the efficient allocation of services; see Frank, Glazer, and McGuire 2000, Glazer and McGuire 2000, and Glazer and McGuire 2002). It also causes risk pooling between plans to limit the effects of the age-based segmentation discussed above.

Herein lies the issue with the insurers’ complaints about the age-based rate band policy. My point is a technical one, but an important one. Rate shock is a big worry, but not because of the age-based premium rating restrictions; rather, rate shock is a worry because of risk adjustment. The risk adjustment models that will be used by HHS to predict cost include age as a predictor of cost. Therefore, the risk adjusting of premiums described above will make a 35 year old and a 55 year old look pretty similar in terms of cost. In a competitive environment, this would mean that a plan would charge them the same premium. In simulations using MEPS data and the CMS-HCC risk adjustment model, McGuire et al. (2012) show that assuming perfect competition, after implementing even partial risk adjustment, the 3:1 rate band is not binding. In other words, plans would charge a 55 year old and a 25 year old premiums that vary by less than 3:1. In simulations I’ve done for another project, if premiums are fully risk adjusted, the premiums plans charge are virtually identical. This means that the premiums charged to the young will spike, but not because of the rate bands. Rather, they’ll spike because of risk adjustment.

All of the policy discussion going on is about the effects of the rate bands and how they will cause extreme adverse selection on the exchanges. Unfortunately, nobody is discussing the real driver of this rate shock and selection: risk adjustment. This is probably because the purpose of risk adjustment is to limit selection, not to expand it. Unfortunately, that only works if mandates work. Since the ACA’s mandate may not be large enough to get the young to buy insurance, it could instead cause more selection due to premium compression. There is a fix to this, though a complex one: Take into account the tradeoff between the gains from risk adjustment (due to improved selection incentives to health plans) and the losses from the premium compression it causes (due to the young dropping out and driving up the costs for those remaining in the exchange; i.e. adverse selection) and use partial risk adjustment instead of full. The optimal amount of risk adjustment that should be used will depend on the demand curves of the young, and it is likely to be far less than 100%. Again, unfortunately nobody is talking about this. Instead, the focus is on rate bands that are unlikely to bind after risk adjustment.

All of this has assumed perfect competition between health plans. In my next post, I’ll talk about what happens when we introduce imperfect competition (which is probably more realistic) and provide an argument for why the insurers may really be arguing for the elimination of the rate bands. Hint: it’s not because of rate shock on the young.


McGuire, Thomas et al. 2012. Integrating Risk Adjustment and Enrollee Premiums in Health Plan Payment

Glazer, Jacob and Thomas McGuire. 2000. Optimal Risk Adjustment. American Economic Review

Frank, Richard, Jacob Glazer, and Thomas McGuire. 2000. Measuring Adverse Selection in Managed Care. Journal of Health Economics

Glazer, Jacob and Thomas McGuire. 2002. Setting Plan Premiums to Ensure Efficient Quality in Health Care: Minimum Variance Optimal Risk Adjustment. Journal of Public Economics

Service-level Selection in Health Insurance Markets

I came across this NY Times article this morning about Chiropractors and Acupuncturists lobbying to be part of the Essential Benefits Package in various states across the country. Being part of this package means that in order to meet the minimum credible coverage requirements in a state, a plan must cover these services. I found it a little odd that these two groups in particular are aggressively lobbying to be covered. You may say, “Of course these groups would want to be covered. They’re often not covered, and of course they would want to use crony capitalism to boost their business.” However, this situation is slightly more complicated.

The essential health benefits requirement is mostly for the individual and small business markets (both of which will move to the new state exchanges in 2014). The purpose is to make sure that healthy people and sick people pool their risks. If people were able to buy policies that were essentially the same as being uninsured, many healthy people would flock to those policies, leaving the sick in the more comprehensive policies, driving up prices, and making the comprehensive insurance less “insurance” against the costs of getting sick per se and more membership fee equal to the average cost of being sick, the classic adverse selection problem. I leave alone whether that argument is a good enough one to justify the essential health benefits provision, which is clearly open to corruption, for now and focus on a different aspect of this situation.

In an exchange, insurers will be required to charge everyone essentially the same premium, whether they’re sick or healthy (there is some latitude for premium variations based on age and smoking status, but they’re restricted). Because of this, sick people will be seen as potential losses to the insurers and healthy people will be seen as potential profits. Naturally, insurers will want to do anything they can to get the healthy folks. There are several ways to do this: Cream-skimming, dumping, or service-level selection. The last of these, service-level selection, is the one I think is most interesting in this situation.

Service-level selection is practiced by managed care organizations (MCOs) when individuals are choosing plans in a market where there are many plans competing for their business. It occurs when the MCOs ration care demanded by high cost individuals, things like diabetes specialists or flagship hospitals. They can do this by increasing cost-sharing for these services or by removing these types of providers from their networks. When a diabetes patient goes to an exchange to purchase coverage, they’re not likely to pick a plan that doesn’t have their (or any good) diabetes specialists. This sends the sick patients to more comprehensive plans for their coverage, driving up prices in those plans, where only the sick go. This is the same classic adverse selection problem (see Frank, Glazer, and McGuire 2000 and Ellis and McGuire 2007).

There are two sides to service-level selection. A plan can restrict access to services demanded by high-cost patients, as described above, or it can loosen access to services demanded by low-cost patients to cause this selection problem. This is where the Times article comes in. It turns out that Chiropractic and Acupuncture services are at the top of the list for services demanded by low-cost patients. Ellis and McGuire (2007) have shown that these are the most attractive services for MCOs in a free market to offer, and Ellis, Jiang, and Kuo (2012) have shown that these services are offered more often by MCOs than non-managed plans. So, it is likely that the free choice available through the exchanges will lead to these services being offered by the MCOs even without a mandate from the state, unless, of course, the risk adjustment schemes in the exchanges (a topic for another post) are good enough that the sick and the healthy are equally attractive to the plans.

So, my point here is that the lobbying of these groups may not be necessary. The groups that should be lobbying for mandated coverage are those that are demanded by the sick, high-cost patients: Diabetes care providers, hospice care, etc. My guess is that the acupuncturists and chiropractors don’t realize this and that they want mandated coverage because they know that individual’s are more price sensitive when it comes to their services than other medical services. I guess we’ll see what happens in the states where acupuncture isn’t mandated compared to the ones where it is. I doubt there will be much difference.


Get every new post delivered to your Inbox.